Agentic AI capability directory
Find the discipline behind your AI-agent decision.
Explore 363 specialist areas across 19 capability families. The directory covers AI agents and their wider operating systems: models, tools, memory, identity, controls, workflows and accountable human decisions.
Use the right scope for the question
Consider a customer refund. AI can help someone understand the request; an AI agent can also work through the steps needed to handle it.
-
Understand information
AI generates an answer
AI can summarise documents, draft text or make predictions. A person decides what to do with the result.
Example: summarise a customer’s refund request.
-
Use tools and take steps
An AI agent works through a task
An AI agent can look up information, choose a next step and use connected software to carry out a task.
Example: check the order, look up the refund policy and prepare a refund.
-
Decide what it may do
Your business sets the controls
Permissions, approval steps and records of actions determine how the agent works with your people and software.
Example: require a manager’s approval before issuing the refund, then record what happened.
We help you choose tasks for AI agents, set limits on their actions and test how they handle real work before expanding their use.
Specialist areas
363 specialist areas
Try a broader question.
Use a shorter search or remove a capability filter to see more specialist areas.
- Agentic AI strategy and transformation
Agentic AI opportunity discovery
Identify a business workflow worth investigating for agentic change. The review connects process observations, exception logs, staff interviews and existing performance measures to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
Enterprise agent strategy
Choose where agentic AI belongs in the enterprise's operating priorities. The review connects business objectives, current initiatives, architecture constraints and risk appetite statements to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
Agentic operating-model design
Assign accountability for work shared by people and agents. The review connects process ownership, decision rights, service responsibilities and escalation paths to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
Workflow and process redesign
Redesign the workflow around its desired business outcome. The review connects process maps, handoff delays, exception categories and acceptance criteria to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
AI readiness and maturity assessment
Understand which prerequisites constrain a credible AI initiative. The review connects data access, process stability, skills, ownership and current control evidence to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
Agent portfolio prioritisation
Allocate attention across competing agent initiatives. The review connects initiative hypotheses, evidence strength, dependencies and capacity estimates to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
Build-versus-buy-versus-partner analysis
Choose an implementation route supported by evidence and constraints. The review connects requirements, internal capacity, vendor capabilities and ownership constraints to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
Agentic transformation roadmap
Sequence transformation work around dependencies and decision gates. The review connects initiative dependencies, capacity assumptions, architecture milestones and learning objectives to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
AI investment and business-case development
Develop an investment case that exposes assumptions instead of manufacturing certainty. The review connects baseline workload, quality, review effort, failure costs and investment assumptions to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
Workforce and role-impact analysis
Understand how agent adoption changes human work and responsibility. The review connects task inventories, work observations, exception demand and role expectations to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
Human–agent collaboration design
Design effective handoffs between people and agents. The review connects handoff screens, reviewer needs, decision authority and queue behaviour to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
Agent adoption and change management
Help intended users adopt an agent through evidence and usable work practices. The review connects user research, workflow fit, training materials and support feedback to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
Agentic center-of-excellence design
Define what a central agent capability team should own. The review connects existing team mandates, common needs, local constraints and demand for support to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
Enterprise AI capability building
Build the organisational capability needed for a defined agent workload. The review connects role competencies, project demands, existing artefacts and practice opportunities to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
Transformation programme assurance
Challenge whether a transformation programme has evidence for its next commitment. The review connects program objectives, dependency status, risk logs and benefit evidence to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
Agentic AI vendor strategy
Shape vendor choices around the enterprise's long-term operating needs. The review connects vendor requirements, contractual dependencies, integration needs and current alternatives to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
Agentic product strategy
Choose the customer problem and product role an agent should serve. The review connects customer problem evidence, current journeys, product economics and support constraints to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic AI strategy and transformation
AI transformation measurement
Measure transformation through useful completed work and realized outcomes. The review connects baselines, adoption records, quality measures, review effort and business outcomes to a bounded review of transformation decisions, giving the business sponsor a clearer basis for the next action.
- Agentic transformation partnerships
Fractional Head of Agentic AI
When agent initiatives need leadership but ownership is fragmented, this specialty examines priority ownership and escalation access. Useful outputs can include leadership mandate draft and transition criteria to support the customer's next decision about what authority and accountability a fractional leadership role should hold.
- Agentic transformation partnerships
Agentic transformation office
When multiple agent initiatives compete for people, evidence, and attention, this specialty examines cross-program sequencing and benefit evidence ownership. Useful outputs can include transformation operating model and review cadence proposal to support the customer's next decision.
- Agentic transformation partnerships
Executive advisory retainer
When leadership needs recurring challenge across changing choices, this specialty examines decision preparation and evidence challenge. Useful outputs can include advisory mandate and action and revisit protocol to support the customer's next decision about what questions an advisory retainer should answer and how advice becomes action.
- Agentic transformation partnerships
Embedded agent engineering pod
When an agent program needs a coherent temporary engineering team, this specialty examines work acceptance and architectural alignment. Useful outputs can include pod work charter and capability handover outline to support the customer's next decision about whether an embedded pod has bounded ownership and a sustainable handover.
- Agentic transformation partnerships
AI assurance partner
When agent releases need a continuing evidence and challenge function, this specialty examines scope continuity and evidence freshness. Useful outputs can include assurance operating proposal and change-triggered refresh plan to support the customer's next decision about how an assurance collaboration should support accountable release decisions.
- Agentic transformation partnerships
Independent transformation reviewer
When a major agent program needs challenge from outside its delivery chain, this specialty examines conflict disclosure and access to adverse evidence. Useful outputs can include review mandate and decision challenge brief to support the customer's next decision.
- Agentic transformation partnerships
AI architecture council support
When agent designs repeatedly need cross-functional architectural decisions, this specialty examines review completeness and reusable decisions. Useful outputs can include council decision pack and reusable pattern register to support the customer's next decision about how council support can make design review more consistent.
- Agentic transformation partnerships
Agent governance office support
When agent oversight is increasing the load on existing control teams, this specialty examines intake consistency and evidence follow-up. Useful outputs can include governance support model and evidence follow-up protocol to support the customer's next decision.
- Agentic transformation partnerships
Enterprise pilot factory
When many business teams want to pilot agents, this specialty examines pilot scope discipline and outcome measurement. Useful outputs can include pilot operating design and transition decision template to support the customer's next decision about how to run bounded pilots that produce comparable learning and stop decisions.
- Agentic transformation partnerships
Agent evaluation centre of excellence
When evaluation knowledge is scattered across product teams, this specialty examines method reuse and reviewer calibration. Useful outputs can include evaluation operating model and adoption and calibration plan to support the customer's next decision about what a centre of excellence should own and what teams should retain.
- Agentic transformation partnerships
Secure-agent launch program
When multiple controls must be coordinated before agent authority expands, this specialty examines control implementation ownership and evidence review. Useful outputs can include launch program plan and readiness decision cadence to support the customer's next decision.
- Agentic transformation partnerships
AI capability transfer program
When a team depends on external help for agent development or evaluation, this specialty examines observed task performance and teach-back quality. Useful outputs can include capability transfer plan and ownership transition criteria to support the customer's next decision.
- Agentic transformation partnerships
Vendor-neutral technology advisor
When agent platform choices need a product-neutral decision process, this specialty examines requirement fit and implementation assumptions. Useful outputs can include technology decision framework and qualification questions to support the customer's next decision about which technology option best fits the customer's requirements and constraints.
- Agentic transformation partnerships
Portfolio evaluation partner
When leadership must choose which agent initiatives to advance or stop, this specialty examines comparable decision criteria and dependency concentration. Useful outputs can include portfolio evaluation rubric and gate recommendation brief to support the customer's next decision.
- Agentic transformation partnerships
Transformation recovery engagement
When an agent program is stalled, disputed, or missing outcomes, this specialty examines root cause separation and recoverable scope. Useful outputs can include recovery diagnostic and restart or stop conditions to support the customer's next decision.
- Agentic transformation partnerships
Post-launch optimisation partnership
When an agent is live but its quality, cost, or adoption needs attention, this specialty examines quality-adjusted outcomes and experiment isolation. Useful outputs can include optimisation backlog and change review cadence to support the customer's next decision.
- Leadership and organisational training
Agentic AI for boards
This area examines a board-level decision about investment, accountability and limits of delegated action. It helps decision owners identify the evidence needed when management seeks approval for an AI program that would expand automated decision-making, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
Agentic AI for executive leadership
This area examines a leadership decision linking a business objective to accountable operating ownership. It helps decision owners identify the evidence needed when different functions sponsor agent initiatives without a shared view of value, control or responsibility, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
AI risk for audit committees
This area examines a committee oversight question about risk reporting, control evidence and assurance coverage. It helps decision owners identify the evidence needed when AI risk reports describe policies and traffic-light ratings without clear evidence of control operation, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
AI governance for senior management
This area examines a management decision at intake, release, change, exception or retirement. It helps decision owners identify the evidence needed when an AI policy has been adopted but teams disagree about approvals, exceptions and escalation, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
Secure AI for CISOs
This area examines a security decision about a defined agent workflow, authority boundary and response capability. It helps decision owners identify the evidence needed when agents are moving from generated answers to authenticated tool use and consequential actions, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
Agent architecture for CTOs
This area examines a system design choice involving orchestration, identity, memory, tools and recovery. It helps decision owners identify the evidence needed when a prototype must become an operable system with tools, state and human oversight, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
AI product strategy for CPOs
This area examines a product hypothesis connecting user intent, delegated actions and an observable outcome. It helps decision owners identify the evidence needed when an agent feature is attractive in demonstrations but the customer decision and acceptable behaviour remain vague, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
AI privacy for legal and privacy leaders
This area examines a data-processing question across an agent's inputs, retained state, providers and outputs. It helps decision owners identify the evidence needed when agent proposals describe model vendors but omit memory, tool, logging and support data flows, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
AI procurement for sourcing teams
This area examines a sourcing decision tied to a defined workflow, operating constraint and supplier dependency. It helps decision owners identify the evidence needed when AI vendors present similar feature claims while operational and exit requirements remain difficult to compare, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
AI assurance for internal audit
This area examines a proposed audit objective with criteria, scope, evidence and responsible reviewers. It helps decision owners identify the evidence needed when an audit plan includes AI without a clear audit object, evidence route or competence boundary, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
Agent evaluation for engineering leaders
This area examines a release requirement traced to test cases, observed behaviour and reviewer disposition. It helps decision owners identify the evidence needed when the team has many tests and traces but no shared basis for deciding whether an agent change is acceptable, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
AI transformation for business-unit leaders
This area examines a business-unit workflow with accountable outcomes, human roles and adoption constraints. It helps decision owners identify the evidence needed when an AI rollout is planned before the unit has defined workflow changes, exceptions and value measurement, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
Human–agent operating-model workshops
This area examines a handoff between a person and an agent within a defined business process. It helps decision owners identify the evidence needed when an agent workflow contains human review points whose responsibilities are unclear in practice, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
AI incident simulation for executives
This area examines a simulated executive decision during a staged agent incident. It helps decision owners identify the evidence needed when a consequential agent incident would require business, technical and communications decisions under uncertainty, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
Agentic AI tabletop exercises
This area examines a facilitated tabletop sequence across a particular workflow and participating roles. It helps decision owners identify the evidence needed when multiple teams own parts of an agent workflow but have not practiced a shared exception, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
Responsible autonomy decision workshop
This area examines a proposed expansion of authority for one consequential agent action. It helps decision owners identify the evidence needed when a team wants to move an agent from recommendations to actions, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
AI investment and ROI masterclass
This area examines a funding hypothesis with an explicit comparator, cost boundary and realization mechanism. It helps decision owners identify the evidence needed when a proposed AI investment case relies on gross time savings or optimistic adoption, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
Vendor evaluation workshop
This area examines a comparative evaluation of candidate approaches for a bounded workflow. It helps decision owners identify the evidence needed when a shortlist exists but participants disagree about criteria and proof, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
Executive AI literacy program
This area examines a role-relevant business choice involving AI capability, authority, evidence and accountability. It helps decision owners identify the evidence needed when leadership discussions are slowed by inconsistent language and unclear assumptions about agents, keeping the analysis tied to a specific business decision.
- Leadership and organisational training
Custom leadership academies
This area examines a role-based learning pathway connected to specific organisational decisions. It helps decision owners identify the evidence needed when separate AI workshops are being requested without a coherent audience, progression or application plan, keeping the analysis tied to a specific business decision.
- Agentic system architecture and engineering
Agent architecture design
When a prototype must become an operable agent workflow, this specialty examines component boundaries and failure containment. Useful outputs can include architecture decision record and validation backlog to support the customer's next decision about which architecture separates reasoning, tools, state, and business authority.
- Agentic system architecture and engineering
Multi-agent system engineering
When several agents are proposed to divide a complex task, this specialty examines overlapping authority and conflicting conclusions. Useful outputs can include role and interaction design and single-agent comparison outline to support the customer's next decision.
- Agentic system architecture and engineering
Agent orchestration design
When agent behaviour must follow a controlled sequence across tools and people, this specialty examines state transition validity and branch ownership. Useful outputs can include orchestration design and exception handling cases to support the customer's next decision.
- Agentic system architecture and engineering
Context engineering
When agent answers vary with how evidence is assembled, this specialty examines source relevance and instruction precedence. Useful outputs can include context assembly specification and context evaluation cases to support the customer's next decision about which context construction strategy supports the task and its authority boundaries.
- Agentic system architecture and engineering
Memory-system engineering
When an agent needs information across interactions, this specialty examines write eligibility and retrieval contribution. Useful outputs can include memory lifecycle design and lifecycle verification plan to support the customer's next decision about what should be remembered, retrieved, corrected, and removed in this deployment.
- Agentic system architecture and engineering
Tool and connector engineering
When an agent must use enterprise tools with reliable semantics, this specialty examines argument validation and effect confirmation. Useful outputs can include tool contract and business outcome test cases to support the customer's next decision about how to expose tools that preserve intent, authority, and recoverable outcomes.
- Agentic system architecture and engineering
Model-routing architecture
When one model is being used for tasks with different needs, this specialty examines routing eligibility and fallback behaviour. Useful outputs can include routing policy design and fallback and monitoring specification to support the customer's next decision.
- Agentic system architecture and engineering
Human-in-the-loop workflow design
When agent actions need meaningful human authorisation, this specialty examines reviewer authority and approval freshness. Useful outputs can include human review workflow and refusal and escalation cases to support the customer's next decision about where people should review, intervene, refuse, and recover.
- Agentic system architecture and engineering
Durable agent workflow engineering
When agent work spans interruptions and external events, this specialty examines checkpoint boundaries and replay behaviour. Useful outputs can include durability design and interruption test plan to support the customer's next decision about how to resume work without losing state or repeating effects.
- Agentic system architecture and engineering
Agent rollback and recovery design
When an agent change could leave inconsistent state or external effects, this specialty examines configuration rollback and business compensation. Useful outputs can include rollback and compensation design and verification checklist to support the customer's next decision.
- Agentic system architecture and engineering
Agent state-management architecture
When agent state is distributed across sessions, stores, and tools, this specialty examines version conflicts and session boundaries. Useful outputs can include state model and state consistency cases to support the customer's next decision about which state is authoritative and how concurrent updates are resolved.
- Agentic system architecture and engineering
Agent observability architecture
When production behaviour cannot be reconstructed from existing logs, this specialty examines correlation across components and authority attribution. Useful outputs can include observability architecture and investigation walkthrough to support the customer's next decision about what observability is necessary to explain a consequential agent action.
- Agentic system architecture and engineering
Agent permission and authority design
When agent capabilities are expanding beyond reading information, this specialty examines least necessary action scope and delegation limits. Useful outputs can include authority model and authorisation test outline to support the customer's next decision about how to constrain delegated authority for each action.
- Agentic system architecture and engineering
Agent sandbox integration
When an agent needs to run code or commands in a controlled environment, this specialty examines credential exposure and output trust. Useful outputs can include sandbox integration design and boundary validation plan to support the customer's next decision.
- Agentic system architecture and engineering
Long-running agent engineering
When an agent must operate across long delays and changing conditions, this specialty examines stale assumptions and permission renewal. Useful outputs can include long-running workflow design and interruption and drift cases to support the customer's next decision.
- Agentic system architecture and engineering
Agent testing infrastructure
When tests need repeatable execution across changing agent components, this specialty examines fixture control and effect isolation. Useful outputs can include testing architecture and integration test backlog to support the customer's next decision about which infrastructure can reproduce meaningful cases without uncontrolled effects.
- Agentic system architecture and engineering
Agent control-plane design
When many agent deployments need common operating controls, this specialty examines configuration authority and emergency disable paths. Useful outputs can include control-plane architecture and control propagation tests to support the customer's next decision about which decisions belong in a shared control plane.
- Agentic system architecture and engineering
Agent deployment architecture
When an agent service must be released into a managed environment, this specialty examines configuration separation and version compatibility. Useful outputs can include deployment architecture and environment acceptance plan to support the customer's next decision about which deployment arrangement supports isolation, recovery, and change control.
- Agentic system architecture and engineering
Production-readiness engineering
When a working agent pilot is approaching production decisions, this specialty examines failure response and ownership and support. Useful outputs can include production readiness backlog and operating handover plan to support the customer's next decision about which engineering gaps prevent a controlled launch.
- Agentic system architecture and engineering
Forward-deployed agent engineering
When an agent implementation must adapt to real operational constraints, this specialty examines access and change boundaries and maintainability. Useful outputs can include embedded engineering charter and maintainer transfer checklist to support the customer's next decision.
- Agent-to-agent interoperability
Agent interoperability assessment
Decide whether agents can exchange work without losing meaning or control. The review connects interaction diagrams, message contracts, identity mappings and task-state definitions to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
MCP architecture and security
Design an MCP integration around explicit resource and authority boundaries. The review connects client and server configurations, tool inventories, identity flows and data paths to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Agent2Agent protocol evaluation
Evaluate a chosen Agent2Agent implementation against workflow expectations. The review connects selected specification revision, implementation documentation, task states and authentication settings to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Agent protocol compatibility testing
Verify compatibility between concrete agent protocol implementations. The review connects version matrices, supported features, message examples and failure responses to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Cross-vendor agent integration
Integrate agents from different vendors around one accountable workflow. The review connects vendor interface documentation, task contracts, identity requirements and exit constraints to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Agent capability discovery design
Design how agents find relevant capabilities without overtrusting descriptions. The review connects capability metadata, registries, eligibility rules and observed task requirements to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Delegation contract design
Make delegated work precise enough to accept, reject or recover. The review connects task requests, input schemas, authority limits and acceptance criteria to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Agent message-schema governance
Control message changes without breaking agent interactions. The review connects schema registries, consumers, version policies and message examples to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Protocol authentication review
Verify who is connecting through an agent protocol. The review connects issuer configuration, audience expectations, peer identity rules and transport setup to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Protocol authorisation review
Define what an authenticated agent may request through a protocol. The review connects task action catalogues, peer roles, resource scopes and enforcement records to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Trust negotiation between agents
Establish what evidence is required before agents rely on each other. The review connects trust policies, identity evidence, capability attestations and exception handling to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Cross-agent identity propagation
Preserve actor and tenant context across agent-to-agent handoffs. The review connects identity claims, delegation chains, tenant mappings and downstream audit records to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Agent communication privacy
Limit information exchanged between cooperating agents. The review connects message payloads, data classifications, purpose statements and retention settings to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Protocol conformance testing
Check implementation behaviour against a named protocol revision. The review connects normative requirements, implementation settings, test fixtures and claimed feature support to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Agent discovery security
Protect the process by which agents discover peers and tools. The review connects registration rules, metadata signatures, endpoint validation and cache policy to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Semantic interoperability assessment
Align the business meaning carried by agent messages. The review connects domain vocabularies, units, status definitions and task acceptance rules to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Multi-agent transaction integrity
Preserve business-state integrity across multi-agent transactions. The review connects transaction boundaries, idempotency keys, state models and compensation rules to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Agent protocol gateway engineering
Design a gateway that mediates agent protocols without hiding important differences. The review connects interface contracts, identity mappings, rate rules and observability requirements to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Interoperability test-suite development
Create reusable tests for cross-agent interaction contracts. The review connects protocol matrices, fixtures, peer simulators and agreed business task contracts to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Agent-to-agent interoperability
Vendor-neutral agent architecture
Retain architectural choices when agent vendors or protocols change. The review connects dependency maps, interface contracts, data export routes and operational requirements to a bounded review of agent interaction contracts, giving the integration owner a clearer basis for the next action.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Enterprise implementation partner
When an agent pilot must connect to enterprise operations, this specialty examines customer access prerequisites and integration responsibility at handoff. Useful outputs can include implementation responsibility matrix and joint delivery acceptance plan to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Secure deployment partner
When a proposed collaboration would deploy agents with sensitive access, this specialty examines secret handling responsibilities and production access approval. Useful outputs can include deployment control allocation and security review handoff to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Evaluation and assurance partner
When evaluation evidence must pass between a vendor and a customer, this specialty examines test ownership and evidence reproducibility. Useful outputs can include evaluation collaboration charter and findings review protocol to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Customer readiness assessment partner
When customers repeatedly reach implementation with missing prerequisites, this specialty examines availability of test data and identity and access readiness. Useful outputs can include customer prerequisite map and proceed-or-hold brief to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Forward-deployed engineering partner
When a customer deployment needs engineering inside unfamiliar systems, this specialty examines decision rights for code changes and architecture escalation. Useful outputs can include embedded work charter and maintainer handover plan to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Integration development partner
When agent adoption depends on connectors spanning several owners, this specialty examines schema compatibility and retry ownership. Useful outputs can include integration interface contract and maintenance responsibility map to support the customer's next decision about how to allocate connector design and maintenance across collaborators.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Reference architecture partner
When several customers ask for a reusable agent deployment pattern, this specialty examines assumption visibility and component substitution points. Useful outputs can include annotated reference architecture and validation backlog to support the customer's next decision about whether a reference design is sufficiently bounded to reuse.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Agent security review partner
When a vendor seeks specialist challenge of an agent integration, this specialty examines permitted attack surfaces and evidence handling. Useful outputs can include review qualification checklist and remediation evidence protocol to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Governance enablement partner
When a technology deployment lacks accountable policy and review processes, this specialty examines policy ownership and evidence production responsibilities. Useful outputs can include governance responsibility matrix and control evidence map to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Migration and modernization partner
When legacy workflows must coexist with a proposed agent implementation, this specialty examines parallel operation and data reconciliation. Useful outputs can include migration boundary map and retirement acceptance criteria to support the customer's next decision about which collaborator should own transition, compatibility, and retirement decisions.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Customer success engineering partner
When post-sale customers need technical help to sustain an agent workflow, this specialty examines problem triage responsibilities and configuration change authority. Useful outputs can include technical support routing map and escalation agreement draft to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Independent benchmark partner
When a comparative benchmark is proposed with commercial sponsorship, this specialty examines case selection influence and reproducibility access. Useful outputs can include benchmark governance proposal and reproducibility requirements to support the customer's next decision about whether benchmark design and reporting can withstand conflicts of interest.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Joint research partner
When a technical question could benefit from shared experiments, this specialty examines experimental responsibility and negative-result treatment. Useful outputs can include research protocol draft and findings publication checklist to support the customer's next decision about whether a joint research scope has useful questions and workable ownership.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Training and certification partner
When staff need practical agent skills and credible assessment, this specialty examines instructor competence and practical assessment design. Useful outputs can include training collaboration scope and credential wording review to support the customer's next decision about which training responsibilities and credential claims a collaborator may support.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Regional delivery partner
When a customer needs local operating coverage for a proposed deployment, this specialty examines named staffing availability and language and stakeholder coverage. Useful outputs can include coverage responsibility plan and cross-region handoff model to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Regulated-industry specialist partner
When an agent proposal touches consequential industry decisions, this specialty examines domain decision authority and evidence interpretation. Useful outputs can include specialist qualification brief and restricted-scope pilot conditions to support the customer's next decision about whether a qualified domain collaborator can define and review professional boundaries.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Marketplace solution partner
When a solution may be offered through a platform marketplace, this specialty examines eligibility evidence and customer contracting route. Useful outputs can include marketplace qualification brief and support and exit plan to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Joint go-to-market partner
When two organisations see a possible shared customer problem, this specialty examines message evidence and lead consent and routing. Useful outputs can include joint experiment brief and learning and stop criteria to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Technical content and lab partner
When a proposed lab should explain a difficult agent behaviour, this specialty examines setup reproducibility and failure demonstration. Useful outputs can include lab collaboration outline and technical claim review to support the customer's next decision about whether collaborators can produce reproducible technical education.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Incident-response partner
When an agent failure may cross organisational support boundaries, this specialty examines activation authority and containment responsibility. Useful outputs can include response collaboration charter and evidence and escalation handoff to support the customer's next decision about whether a response collaborator has a defined and authorised role.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Production Readiness Review for the partner’s customers
When a partner's customer needs a decision before production access expands, this specialty examines readiness evidence completeness and unresolved blocking dependencies. Useful outputs can include readiness evidence map and customer decision brief to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Secure Agent Launch package
When an agent launch requires coordinated engineering and security work, this specialty examines permission boundaries and safe failure behaviour. Useful outputs can include launch control plan and operational handover requirements to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Evaluation Foundation implementation
When a platform implementation lacks a maintainable evidence foundation, this specialty examines case ownership and evaluation integration. Useful outputs can include evaluation foundation design and release evidence integration plan to support the customer's next decision about how collaborators should build cases, execution, and review into customer workflows.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Enterprise Governance Accelerator
When agent adoption needs repeatable approvals and evidence handling, this specialty examines approval routing and evidence responsibilities. Useful outputs can include governance adaptation map and operating owner checklist to support the customer's next decision about whether a proposed governance accelerator fits existing decision structures.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Sandbox platform assessment
When a sandbox platform is considered for a specific agent deployment, this specialty examines customer configuration assumptions and evidence access. Useful outputs can include sandbox boundary assessment plan and unresolved proof requests to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Agent Identity Architecture
When agents need identities and delegated authority across several systems, this specialty examines principal attribution and permission issuance. Useful outputs can include identity collaboration design and lifecycle verification outline to support the customer's next decision about how collaborators should divide identity design and implementation.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
MCP Security Assessment
When an MCP integration is proposed with tools from another organisation, this specialty examines tool trust boundary and authorised action limits. Useful outputs can include MCP assessment boundary and findings handoff requirements to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Regulated Industry Deployment Blueprint
When a proposed agent needs a reusable deployment starting point, this specialty examines professional decision boundaries and evidence retention needs. Useful outputs can include domain deployment blueprint and specialist review checkpoints to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Migration to Agentic Workflows
When an existing workflow is being redesigned around agent assistance, this specialty examines human fallback availability and data reconciliation. Useful outputs can include workflow migration plan and adoption acceptance brief to support the customer's next decision.
- Partnerships with agentic IaaS, PaaS, and SaaS companies
Joint benchmarks, workshops, reference architectures, and executive roundtables
When several educational collaboration formats are being considered, this specialty examines format fit and participant expectations. Useful outputs can include format selection brief and learning evaluation outline to support the customer's next decision about which format best tests a shared customer question.
- Self-evolving and long-lived agent assurance
Extended-task agent evaluation
This area examines a complete extended task with checkpoints, interruptions, dependencies and a defined stopping condition. It helps decision owners identify the evidence needed when an agent performs well on short examples but must sustain work through many steps and interruptions, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Self-modifying agent risk assessment
This area examines a proposed or executed modification with a target, authority, validation and deployment path. It helps decision owners identify the evidence needed when an agent can alter prompts, code, tools or operating policies as part of its workflow, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Agent behavioral-drift monitoring
This area examines a behaviour measure or trajectory pattern compared across declared conditions and versions. It helps decision owners identify the evidence needed when an agent's behaviour changes over time without a clear model, prompt or workload explanation, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Capability-growth evaluation
This area examines a capability claim compared across agent states using protected tasks and declared resources. It helps decision owners identify the evidence needed when an agent is reported to improve after experience, tool changes or self-generated adaptations, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Recursive improvement control review
This area examines a recursive improvement cycle from candidate change to evaluation and promotion. It helps decision owners identify the evidence needed when an agent repeatedly proposes, tests and adopts changes intended to improve itself, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Learning-policy assessment
This area examines a learning event that selects information, changes state and affects later behaviour. It helps decision owners identify the evidence needed when an agent can retain feedback or adapt behaviour without an explicit learning policy, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Experience replay and memory evolution testing
This area examines a replay or consolidation event and its effect on later retrieval and action. It helps decision owners identify the evidence needed when past interactions are replayed, summarised or consolidated into evolving agent memory, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Goal stability evaluation
This area examines a goal and its constraints carried through a multi-step trajectory. It helps decision owners identify the evidence needed when an agent works through changing evidence, interruptions and delegated subtasks, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Reward-hacking detection
This area examines a reward-producing action sequence compared with the underlying task objective. It helps decision owners identify the evidence needed when an agent is optimised against a reward, judge or success metric, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Emergent strategy monitoring
This area examines a recurring action pattern or strategy within a declared task and authority boundary. It helps decision owners identify the evidence needed when an agent begins using unfamiliar sequences of tools, delegation or workarounds, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Long-term alignment testing
This area examines a longitudinal sequence of tasks and changes evaluated against declared behavioural requirements. It helps decision owners identify the evidence needed when an agent's goals, experience and environment evolve across repeated tasks, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Model and prompt mutation governance
This area examines a model or prompt revision linked to lineage, evaluation evidence and deployment approval. It helps decision owners identify the evidence needed when models or prompts can change between releases through automated optimisation or provider updates, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Evaluation contamination detection
This area examines a test item or evaluation artefact traced through access, training, memory and optimisation pathways. It helps decision owners identify the evidence needed when an agent or optimisation loop may have seen examples used to judge improvement, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Capability regression tracking
This area examines a previously supported capability compared across agent versions and task strata. It helps decision owners identify the evidence needed when updates improve a target task but may weaken previously useful behaviour, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Agent lineage and versioning
This area examines a composite agent version linking behavior-relevant artefacts and state dependencies. It helps decision owners identify the evidence needed when an agent's behaviour depends on changing prompts, tools, memory and configuration that are not versioned together, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Safe experimentation boundaries
This area examines a proposed experiment with data, tools, resources, authority and exit conditions. It helps decision owners identify the evidence needed when an adaptive agent is permitted to try new strategies, code or tool combinations, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Evolution rollback design
This area examines a rollback across model, prompt, memory, tools and dependent external state. It helps decision owners identify the evidence needed when an update or learning event produces unacceptable behaviour and the team needs to restore a known operating state, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Continuous assurance for adaptive agents
This area examines a change or operating signal linked to affected requirements and assurance activities. It helps decision owners identify the evidence needed when an agent changes through prompts, memory, tools or learning faster than periodic reviews can follow, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Multi-month agent simulation
This area examines a simulated operational timeline with state evolution, external events and explicit time assumptions. It helps decision owners identify the evidence needed when an agent is expected to operate across long periods but real-time evidence is not yet available, keeping the analysis tied to a specific business decision.
- Self-evolving and long-lived agent assurance
Longitudinal agent audit
This area examines a longitudinal record of agent versions, behaviour, findings and accountable decisions. It helps decision owners identify the evidence needed when an agent has accumulated changes, incidents and decisions over an extended period, keeping the analysis tied to a specific business decision.
- Agent evaluation and assurance
Agent evaluation strategy
Choose which evaluation evidence should govern the next release. The review connects workflow requirements, incident notes, user groups and existing test inventories to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Agent evaluation harness development
Build a harness that preserves comparable evidence across agent revisions. The review connects runner code, configuration manifests, fixtures, tool mocks and retained execution traces to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Task and capability evaluation
Determine which tasks an agent can complete within defined operating limits. The review connects task definitions, accepted examples, difficult exceptions and domain scoring rubrics to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
End-to-end workflow evaluation
Decide whether the complete workflow reaches the intended business state. The review connects workflow state diagrams, handoff records, tool events and final business-system states to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Agent reliability benchmarking
Compare agent reliability under a representative and declared workload. The review connects repeated-run records, workload strata, dependency versions and failure classifications to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Agent regression testing
Detect changes that break previously accepted agent behaviour. The review connects versioned baselines, accepted case results, change notes and release configurations to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Golden trace development
Define reference traces without prescribing one brittle reasoning path. The review connects expert demonstrations, tool contracts, permitted alternatives and terminal-state evidence to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Adversarial agent evaluation
Understand how an agent responds to deliberately difficult task conditions. The review connects authorised attack scenarios, protected actions, synthetic records and escalation rules to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Tool-use evaluation
Verify that tool actions match the request and the agent's authority. The review connects tool schemas, permission rules, request examples and tool-side execution logs to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Multi-agent coordination evaluation
Determine whether cooperating agents preserve ownership and complete shared work. The review connects role contracts, message histories, shared-state rules and ownership transitions to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Human–agent interaction evaluation
Design evidence for the moments people approve, correct or interrupt an agent. The review connects interaction flows, approval screens, reviewer instructions and intervention records to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Cost and latency evaluation
Set useful cost and response-time budgets for a complete agent task. The review connects request traces, model and tool usage, retry events and human-review effort to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Agent failure-recovery evaluation
Decide how an agent should recover when part of a workflow fails. The review connects retry policies, state checkpoints, error contracts and compensation rules to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Uncertainty and calibration evaluation
Determine whether uncertainty signals support safe task routing. The review connects confidence outputs, labelled cases, abstention rules and observed correctness to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Autonomy-level evaluation
Choose the permitted level of agent action for a particular workflow. The review connects action inventories, reversibility, approval rules and operator intervention paths to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Policy-compliance evaluation
Translate operational policies into testable agent behaviour. The review connects approved policy text, precedence rules, exceptions and action logs to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Agent release-gate design
Make the release decision depend on explicit evidence and accountable exceptions. The review connects release criteria, critical failure definitions, test reports and exception records to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Continuous production evaluation
Refresh evaluation evidence as live workflows and dependencies change. The review connects sanitised production samples, change events, drift signals and incident classifications to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Independent agent assurance
Commission a bounded review whose conclusions can be traced to evidence. The review connects scope statements, prior assessments, conflict disclosures and reproducible evaluation records to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Agent evaluation and assurance
Agent benchmark development
Build a benchmark that measures the decision owner's intended use. The review connects task taxonomy, representative cases, scoring rules and contamination controls to a bounded review of evaluation evidence, giving the release owner a clearer basis for the next action.
- Component-level agent evaluation
Memory accuracy and relevance testing
When remembered information appears irrelevant or incorrect, this specialty examines fact correctness and task contribution. Useful outputs can include memory accuracy case set and deployment remediation priorities to support the customer's next decision about whether retrieved memories improve the deployment's actual task outcomes.
- Component-level agent evaluation
Memory retention and forgetting evaluation
When the deployment needs rules for preserving or forgetting information, this specialty examines expiry enforcement and purposeful persistence. Useful outputs can include retention and forgetting cases and propagation gap findings to support the customer's next decision.
- Component-level agent evaluation
Cross-user memory leakage testing
When a memory-enabled agent serves several users or tenants, this specialty examines cross-user queries and tenant switching. Useful outputs can include isolation test matrix and boundary remediation findings to support the customer's next decision about whether identity boundaries hold across writing, retrieval, and reuse.
- Component-level agent evaluation
Memory poisoning assessment
When untrusted information may enter persistent agent memory, this specialty examines write acceptance and provenance and trust handling. Useful outputs can include memory poisoning case set and control gap recommendations to support the customer's next decision.
- Component-level agent evaluation
Temporal consistency evaluation
When remembered facts change over time, this specialty examines effective dates and update precedence. Useful outputs can include temporal case matrix and temporal consistency findings to support the customer's next decision about whether the agent uses the correct fact for the relevant time and task.
- Component-level agent evaluation
Memory provenance and lineage review
When teams cannot explain where a remembered assertion came from, this specialty examines source attribution and transformation visibility. Useful outputs can include memory lineage map and correction traceability backlog to support the customer's next decision about whether provenance remains usable from source through later decisions.
- Component-level agent evaluation
Memory deletion and correction testing
When a user or data owner requests deletion or correction, this specialty examines primary store update and derived copy propagation. Useful outputs can include deletion and correction test pack and unresolved lifecycle findings to support the customer's next decision.
- Component-level agent evaluation
Long-term personalization evaluation
When an agent adapts to users over repeated interactions, this specialty examines preference stability and user correction. Useful outputs can include personalization evaluation plan and user control recommendations to support the customer's next decision about whether personalization remains useful, correctable, and appropriate over time.
- Component-level agent evaluation
Memory cost and retrieval performance testing
When memory adds latency or cost without a clear contribution, this specialty examines retrieval contribution and tail behaviour. Useful outputs can include memory performance profile and configuration experiment backlog to support the customer's next decision about which retrieval configuration earns its operating overhead.
- Component-level agent evaluation
Episodic versus semantic memory comparison
When the deployment must choose how to represent experience and reusable knowledge, this specialty examines event detail preservation and abstraction correctness. Useful outputs can include representation comparison and memory design decision brief to support the customer's next decision.
- Component-level agent evaluation
Context relevance evaluation
When answers depend on a large mix of retrieved and supplied context, this specialty examines necessary evidence inclusion and distracting material. Useful outputs can include context relevance case set and assembly improvement backlog to support the customer's next decision.
- Component-level agent evaluation
Context-window utilization testing
When context length is growing or prompts are being truncated, this specialty examines truncation behaviour and evidence prioritisation. Useful outputs can include context budget profile and utilization test cases to support the customer's next decision about whether the available context budget preserves the information needed for the task.
- Component-level agent evaluation
Context poisoning and injection assessment
When untrusted documents or tool outputs enter the agent's context, this specialty examines instruction confusion and source trust handling. Useful outputs can include context injection cases and mitigation verification backlog to support the customer's next decision.
- Component-level agent evaluation
Retrieval quality evaluation
When the agent cannot reliably find evidence needed to answer tasks, this specialty examines missing relevant items and irrelevant ranking. Useful outputs can include retrieval quality case set and retrieval remediation priorities to support the customer's next decision.
- Component-level agent evaluation
Permission-aware retrieval testing
When retrieval must respect changing user or tenant access, this specialty examines access revocation and inherited permissions. Useful outputs can include permission-aware retrieval matrix and revocation verification cases to support the customer's next decision about whether authorised context is selected for the acting identity.
- Component-level agent evaluation
Citation and grounding evaluation
When users need to verify the agent's factual answers, this specialty examines citation existence and claim entailment. Useful outputs can include claim-to-source review and citation failure taxonomy to support the customer's next decision about whether claims are supported by the cited evidence.
- Component-level agent evaluation
Context compression testing
When the deployment compresses long context before reasoning, this specialty examines loss of qualifiers and source and time preservation. Useful outputs can include compression comparison cases and preservation requirements to support the customer's next decision about whether compression preserves the facts, exceptions, and authority needed downstream.
- Component-level agent evaluation
Stale-context detection
When agents continue using information after it becomes outdated, this specialty examines freshness signals and invalidation behaviour. Useful outputs can include staleness evaluation cases and freshness decision rules to support the customer's next decision about whether stale context is detected and handled before affecting a task.
- Component-level agent evaluation
Context-conflict resolution testing
When the agent receives sources that disagree, this specialty examines authority precedence and temporal distinctions. Useful outputs can include conflict resolution test set and source policy recommendations to support the customer's next decision about whether conflicts are resolved using explicit authority and context rules.
- Component-level agent evaluation
Lost-in-the-middle evaluation
When important evidence may be overlooked inside lengthy inputs, this specialty examines position sensitivity and ordering confounds. Useful outputs can include position-controlled case set and context assembly recommendations to support the customer's next decision about whether evidence position changes task performance in this deployment.
- Component-level agent evaluation
Workflow-routing evaluation
When requests are sent to different workflows or agents, this specialty examines ambiguous intent and boundary categories. Useful outputs can include routing evaluation matrix and clarification and fallback cases to support the customer's next decision about whether routing selects an appropriate path for the task and authority.
- Component-level agent evaluation
Planner quality assessment
When a planner produces steps that look plausible but fail in execution, this specialty examines missing prerequisites and invalid step order. Useful outputs can include planner quality rubric and replanning test cases to support the customer's next decision.
- Component-level agent evaluation
Delegation correctness testing
When one agent delegates work to another, this specialty examines scope preservation and authority transfer limits. Useful outputs can include delegation contract review and corrective design findings to support the customer's next decision about whether delegated tasks preserve the intended scope and authority.
- Component-level agent evaluation
Multi-agent handoff evaluation
When work passes between agents with different roles or state, this specialty examines missing context and duplicate ownership. Useful outputs can include handoff evaluation cases and handoff contract recommendations to support the customer's next decision about whether handoffs transfer the necessary context, status, and responsibility.
- Component-level agent evaluation
Loop and deadlock detection
When workflows repeat steps or wait indefinitely, this specialty examines productive versus unproductive repetition and circular waits. Useful outputs can include loop and deadlock cases and timeout and escalation recommendations to support the customer's next decision.
- Component-level agent evaluation
Termination-condition testing
When agents stop too early or continue after the task is complete, this specialty examines premature success and unresolved effects. Useful outputs can include termination case set and stopping and escalation rules to support the customer's next decision.
- Component-level agent evaluation
Retry and compensation testing
When external failures trigger retries or compensating actions, this specialty examines retry eligibility and compensation ordering. Useful outputs can include retry and compensation matrix and recovery recommendations to support the customer's next decision about whether recovery preserves the intended business outcome.
- Component-level agent evaluation
Workflow determinism assessment
When the same task produces materially different execution paths, this specialty examines outcome variation and consequential path changes. Useful outputs can include repeatability study and acceptable-variation criteria to support the customer's next decision about which variation is acceptable and which undermines reliability.
- Component-level agent evaluation
Supervisor-agent evaluation
When a supervisor allocates work and judges other agents' results, this specialty examines erroneous worker acceptance and task allocation. Useful outputs can include supervisor evaluation rubric and oversight gap findings to support the customer's next decision.
- Component-level agent evaluation
Failure-propagation analysis
When one component failure creates unexpected downstream behaviour, this specialty examines error translation and state contamination. Useful outputs can include failure propagation map and containment and recovery priorities to support the customer's next decision about how failures spread and where containment or recovery should occur.
- Component-level agent evaluation
Tool-selection accuracy
When the agent has several tools with overlapping descriptions, this specialty examines semantic fit and unnecessary tools. Useful outputs can include tool selection case set and tool catalogue recommendations to support the customer's next decision about whether it selects a tool that fits the task and authority.
- Component-level agent evaluation
Tool-argument correctness
When tool calls fail or affect the wrong records despite valid syntax, this specialty examines entity resolution and required constraints. Useful outputs can include argument correctness cases and validation and clarification recommendations to support the customer's next decision.
- Component-level agent evaluation
Unauthorised tool-use testing
When an agent can invoke tools with consequential permissions, this specialty examines role boundaries and revoked access. Useful outputs can include unauthorized-use test matrix and authority control findings to support the customer's next decision about whether attempted tool use stays within the acting principal's authority.
- Component-level agent evaluation
Partial-execution recovery
When a multi-step task fails after some external effects occur, this specialty examines completed effect detection and safe continuation. Useful outputs can include partial-execution case set and recovery and communication recommendations to support the customer's next decision.
- Component-level agent evaluation
Idempotency evaluation
When requests may be repeated through retries, reconnects, or user resubmission, this specialty examines duplicate request handling and idempotency scope. Useful outputs can include idempotency evaluation cases and operation identity recommendations to support the customer's next decision.
- Component-level agent evaluation
Tool-result interpretation
When tools return nuanced, partial, or ambiguous responses, this specialty examines success versus acceptance and missing fields. Useful outputs can include tool-result interpretation cases and response contract improvements to support the customer's next decision about whether the agent interprets results according to business meaning.
- Component-level agent evaluation
Connector reliability testing
When connector behaviour varies under outages or changing upstream systems, this specialty examines connection loss and schema changes. Useful outputs can include connector reliability profile and monitoring and recovery backlog to support the customer's next decision.
- Component-level agent evaluation
Tool-description poisoning tests
When tool descriptions can be supplied or changed by external sources, this specialty examines instruction-bearing descriptions and selection manipulation. Useful outputs can include tool description poisoning cases and review and enforcement recommendations to support the customer's next decision.
- Component-level agent evaluation
High-impact action confirmation
When the agent may perform actions with material consequences, this specialty examines specificity of consent and changed action after approval. Useful outputs can include confirmation evaluation cases and oversight design recommendations to support the customer's next decision.
- Component-level agent evaluation
Tool-chain integrity evaluation
When a task chains several tools across different systems, this specialty examines data transformation integrity and authority continuity. Useful outputs can include tool-chain integrity map and end-to-end control findings to support the customer's next decision about whether intent, data integrity, and authority survive the full chain.
- Component-level agent evaluation
Agent context system evaluation
When the team needs an integrated view of context-related failures, this specialty examines selection and grounding and trust and permissions. Useful outputs can include context system evaluation plan and prioritised deployment improvements to support the customer's next decision.
- Component-level agent evaluation
Agent memory system evaluation
When a release or incident raises questions about persistent context, this specialty examines task contribution and accuracy and isolation and poisoning. Useful outputs can include memory deployment evaluation and decision and remediation brief to support the customer's next decision.
- Component-level agent evaluation
Agent orchestration system evaluation
When routing, planning, and recovery issues appear across a workflow, this specialty examines planning and routing and delegation and handoff. Useful outputs can include orchestration system evaluation and coordination remediation priorities to support the customer's next decision.
- Distinctive research-led offerings
Agent Reliability Index
This area examines a completed task under declared tools, permissions, interruption conditions and review rules. It helps decision owners identify the evidence needed when several candidate agents appear strong on demonstrations but have been tested on different work, keeping the analysis tied to a specific business decision.
- Distinctive research-led offerings
Agent Authority Map
This area examines a chain from human mandate through agent identity to a consequential tool action. It helps decision owners identify the evidence needed when agent demonstrations obscure who can authorise an action or withdraw that authority, keeping the analysis tied to a specific business decision.
- Distinctive research-led offerings
Agent Evolution Observatory
This area examines a versioned agent configuration observed across successive declared periods. It helps decision owners identify the evidence needed when an agent changes prompts, tools or retained experience between evaluation snapshots, keeping the analysis tied to a specific business decision.
- Distinctive research-led offerings
Sandbox Assurance Benchmark
This area examines a sandbox configuration under an authorised, reproducible workload and threat model. It helps decision owners identify the evidence needed when several isolation approaches are being considered for agents that execute code or use computers, keeping the analysis tied to a specific business decision.
- Distinctive research-led offerings
Agent Interoperability Test Lab
This area examines a defined agent-to-tool or agent-to-agent interaction across versioned interfaces. It helps decision owners identify the evidence needed when a connector or protocol integration works in a demonstration but has unclear failure semantics, keeping the analysis tied to a specific business decision.
- Distinctive research-led offerings
Memory Safety Benchmark
This area examines a customer-style memory deployment including prompts, identities, retrieval and downstream tools. It helps decision owners identify the evidence needed when persistent memory is being introduced or replaced without evidence for correction, isolation and task contribution, keeping the analysis tied to a specific business decision.
- Distinctive research-led offerings
Agentic Impact Index
This area examines a workflow outcome measured against a declared baseline and relevant resource use. It helps decision owners identify the evidence needed when AI initiatives report activity and savings using incompatible definitions, keeping the analysis tied to a specific business decision.
- Distinctive research-led offerings
Responsible Autonomy Framework
This area examines a business action with defined authority, consequence, reversibility and human oversight. It helps decision owners identify the evidence needed when teams use the same autonomy label for agents with very different rights and consequences, keeping the analysis tied to a specific business decision.
- Distinctive research-led offerings
Agent Incident Simulation
This area examines a simulated incident progressing through detection, containment, investigation and restoration. It helps decision owners identify the evidence needed when an organisation has response policies but has not rehearsed an agent-specific failure across teams, keeping the analysis tied to a specific business decision.
- Distinctive research-led offerings
Independent Agent Assurance Report
This area examines a bounded report conclusion linked to criteria, evidence, reviewer competence and disclosed conflicts. It helps decision owners identify the evidence needed when an assurance-style report is requested without agreement on criteria, independence or intended reliance, keeping the analysis tied to a specific business decision.
- Agent security and cybersecurity
Agentic AI threat modelling
Prioritise credible attack paths against the agent's business authority. The review connects data-flow diagrams, actor assumptions, assets and action privileges to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
Secure agent architecture assessment
Assess where security controls should sit in an agent architecture. The review connects component diagrams, identity flows, policy enforcement and data paths to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
Prompt-injection testing
Test whether direct instructions can redirect an agent beyond its approved task. The review connects instruction hierarchy, protected behaviours, synthetic prompts and tool audit records to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
Indirect prompt-injection assessment
Examine instructions hidden inside information an agent consumes. The review connects content provenance, retrieval paths, trust labels and consequential action rules to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
Tool and capability abuse testing
Understand whether legitimate tools can be combined for unintended actions. The review connects tool inventories, permitted use cases, action sequences and resource constraints to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
Agent privilege-escalation testing
Test whether agent authority can increase across a workflow. The review connects permission graphs, delegation paths, privileged interfaces and test identities to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
Cross-agent attack analysis
Identify how one agent can manipulate another agent's decisions. The review connects agent roles, message provenance, shared resources and delegation rules to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
Agent identity spoofing assessment
Verify that agent identity claims cannot substitute for authentication. The review connects peer registries, identity proof, issuer rules and role-assignment records to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
Memory poisoning tests
Examine whether persistent memory can carry attacker-controlled instructions. The review connects memory write paths, provenance, retrieval rules and correction controls to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
Retrieval poisoning tests
Test how corrupted retrieval sources influence agent outputs and actions. The review connects corpus permissions, indexing rules, ranking behaviour and source-version records to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
Data-exfiltration testing
Validate whether information can leave the intended agent boundary. The review connects data classifications, outbound tools, recipient controls and synthetic canary records to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
Agent supply-chain security
Review the chain of components that supplies an agent deployment. The review connects dependency inventories, provenance records, update paths and approval policies to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
Model and dependency security
Assess security-relevant changes in model and software dependencies. The review connects version manifests, dependency advisories, integration tests and provider configuration to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
MCP server security review
Inspect the authority exposed by a particular MCP server integration. The review connects server version, exposed tool definitions, authentication settings and resource permissions to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
Adversarial agent testing
Challenge a defined agent workflow using authorised adversarial scenarios. The review connects rules of engagement, threat hypotheses, synthetic data and stop conditions to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
AI incident-response planning
Prepare an operational response to agent-related incidents. The review connects system ownership, tool privileges, incident classes and evidence-retention controls to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
Agent kill-switch and containment design
Design a reliable way to stop agent action and contain remaining authority. The review connects worker queues, credential paths, scheduler controls and cancellation semantics to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
AI security monitoring
Observe security-relevant behaviour without drowning operators in noise. The review connects event models, known incident patterns, action inventories and analyst workflows to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
Secure coding review for agent systems
Review application code that translates agent choices into actions. The review connects source code, tool schemas, validation logic and error-handling paths to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent security and cybersecurity
Security release assurance
Decide what security evidence is required before an agent release. The review connects threat models, control tests, unresolved findings and release changes to a bounded review of agent security, giving the security decision owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Agent identity architecture
Choose how agents should be identified across business systems. The review connects principal inventories, trust boundaries, account ownership and authentication paths to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Workload identity for agents
Bind execution authority to a verifiable agent workload. The review connects runtime inventory, workload attributes, credential exchanges and resource trust policies to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Agent credential lifecycle design
Control agent credentials from issuance through retirement. The review connects credential metadata, issuance flows, rotation records and revocation dependencies to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Agent authentication review
Verify the identity checks that precede agent access. The review connects authentication sequence diagrams, token validation configuration and failed-login records to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Agent authorisation modelling
Define which resources and actions an agent may use in context. The review connects resource models, action catalogues, subject attributes and policy decision records to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Delegated authority design
Make delegated authority explicit and bounded. The review connects delegation records, action limits, expiry rules and authority transfer diagrams to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
User-to-agent delegation controls
Preserve a person's intended limits when an agent acts for them. The review connects user journeys, approval records, consent text and action parameters to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Agent-to-agent delegation controls
Prevent authority from expanding as agents delegate to one another. The review connects delegation chains, agent role definitions, resource policies and downstream tool logs to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Non-human identity governance
Put accountable ownership around non-human identities. The review connects identity inventories, owners, usage metadata, approvals and retirement records to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Least-privilege agent design
Reduce agent permissions to those required by the agreed workflow. The review connects observed action traces, business requirements, permission grants and deny logs to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Temporary credential architecture
Design short-lived authority that remains usable during real work. The review connects issuance policy, validity periods, renewal paths and task-duration traces to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Agent impersonation testing
Test whether one actor can convincingly masquerade as an authorised agent. The review connects trusted identity attributes, validation points, test principals and attribution records to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Identity propagation across tools
Carry the correct actor context through tool integrations. The review connects identity claims, gateway mappings, tool requests and resource audit logs to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Cross-tenant identity isolation
Verify that agent identity cannot cross tenant boundaries. The review connects tenant-principal mappings, token claims, resource partitions and synthetic tenant fixtures to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Agent attribution and non-repudiation
Make consequential agent actions attributable to an accountable chain. The review connects event schemas, timestamps, identity evidence, delegation records and integrity controls to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Authorisation policy testing
Verify authorisation policy decisions against intended rules. The review connects policy versions, subject-resource fixtures, action catalogues and expected decision tables to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Consent and mandate management
Manage mandates as records that can change and expire. The review connects mandate versions, consent interactions, withdrawal events and execution queues to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Agent transaction signing
Bind transaction approval to the exact action being executed. The review connects canonical transaction fields, signing flows, verifier rules and replay controls to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Identity revocation testing
Verify that withdrawn authority stops future agent action. The review connects revocation events, token caches, queued jobs and downstream permission checks to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent identity, authorisation, and delegation
Agent identity maturity assessment
Prioritise the next improvements in agent identity controls. The review connects identity inventories, architecture diagrams, lifecycle evidence and control-test results to a bounded review of identity and authority, giving the identity control owner a clearer basis for the next action.
- Agent sandbox assurance
Sandbox architecture review
Decide whether the proposed execution boundary matches the agent's privileges. The review connects deployment diagrams, mount configuration, network routes and privilege settings to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Isolation-boundary testing
Verify the resource boundaries an execution environment is meant to enforce. The review connects approved boundary claims, synthetic fixtures, runtime policies and host observations to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Escape-resistance assessment
Assess evidence for resisting escape from a defined execution environment. The review connects runtime versions, privilege configuration, host interfaces and an authorised threat model to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Filesystem-access testing
Verify what files an agent execution environment can read or change. The review connects mount maps, path permissions, symlink rules and synthetic sensitive-file markers to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Network-egress control review
Define and validate where a sandbox may send network traffic. The review connects destination policies, DNS behaviour, proxy routes and connection logs to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Secret-exposure testing
Check whether sandboxed work can expose credentials or other secret material. The review connects secret-injection design, redaction rules, access paths and synthetic canary values to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Resource-exhaustion testing
Establish what happens when agent execution consumes excessive resources. The review connects resource quotas, scheduler settings, workload limits and telemetry to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Malicious-code execution testing
Validate controls around potentially harmful generated code. The review connects execution policies, allowed interpreters, network limits and harmless test payloads to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Persistence and cleanup verification
Verify that finished agent sessions leave no unintended reusable state. The review connects session lifecycle events, volume policies, artefact lists and synthetic state markers to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Tenant-isolation assessment
Assess separation between tenants sharing agent execution infrastructure. The review connects tenant allocation rules, storage namespaces, cache keys and synthetic tenant cases to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Sandbox image supply-chain review
Control the provenance and contents of sandbox images. The review connects image manifests, build recipes, dependency inventories and signing or provenance records to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Runtime-policy validation
Confirm that runtime restrictions are enforced in the deployed configuration. The review connects policy files, launch parameters, enforcement logs and representative workload fixtures to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Snapshot and rollback testing
Determine what a snapshot can restore after an agent failure. The review connects snapshot scope, external tool state, storage dependencies and recovery procedures to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Auditability assessment
Make sandbox activity reconstructable for operational review. The review connects event schemas, timestamp sources, retention rules and correlation identifiers to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Comparative sandbox benchmarking
Compare sandbox options against the same workload and control requirements. The review connects candidate configurations, common workload fixtures, isolation criteria and cost inputs to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Secure computer-use environment design
Define a controlled environment for agents that operate a computer interface. The review connects application inventory, account permissions, clipboard rules and download paths to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Browser sandbox evaluation
Verify the browser state and access boundaries used by an agent. The review connects browser profile settings, cookie stores, extension inventory and navigation policies to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- Agent sandbox assurance
Code-execution sandbox certification readiness
Prepare evidence for a separately defined sandbox certification assessment. The review connects target scheme requirements, scope statements, architecture evidence and control-test records to a bounded review of execution boundaries, giving the runtime owner a clearer basis for the next action.
- AI governance and regulatory readiness
AI governance maturity assessment
When leadership needs a baseline before choosing governance investments, this specialty examines documented versus practiced controls and evidence availability. Useful outputs can include governance baseline and prioritised operating improvements to support the customer's next decision about which governance gaps materially affect current agent decisions.
- AI governance and regulatory readiness
AI system inventory creation
When the organisation cannot reliably identify its AI systems and agents, this specialty examines inventory completeness and lifecycle status. Useful outputs can include AI inventory model and maintenance ownership rules to support the customer's next decision.
- AI governance and regulatory readiness
Agent risk classification
When a new agent needs proportionate review and control requirements, this specialty examines autonomy level and reversibility. Useful outputs can include internal classification rubric and review routing recommendation to support the customer's next decision about how to classify the deployment's operational risk for internal decisions.
- AI governance and regulatory readiness
AI policy development
When existing policies do not explain how agent use should be governed, this specialty examines policy applicability and operational feasibility. Useful outputs can include AI policy draft and exception and review provisions to support the customer's next decision.
- AI governance and regulatory readiness
Agent autonomy policy
When an agent is being allowed to act with less human intervention, this specialty examines authority thresholds and changing context. Useful outputs can include autonomy policy draft and escalation and renewal rules to support the customer's next decision.
- AI governance and regulatory readiness
Acceptable-use policy
When staff need clear rules for using AI tools in everyday work, this specialty examines comprehensible examples and prohibited data handling. Useful outputs can include acceptable-use policy draft and exception and support guide to support the customer's next decision.
- AI governance and regulatory readiness
Human-oversight framework
When human oversight is required but responsibilities are unclear, this specialty examines review information quality and refusal authority. Useful outputs can include oversight role model and oversight effectiveness cases to support the customer's next decision about what people must observe, decide, and be able to stop.
- AI governance and regulatory readiness
AI control-library development
When teams need reusable controls with clear evidence expectations, this specialty examines control objective clarity and testability. Useful outputs can include control library structure and evidence and testing conventions to support the customer's next decision about which control library structure supports implementation and testing.
- AI governance and regulatory readiness
Governance workflow implementation
When policy approvals are managed through inconsistent informal exchanges, this specialty examines routing correctness and approval state. Useful outputs can include governance workflow design and operating acceptance cases to support the customer's next decision about how to implement review and approval workflows that preserve accountability.
- AI governance and regulatory readiness
Evidence-register design
When reviewers cannot locate the evidence behind important AI decisions, this specialty examines artefact provenance and version linkage. Useful outputs can include evidence register schema and refresh and supersession workflow to support the customer's next decision.
- AI governance and regulatory readiness
Model and agent change management
When changes to models, prompts, tools, or data alter agent behaviour, this specialty examines impact classification and regression evidence. Useful outputs can include agent change policy and approval and rollback workflow to support the customer's next decision.
- AI governance and regulatory readiness
Third-party AI risk management
When an agent depends on external AI products or services, this specialty examines customer configuration responsibility and change notification. Useful outputs can include third-party AI risk profile and shared responsibility and exit map to support the customer's next decision.
- AI governance and regulatory readiness
AI procurement controls
When an AI purchase is approaching approval, this specialty examines decision evidence completeness and ownership after purchase. Useful outputs can include AI procurement checklist and stakeholder responsibility map to support the customer's next decision about what procurement controls should establish before contract and deployment decisions.
- AI governance and regulatory readiness
Exception-management workflow
When teams request exceptions to AI controls, this specialty examines approval authority and expiry conditions. Useful outputs can include exception workflow and expiry and closure checks to support the customer's next decision about how exceptions should be assessed, authorised, monitored, and closed.
- AI governance and regulatory readiness
Algorithmic impact assessment
When an agent may materially affect people or business outcomes, this specialty examines distribution of consequences and contestability and correction. Useful outputs can include impact assessment draft and mitigation and monitoring plan to support the customer's next decision.
- AI governance and regulatory readiness
Regulatory applicability assessment
When a deployment may fall within unfamiliar regulatory requirements, this specialty examines fact completeness and role and use distinctions. Useful outputs can include applicability fact pack and operational dependency map to support the customer's next decision.
- AI governance and regulatory readiness
EU AI Act readiness
When an organisation needs to prepare for potentially applicable EU AI Act requirements, this specialty examines system inventory alignment and responsibility allocation. Useful outputs can include readiness evidence map and qualified-review dependencies to support the customer's next decision.
- AI governance and regulatory readiness
ISO/IEC 42001 readiness
When the organisation wants to assess preparation for ISO/IEC 42001, this specialty examines scope consistency and documented versus operating processes. Useful outputs can include readiness gap assessment and improvement and review plan to support the customer's next decision.
- AI governance and regulatory readiness
NIST AI RMF implementation
When the organisation wants to apply NIST AI RMF in its operating workflow, this specialty examines contextual mapping and measurable control activity. Useful outputs can include contextual framework mapping and evidence and review model to support the customer's next decision.
- AI governance and regulatory readiness
AI audit preparation
When an audit or formal review needs a coherent AI evidence pack, this specialty examines traceability and completeness and freshness. Useful outputs can include audit preparation index and owner and response plan to support the customer's next decision.
- AI governance and regulatory readiness
Continuous compliance monitoring
When approved controls and obligations need ongoing evidence refresh, this specialty examines signal relevance and evidence staleness. Useful outputs can include monitoring design and review and exception workflow to support the customer's next decision about which signals should trigger review when deployments or requirements change.
- AI governance and regulatory readiness
Board-level AI risk reporting
When directors need a clear view of agent exposure and management decisions, this specialty examines materiality and uncertainty visibility. Useful outputs can include board reporting structure and decision and action summary to support the customer's next decision.
- Privacy engineering for AI systems
AI privacy impact assessment
This area examines a defined processing activity across collection, inference, action, retention and disposal. It helps decision owners identify the evidence needed when a new agent workflow introduces personal information, inference or additional recipients, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Agent privacy threat modelling
This area examines a privacy threat scenario linking an actor, information flow, action and adverse effect. It helps decision owners identify the evidence needed when an agent handles personal information across tools, memory and multiple users, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Privacy-by-design architecture
This area examines a data and control boundary across collection, processing, storage and tool execution. It helps decision owners identify the evidence needed when a proposed agent design distributes personal information before privacy requirements have been translated into system boundaries, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Sensitive-data flow mapping
This area examines a data item or category moving between source, context, model, tool, memory and logs. It helps decision owners identify the evidence needed when the organisation cannot explain which systems receive or retain information used by an agent, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Purpose-limitation control design
This area examines a purpose-bound request for information or action within an agent workflow. It helps decision owners identify the evidence needed when an agent can reuse available data across tasks without a clear restriction on why it may do so, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Data-minimisation review
This area examines a data field, document segment or retained context element used for a defined task. It helps decision owners identify the evidence needed when agents receive broad records or full conversation histories for relatively narrow tasks, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
AI data-retention design
This area examines a retained data class with a purpose, owner, expiry condition and dependent copies. It helps decision owners identify the evidence needed when retention policies exist but agent prompts, memory, caches and logs have different lifecycles, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Memory privacy review
This area examines a memory item from creation through retrieval, inference, correction and deletion. It helps decision owners identify the evidence needed when persistent memory is introduced, changed or questioned after a correction or cross-user concern, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Consent management for AI interactions
This area examines a user choice or consent state governing a defined processing activity. It helps decision owners identify the evidence needed when users can express choices about AI processing but the effect on tools and retained state is unclear, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Cross-border AI data-flow assessment
This area examines a data flow or remote access path across declared processing and storage locations. It helps decision owners identify the evidence needed when an agent uses providers, subprocessors or support arrangements in multiple locations, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Model-provider privacy assessment
This area examines a particular provider service, account arrangement, configuration and data path. It helps decision owners identify the evidence needed when a model provider is selected or changed for an agent handling personal information, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Training-data exposure analysis
This area examines a training or adaptation data source and its potential relationship to model outputs. It helps decision owners identify the evidence needed when personal or confidential data may have entered model training or adaptation datasets, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Inference privacy review
This area examines a request from context assembly through provider processing, tool calls and response delivery. It helps decision owners identify the evidence needed when an agent sends personal information to models and tools during live inference, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
PII detection and redaction
This area examines a sensitive span or attribute processed through detection, transformation and downstream use. It helps decision owners identify the evidence needed when personal information must be detected or removed before agent processing or external transfer, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Privacy-preserving retrieval
This area examines a retrieval request and returned context under a specific identity, purpose and source state. It helps decision owners identify the evidence needed when an agent retrieves relevant material from sources with different access and purpose restrictions, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Tenant privacy isolation
This area examines a tenant-scoped request across identity, storage, retrieval, memory and output boundaries. It helps decision owners identify the evidence needed when agents share models, indexes, caches or operating services across customer tenants, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Right-to-access and deletion workflows
This area examines a verified request linked to relevant data stores, transformations and authorised response steps. It helps decision owners identify the evidence needed when an access or deletion request must reach agent memory, derived records and downstream systems, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Privacy leakage red teaming
This area examines a controlled adversarial case with synthetic targets, allowed techniques and an evidence path. It helps decision owners identify the evidence needed when a sensitive agent workflow needs adversarial examination beyond ordinary functional tests, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Confidential-computing assessment
This area examines a defined computation and trust boundary including inputs, execution, attestation and outputs. It helps decision owners identify the evidence needed when a protected-compute proposal is being considered for agent or model processing, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Synthetic-data strategy
This area examines a synthetic dataset designed for a specific development, evaluation or sharing purpose. It helps decision owners identify the evidence needed when representative data is needed for development or evaluation but real records are sensitive, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
Differential privacy advisory
This area examines a defined data release or analysis with a privacy unit, mechanism and accounting model. It helps decision owners identify the evidence needed when a project proposes differential privacy without a clear protected unit, release mechanism or utility requirement, keeping the analysis tied to a specific business decision.
- Privacy engineering for AI systems
AI privacy control validation
This area examines a stated privacy control mapped to implementation, test conditions and observed behaviour. It helps decision owners identify the evidence needed when privacy requirements have been translated into controls but their runtime behaviour is unevidenced, keeping the analysis tied to a specific business decision.
- Measuring AI impact
AI value measurement design
This area examines a workflow-level value hypothesis with an outcome, comparator and accountable beneficiary. It helps decision owners identify the evidence needed when several stakeholders use different definitions of success for the same agent initiative, keeping the analysis tied to a specific business decision.
- Measuring AI impact
AI program baseline assessment
This area examines a prechange workflow population measured over a declared period and case mix. It helps decision owners identify the evidence needed when an AI rollout is approaching without a reliable picture of current workflow performance, keeping the analysis tied to a specific business decision.
- Measuring AI impact
Productivity-impact measurement
This area examines a quality-accepted unit of work including preparation, review, correction and exception handling. It helps decision owners identify the evidence needed when an agent appears to accelerate individual tasks but the effect on overall work is unclear, keeping the analysis tied to a specific business decision.
- Measuring AI impact
Workflow outcome measurement
This area examines a business case from entry to accepted completion, including downstream and reopened work. It helps decision owners identify the evidence needed when component metrics improve while customers or downstream teams still experience incomplete work, keeping the analysis tied to a specific business decision.
- Measuring AI impact
AI quality-adjusted ROI
This area examines a net-value model tied to accepted outcomes and a declared financial comparison. It helps decision owners identify the evidence needed when an ROI narrative counts gross benefits but omits rejected outputs, review and operating burden, keeping the analysis tied to a specific business decision.
- Measuring AI impact
Human-effort and review-cost measurement
This area examines a human activity supporting or correcting one agent-assisted case. It helps decision owners identify the evidence needed when AI cost estimates omit checking, escalation and the effort required to repair agent output, keeping the analysis tied to a specific business decision.
- Measuring AI impact
Failure-cost modelling
This area examines a failure event with a causal path, operational consequence and cost boundary. It helps decision owners identify the evidence needed when an AI proposal includes expected benefits but treats failures as a single undefined contingency, keeping the analysis tied to a specific business decision.
- Measuring AI impact
AI total-cost-of-ownership analysis
This area examines a deployment option over a declared planning horizon and cost boundary. It helps decision owners identify the evidence needed when vendor usage charges are known but integration, operation, change and exit costs are not, keeping the analysis tied to a specific business decision.
- Measuring AI impact
Agent unit-economics assessment
This area examines a completed, accepted workflow unit including unsuccessful attempts and required human work. It helps decision owners identify the evidence needed when cost-per-call dashboards do not explain whether useful work is economically viable, keeping the analysis tied to a specific business decision.
- Measuring AI impact
Adoption and utilization measurement
This area examines a user or team completing an eligible workflow with declared agent assistance. It helps decision owners identify the evidence needed when licence activation or login counts are being reported as evidence of successful adoption, keeping the analysis tied to a specific business decision.
- Measuring AI impact
Customer-experience impact
This area examines a customer journey through resolution, follow-up and relevant handoffs. It helps decision owners identify the evidence needed when internal efficiency appears to improve but the effect on customer effort and resolution is uncertain, keeping the analysis tied to a specific business decision.
- Measuring AI impact
Employee-experience impact
This area examines a role's experience of workload, control, skills and responsibility within a defined process. It helps decision owners identify the evidence needed when an AI rollout changes tasks and review duties without clear evidence of its effect on everyday work, keeping the analysis tied to a specific business decision.
- Measuring AI impact
Risk-adjusted value measurement
This area examines a value scenario combining benefit, operating cost, control effort and downside exposure. It helps decision owners identify the evidence needed when an AI proposal has attractive expected benefits but material uncertainty about failure and control costs, keeping the analysis tied to a specific business decision.
- Measuring AI impact
Time-to-decision measurement
This area examines a decision episode from a defined trigger to an accepted decision and relevant revision. It helps decision owners identify the evidence needed when an agent accelerates information preparation but the time to an accountable decision remains unchanged, keeping the analysis tied to a specific business decision.
- Measuring AI impact
Revenue-attribution design
This area examines a defined commercial outcome and exposure condition within an attribution design. It helps decision owners identify the evidence needed when an AI sales or customer initiative is credited with growth without a credible comparison, keeping the analysis tied to a specific business decision.
- Measuring AI impact
Cost-avoidance validation
This area examines a claimed future cost avoided against an evidenced alternative plan. It helps decision owners identify the evidence needed when an AI program claims savings from expenditure that has not actually been incurred or removed, keeping the analysis tied to a specific business decision.
- Measuring AI impact
AI portfolio value realization
This area examines a portfolio initiative and its contribution to shared business outcomes and dependencies. It helps decision owners identify the evidence needed when many AI initiatives report benefits independently with overlapping assumptions and uneven evidence, keeping the analysis tied to a specific business decision.
- Measuring AI impact
AI impact dashboard design
This area examines a decision-oriented dashboard measure with a definition, source, owner and threshold. It helps decision owners identify the evidence needed when AI reporting shows usage and charts without agreed definitions or an action path, keeping the analysis tied to a specific business decision.
- Measuring AI impact
Benefits-realization assurance
This area examines a reported benefit traced to its baseline, realization mechanism and observed outcome. It helps decision owners identify the evidence needed when an AI program reports delivery completion and expected benefits as though they were the same thing, keeping the analysis tied to a specific business decision.
- Measuring AI impact
Independent AI investment review
This area examines a bounded investment proposition assessed against sponsor criteria and available evidence. It helps decision owners identify the evidence needed when an existing AI investment case needs an external challenge or renewal decision, keeping the analysis tied to a specific business decision.
- AI product-market fit and market research
AI opportunity and problem discovery
Decide which problem an AI product should address, and for whom. The work examines the work as it is done now, its cost, its failure points and who feels them, for founders and product leaders shaping an AI product direction, and produces an opportunity register with evidence, severity ratings and a recommended focus.
- AI product-market fit and market research
AI product market research
Decide whether the market for a proposed AI product is reachable, competitive and worth entering. The work examines who holds the budget, what they use today, what a switch would cost them and what would make them look, for product and commercial leaders assessing a market entry, and produces a market research report with segment definition, alternatives, reachability and price evidence.
- AI product-market fit and market research
Customer and workflow research
Decide how the target workflow actually runs and where an AI product could fit. The work examines the steps, the decisions, the inputs people gather informally and the consequences of each type of error, for product managers and designers grounding an AI product in real work, and produces a workflow model with observed decision points, exceptions and candidate intervention points.
- AI product-market fit and market research
AI product positioning and proposition design
Decide the product's position, its comparison set and the claim it can honestly make. The work examines the comparison the buyer will make anyway, and whether the proposition wins or loses it honestly, for product marketing and founding teams, and produces a positioning statement with a comparison set, substantiated claims and buyer-tested language.
- AI product-market fit and market research
Product-market-fit hypothesis design
Decide what result would confirm fit, and what result would refute it. The work examines whether each belief is stated so that a specific observation could refute it, for founders and product leaders running towards a fit decision, and produces a fit hypothesis set with thresholds, disconfirming results and a measurement design.
- AI product-market fit and market research
AI pilot and experiment design
Decide what a pilot must measure for its result to justify the next commitment. The work examines whether the pilot's population, case mix and measures could produce a result that changes the decision, for product leaders and sponsors preparing a pilot, and produces a pilot design with a named decision, representative population, measures and stopping rules.
- AI product-market fit and market research
Adoption-barrier and buyer-objection analysis
Decide which barriers actually block adoption and which are surface objections. The work examines where each stalled case actually stopped, who stopped it and what they needed that was missing, for commercial and product leaders facing stalled adoption, and produces a barrier analysis with reconstructed stalls, operative causes and ranked responses.
- AI product-market fit and market research
AI product launch and market-readiness assessment
Decide whether a product is ready to face buyers, and what must be true first. The work examines each claim the launch will make and each promise the organisation will have to keep afterwards, for product and commercial leaders approaching a launch, and produces a launch readiness review with claim-by-claim evidence and blocking gaps.
- Security software engineering
Secure software architecture and threat modelling
Agree which trust boundaries and attack paths the architecture must defend. The work examines how data, credentials and privileged operations cross each boundary in the intended deployment, for software architects and product security leads, and produces a threat model with ranked attack paths and the architectural mitigations attached to them.
- Security software engineering
Secure development lifecycle and DevSecOps
Decide which security activities belong in the delivery pipeline and which need human judgement. The work examines where a security activity changes a decision, and where it only produces an unread artefact, for engineering leaders and platform teams, and produces a secure development lifecycle design with placed controls, ownership and exception handling.
- Security software engineering
Product security requirements and security-by-design
State the product's security requirements in terms a test can confirm or refute. The work examines whether each stated security property names an actor, a condition and an observable outcome, for product managers and security architects, and produces a product security requirement set with verification methods and owners.
- Security software engineering
Application, API and integration security engineering
Decide how each interface enforces authorisation and handles untrusted input. The work examines whether each operation checks the caller's authority at the point where the effect occurs, for application engineers and integration owners, and produces an interface security review with enforcement changes and regression cases.
- Security software engineering
Secure code review and vulnerability remediation
Decide which code weaknesses matter in this deployment and in what order to fix them. The work examines reachability, exploitability and blast radius rather than the severity label a tool assigned, for engineering managers and security engineers, and produces a prioritised remediation plan with reachability evidence for each finding.
- Security software engineering
Software supply-chain security and dependency control
Decide what the build is allowed to consume and what evidence the release must carry. The work examines every input the build consumes, every identity that can change it and every claim a consumer can independently verify, for platform engineers and release owners, and produces a supply-chain control design with an inventory, provenance and intake rules.
- Security software engineering
Embedded and IoT security engineering
Decide which device threats the hardware and software design must resist in the field. The work examines physical access, debug interfaces, storage, radio surfaces and what a single compromised device reveals about the rest of the fleet, for embedded engineering leads and product security owners, and produces a device security design with a stated attacker model and fleet-level consequences.
- Security software engineering
Firmware security, secure boot and signed updates
Decide what the device verifies before it executes code, and how an update is authorised. The work examines each transition between boot stages, the root of trust it depends on, and the device's behaviour when verification fails, for firmware engineers and device platform owners, and produces a boot and update architecture review with verification gaps and custody decisions.
- Security software engineering
Security protocol design, review and implementation
Decide whether a protocol's design and implementation deliver the properties it claims. The work examines message authentication, replay, ordering, downgrade, session binding and the behaviour of each party on an unexpected message, for protocol designers and platform security engineers, and produces a protocol review with property claims, attacker-path findings and implementation observations.
- Security software engineering
Identity, authentication, authorisation and secure-session engineering
Decide how identity is proved, carried and revoked across the product's components. The work examines how an identity assertion is produced, validated, propagated, narrowed and revoked across every hop, for platform engineers and identity owners, and produces an identity and session design review with propagation, scope and revocation findings.
- Security software engineering
Security testing and release assurance
Decide what security evidence a specific release requires before it ships. The work examines whether the available evidence covers the code and configuration actually being released, for release owners and security assurance leads, and produces a release security evidence pack with coverage, residual findings and named dispositions.
- Cryptographic engineering
Applied cryptography architecture
Decide which cryptographic mechanisms protect which data against which adversary. The work examines what each mechanism protects against, who holds the key, and where plaintext still exists, for security architects and engineering leads, and produces a cryptographic architecture assessment with mechanism decisions and residual exposure.
- Cryptographic engineering
Key management, PKI and HSM engineering
Decide how keys and certificates are generated, held, rotated and revoked across their lifetime. The work examines every stage of the lifecycle, including the stages nobody has had to exercise yet, for platform security engineers and PKI owners, and produces a key-management architecture review with an inventory, lifecycle findings and operable procedures.
- Cryptographic engineering
Cryptographic agility assessment
Decide how quickly the system could change a cryptographic algorithm or trust anchor if it had to. The work examines how algorithm choice is expressed, how long ciphertext and signatures persist, and what a peer or a stored record forces you to keep supporting, for architects and engineering leaders planning for algorithm change, and produces a cryptographic agility assessment with a dependency inventory and change-cost ratings.
- Cryptographic engineering
Post-quantum cryptography readiness assessment
Decide which systems need post-quantum work first, and which do not need it yet. The work examines which mechanisms are quantum-vulnerable, how long their data must stay confidential, and which dependencies are outside your control, for security leaders and architects responding to post-quantum guidance, and produces a post-quantum readiness diagnostic with a vulnerable-mechanism inventory and a prioritised position.
- Cryptographic engineering
Post-quantum cryptography migration architecture and implementation
Decide the sequence, mechanism and rollback position for moving a system to quantum-resistant cryptography. The work examines how each system reaches the target state while remaining interoperable, reversible and measurable throughout, for engineering leads delivering a cryptographic migration, and produces a post-quantum migration blueprint with sequencing, interoperability evidence and rollback positions.
The bigger picture
Explore a capability family.
Read the broader context behind each specialist discipline.
- Agentic AI strategy and transformation
- Agentic transformation partnerships
- Leadership and organisational training
- Agentic system architecture and engineering
- Agent-to-agent interoperability
- Partnerships with agentic IaaS, PaaS, and SaaS companies
- Self-evolving and long-lived agent assurance
- Agent evaluation and assurance
- Component-level agent evaluation
- Distinctive research-led offerings
- Agent security and cybersecurity
- Agent identity, authorisation, and delegation
- Agent sandbox assurance
- AI governance and regulatory readiness
- Privacy engineering for AI systems
- Measuring AI impact
- AI product-market fit and market research
- Security software engineering
- Cryptographic engineering