AI opportunity and problem discovery
Decide which problem an AI product should address, and for whom
Several AI ideas are competing for the same engineering capacity and none has been tested against a real problem. The decision is which problem an AI product should address, and for whom. The work starts from the work as it is done now, its cost, its failure points and who feels them and ends with an opportunity register with evidence, severity ratings and a recommended focus. It is most relevant to founders and product leaders shaping an AI product direction. A useful result states what was examined, what remains uncertain and which decision the evidence can support.
Decide which problem an AI product should address, and for whom
- Candidate problem
- The starting condition this work depends on.
- Observed cost
- Where the consequence actually arises.
- Selection rationale
- The finding and its open dependencies.
Decision
Frame the decision before scoping the work
Questions and the evidence each one needs
- Candidate problem
- Inspect the current product hypotheses and their origin.
- Observed cost
- Inspect access to people who do the work today.
- Selection rationale
- Inspect existing support, sales or usage signals.
The decision is which problem an AI product should address, and for whom. This matters to founders and product leaders shaping an AI product direction when several AI ideas are competing for the same engineering capacity and none has been tested against a real problem. Name the affected proposition, the consequence of a wrong decision and the owner who can act. The opening conversation should establish which evidence already exists and which uncertainty is material enough to investigate.
The specific failure hypothesis is that the idea describes a capability the technology makes possible rather than a problem someone is paying to solve. It stays a hypothesis until it is examined in your own case. Existing research notes, usage data or customer records may already resolve it, so identify the uncovered question before commissioning further work.
Mechanism
Follow the relationship that carries the consequence
Where the boundary is actually crossed
- Candidate problem
- The condition the design assumes.
- Observed cost
- Where authority or information changes hands.
- Selection rationale
- The effect the owner must answer for.
What this specialty examines
AI opportunity and problem discovery rests on the work as it is done now, its cost, its failure points and who feels them. Work from what people actually did and said rather than from the plan’s description of them, and record where the two differ.
The examination draws on the current product hypotheses and their origin; access to people who do the work today; and existing support, sales or usage signals. Identify the owner, revision and operating context of each source. Keep an incomplete input visible, with an explanation of what its absence prevents the owner from concluding, rather than quietly treating a gap as a satisfactory result.
The tradeoff to make explicit
The tradeoff here is a broad opportunity with many possible buyers against a narrow one you can serve completely. Compare the available options against the same intended outcome. If one option removes a control or moves a boundary, describe that difference before treating its lower effort as an improvement. The decision should expose the tradeoff rather than bury it in a single rating.
Preparation
Connect the inputs to the boundary examined
Preparation record for the engagement
Illustrative working format
- The current product hypotheses and their origin
- Bounds the context.
- Access to people who do the work today
- Supplies the material examined.
- Existing support, sales or usage signals
- Exposes conditions that qualify a result.
Start with the current product hypotheses and their origin; access to people who do the work today; and existing support, sales or usage signals. Each has a different job: the first establishes context, the second supplies the material to examine, and the third exposes operating conditions that may qualify a conclusion. Sanitised examples can support early work, provided the differences from the real configuration stay explicit.
The scope also needs founders and product leaders shaping an AI product direction with the authority to reach the relevant participants, and agreement on what evidence would distinguish the failure in which the idea describes a capability the technology makes possible rather than a problem someone is paying to solve from acceptable behaviour. Where stakeholders disagree about intended behaviour, resolve that interpretation before classifying an observed difference as a defect.
Working sequence
Carry the question through to an accountable finding
From scoped material to a decision record
- Collect the candidate problems
Set the conditions assumed.
- Observe the work
Apply the scoped check.
- Rate and select
Record evidence and the next decision.
Collect the candidate problems
Gather the existing hypotheses and their sources, separating an observed complaint from an internally generated idea. Record what is deliberately excluded.
Observe the work
Interview and, where possible, watch the people doing the task, recording the steps, the workarounds, the rework and what each failure costs. Keep an unreproducible observation in the record with its reason.
Rate and select
Compare candidates by severity, frequency, willingness to pay, reachability of the buyer and the team’s ability to serve them. Hand the result over with its open dependencies attached.
Indicative artefact
Make the artefact usable by the person who owns the decision
An opportunity register with evidence, severity ratings and a recommended focus
Illustrative working format
- Candidate problems and their evidential basis
- What was examined, and under which assumptions.
- Observed workflow cost and failure points
- Each observation, linked to evidence.
- Comparative ratings and a recommended focus
- The next decision, and its limits.
The proposed decision artefact is an opportunity register with evidence, severity ratings and a recommended focus. Depending on the agreed scope, the pack may include:
-
Candidate problems and their evidential basis. Basis: the definitions, boundaries and assumptions another reviewer would need.
-
Observed workflow cost and failure points. Observation: the findings and their interpretation, preserving adverse results and disagreement.
-
Comparative ratings and a recommended focus. Decision: what the owner could do next, and what change would justify revisiting it.
Retain the selection rationale beside the result. Without it, a bounded finding is easily mistaken for complete coverage.
Illustrative scenario
Examine the question through a constructed case
A constructed case and its decision paths
Illustrative scenario
- Candidate problem
A team plans an AI assistant for a task that practitioners say takes them four minutes a week.
- Observed cost
How severe and frequent is this problem compared with the ones they raised unprompted?
- Selection rationale
Redirect the effort to the problem they described without being asked, or test the original assumption directly.
Consider an illustrative situation: a team plans an AI assistant for a task that practitioners say takes them four minutes a week. The example is constructed to show how the question is framed. No client engagement, executed test or measured improvement is implied.
The question to examine is how severe and frequent is this problem compared with the ones they raised unprompted. A reasonable next action would be to redirect the effort to the problem they described without being asked, or test the original assumption directly. The decision record should state which configuration was accepted and which limitations came with it.
Scope and next action
Keep the conclusion inside the agreed boundary
Confirm fit before the next commitment
- Candidate problem
- Limited to the agreed scope and sample.
- The current product hypotheses and their origin
- The owner confirms access and conditions.
- Selection rationale
- Discovery produces evidence about a problem, not a forecast of revenue or a guarantee that a market exists.
Discovery produces evidence about a problem, not a forecast of revenue or a guarantee that a market exists. Any engagement depends on an agreed scope, a qualified team and authorised access.
Bring the proposition, the decision you face and the evidence you already hold. The scope question follows from the situation in which several AI ideas are competing for the same engineering capacity and none has been tested against a real problem. Confirm which missing observation would change the owner’s next action.