How do I ask AI to find gaps in a plan?

Watch the briefing

Watch the full video on YouTube

Full briefing transcript

The craft-fair setup plan lists tables, signs, arrival times, and cleanup. It looks ready, but one stall needs powered equipment and nobody has confirmed venue-approved power access. Ask AI to find gaps against the actual goal and constraints, not merely to say whether the plan looks good. The early lesson is simple: a polished list can still omit a dependency that matters to whether the event can proceed.

Our fictional event is an indoor adult craft fair on October seventeenth, twenty twenty six, from ten until two. Six stalls are planned, and stall leads arrive at eight thirty. The organizer confirms the room booking, the layout lead marks positions, and the welcome lead prepares signs. Cleanup is assigned. The constraints say that one demonstration stall needs powered equipment and all arrangements must be approved by the venue.

Give the reviewer both the plan and those constraints. Ask for omissions, their possible impact, the person who must confirm them, and the evidence needed to close each gap. Do not ask the assistant to fill unknowns with confident instructions. In particular, our prompt prohibits invented venue approval and electrical setup advice. The assistant's job here is to expose a missing decision, not design an installation or declare the event safe.

Our illustrative gap review says the powered demonstration lacks a confirmed venue-approved power arrangement. The impact is that the demonstration might not be ready or permitted. The owner is the organizer working with the venue representative. Completion evidence is recorded confirmation of approved arrangements. We wrote this mock review for teaching, not as a model test. The status remains unconfirmed until the responsible people obtain and inspect that evidence.

Notice what the response must not do. It must not tell the organizer to run a cable across the room, select electrical equipment, or treat an outlet symbol as safety approval. Those would introduce a different and potentially risky task. Keep the visible arrow at venue and competent-person review. The organizer can decide to change or remove the demonstration if approval is unavailable; a draft review cannot make that decision on their behalf.

Use a four-part gap table: omission, impact, owner, and evidence. Compare each entry to the source plan and constraints. Some suggestions may be irrelevant, and the reviewer may miss other important omissions. Official prompting and evaluation guidance supports explicit task requirements and defined checks; this event example is our editorial adaptation. For real events, follow venue requirements and competent local advice. A general AI review is not a safety inspection or certification.

Ask for omissions against a defined goal and constraints. Our plan now shows an unresolved power dependency rather than hiding it behind a complete-looking checklist. The responsible organizer retains the confirmation and go-ahead decision. No event arrangement was changed and no technical instructions were supplied. Next, we turn a seed-swap idea into checks with owners, dependencies, and visible completion evidence.