There is an empty chair at the end of the meeting table. Before you send the proposal, imagine someone sitting in it. Not your biggest supporter. The person who will ask the question you hoped nobody would.

Perhaps it is the customer who does not understand the offer. The colleague who has to make it work on Monday morning. Or the budget holder who likes the idea but needs to know what happens if it fails.

That chair can be a useful place to put AI. Not as a fortune-teller, and certainly not as a substitute for talking to real people. As a rehearsal partner.

Give the chair a job

“What do you think?” is a comfortable question. It leaves plenty of room for a comfortable answer. A more useful brief gives the reviewer a particular responsibility and something concrete to inspect.

Imagine you are proposing a new online booking system. Ask for a review from the perspective of a receptionist handling a queue, then from a customer booking on a small phone screen. One might ask how to correct a mistaken booking. The other might ask why an account is needed before they can see an available time.

Neither response proves what real users will think. But each gives you a question worth investigating before you build.

Ask for the awkward questions, not a verdict

Try a brief like this, using a draft you are authorised to share:

Review this proposal from the perspective of the person responsible for putting it into practice. Ask five specific questions that the draft does not answer. For each question, point to the relevant passage or missing detail. Separate problems you can identify in the text from assumptions we need to check with real people. Do not invent evidence or rewrite the proposal yet.

The last sentence matters. If you ask for a rewrite immediately, a polished paragraph can cover up a decision you have not actually made. Find the gap first. Decide what belongs there. Then improve the wording.

Leave room for disagreement

Now change the occupant. A finance reviewer might care about ongoing costs. A first-time customer might care about whether the next step is clear. The person delivering the service might care about exceptions that never appear in the happy-path demonstration.

You do not need to satisfy every imagined objection. Some will be irrelevant; others may rely on an incorrect assumption. Keep a short list: change this, check this, leave this alone. That turns a pile of criticism into a manageable next step.

A rehearsal is not an audience

AI can miss an obvious problem or confidently suggest one that does not exist. Three simulated perspectives are still outputs from a model, not three independent customer interviews. Take the useful questions to the people who can answer them.

The empty chair earns its place when it helps you arrive better prepared: a clearer offer, a more honest limitation, a practical answer to “what happens next?”

At DevConductor, that is the kind of work we want AI to support. Not just making something sound ready. Helping you work out whether it is ready.

Have a proposal, website or idea that could use a rehearsal? Tell us what you are preparing. We will help you find the questions worth asking.

The DevConductor Team