Use case 02 · Product Management
The evidence exists. Nothing connects it to the decision it was supposed to inform.
Interviews sit in one tool, tickets in another, the roadmap in a third, and the reason a feature was cut lives in a call nobody recorded. Six weeks later the same debate reopens, and no one can say what changed — the evidence or the argument.
For a product leader whose team runs real research and still argues from memory in prioritization.
PDF · the workflow, the review points, and what a first bounded case covers
The workflow
What the workflow becomes
Prioritization stays a human judgment call. What changes is that the judgment is made against an assembled record instead of whoever remembers most vividly.
One product record
People carry this
Customer evidence, the current decisions, and the requirements they produced live in one place with the link between them intact. A decision without its evidence is visibly incomplete.
AI agents assemble the evidence
AI agents carry this
AI agents gather what was said, group it, and attach each finding to its source — the call, the ticket, the message. They surface where the evidence contradicts itself instead of averaging it away.
Named people set priority
People carry this
Priority is set by the person accountable for the outcome, reading the assembled evidence. The AI agent can argue against the ranking; it does not hold it.
The decision keeps its reason
People and AI agents together
What was decided, by whom, on what evidence, and what would change it are recorded together. When the debate reopens, it reopens against the record.
Where the manual work goes
Where the manual work goes
Each row is work a person is doing right now because the workflow does not do it. That is the cost the first case is measured against.
| Problem | What a person does today | What the workflow change does |
|---|---|---|
| Evidence in four tools | Re-read everything before each planning cycle. | Assembled once, attached to its source, kept current. |
| The same debate, twice | Argue from whoever remembers most confidently. | The prior decision carries its evidence and its reversal condition. |
| Contradictory research | Quietly average it into a compromise. | The contradiction is surfaced as a contradiction and given an owner. |
| Requirements drift from evidence | Discover the gap during build. | The requirement links to the finding it came from. |
The working result
What you end up holding.
A product record where evidence, priority and decision are connected; a list of the contradictions the evidence actually contains; and, for each live decision, the condition that would change it.
What this does not claim
This describes a workflow design. It is not a claim that AI agents should set product priority, and it is not a measured delivery outcome. Prioritization stays with the accountable person by design, not as a temporary caution.
A bounded first case
Bring us this workflow.
Twenty minutes, one recurring workflow, and an honest answer about whether a bounded first case is worth running. If it is not, we will say so.
PDF · the workflow, the review points, and what a first bounded case covers