TL;DR: Choose a segment because its task and constraints lead to a different product decision. Name the person doing the work, the buyer and anyone affected by mistakes. Compare the current workaround, select one problem and say what observation would change your choice.
A persona must change the decision
“Busy professionals who like AI” does little to constrain a design. A maintenance coordinator taking over twelve unfinished repair requests at a shift change has a specific job: identify what remains open, what blocks it and who owns the next action.
This is a fictional teaching scenario. We have not interviewed these users. Segmentation is a hypothesis to investigate, not research performed by imagining a plausible person.
| Candidate segment | Circumstance that changes the design | First research question |
|---|---|---|
| Incoming coordinator | Must act on someone else's incomplete notes | Which missing fact causes a repeat call? |
| Outgoing technician | Has little time to record unfinished work | Which details are already captured elsewhere? |
| Operations manager | Purchases software and handles escalations | What evidence establishes a reliable handover? |
| Customer awaiting repair | Bears the cost of a missed commitment | How are changed promises communicated? |
These roles can overlap. Do not add them together as four independent markets. For the first version, choose the incoming coordinator, while treating the other roles as dependencies and people whose outcomes matter.
Compare problems before solutions
Suppose observation could reveal three different causes of delay: missing status, unavailable parts or no assigned technician. A summary may help with the first. It cannot manufacture a part or create staffing capacity.
Ask a coordinator to show the last difficult handover. Follow the current process through tickets, calls and spreadsheets, with permission. Record the repeated decision, the missing information and the consequence. Include someone who tried the existing tool and abandoned it; satisfied users alone give an incomplete picture.
For this fixture, assume stale status is the main problem. Compare a required handover checklist with a source-linked AI draft. Both should preserve unresolved questions. The draft earns further investment only if it reduces the total work of producing and verifying an accurate handover.
Follow the decision through the workflow
Follow the flow, one stage at a time.
Illustrated sequence; animation speed does not represent processing time.Read diagram description
Name the business goal, separate the people involved, observe the current job, compare problems, choose the first segment, compare interventions, test the whole handover and revise the commitment. New evidence returns the team to observation.
Choose a first scope without erasing other people
An initial pilot might cover routine repair requests whose source records are accessible to both shifts. Display timestamps, sources and unresolved items; the incoming coordinator acknowledges ownership. A generated sentence must not silently mark a repair complete or assign someone work they have not accepted.
That scope excludes emergency dispatch. It also needs testing with different language, input and accessibility needs. Ask participants about their actual constraints rather than assuming a demographic group needs one simplified interface.
Measure the share of eligible handovers that reach a correct next action without a repeat clarification, with a defined observation window. Check missed commitments, verification time and unacknowledged work separately. Summary opens are a diagnostic signal, not the outcome.
The interview answer
Explain why this segment deserves the first investment, name the strongest alternative and identify the evidence you lack. If observation shows that parts shortages dominate delays, redirect the proposal toward availability and coordination. Keeping the original AI summary would require a separate value case.
Practise designing a shift-handover assistant, then use the AI fit test to compare the chosen intervention with a simpler baseline. Sizing an AI opportunity turns the task into a value estimate.