Gate Review Preparation With AI: Check the Evidence, Then Rehearse the Questions
For gate review preparation with AI, run two checks on the pack before the meeting. First, turn each gate criterion into a yes-or-no question and have the AI answer it from your pack and its source documents, quoting the line behind each answer and flagging any figure that doesn’t match its source. Second, give it the minutes of the last two gate reviews and ask it to predict what each board member will ask, based on what they asked before. Put the results on one page: the likeliest questions, your answer, the evidence behind it, and the gaps you’ll raise yourself. It takes about an hour.
The Question You Wish You’d Seen Coming
You know the moment. Slide four, and the Finance Director asks how much contingency you’ve drawn since the last gate. The number is in the cost report. It isn’t in the pack, and she asked the same thing at the last two gates.
Your gate pack usually gets checked by the person who wrote it. You read the slides for what you meant them to say. The board reads them against the criteria, and against what they asked last time.
In You Already Have a Senior Advisor we suggested asking AI to play your most sceptical stakeholder, and the Red Team prompt in the PM Prompt Library still does that well. The assistant you already use can read a full pack and its source documents together, and answer a fixed list of questions about them. So the rehearsal can start from evidence instead of a persona.
Check the Evidence Against the Gate Criteria
Start with the criteria the board will use, because that’s the test the pack has to pass. Give the assistant the criteria, the pack (or the one-sentence-per-slide storyline from AI Builds the Deck in Minutes Now), and the sources the pack draws on: the cost report, schedule, RAID log and approvals.
Below are the criteria for our [gate name], our gate pack, and the source documents behind it. Turn each criterion into a yes-or-no question. Answer each one yes, no or not shown. For each answer, quote the line in the pack and the line in the source that supports it. Flag every figure or claim in the pack that doesn’t match its source, or has no source. Don’t judge whether the project should pass.
The last line keeps the tool on the evidence. You want a list of what the documents show, and the call on what that means stays with you.
Predict the Board’s Questions From Past Minutes
Then give it the minutes of the last two or three gates, in the same chat so it still has the first answer.
Below are the minutes of our last gate reviews. For each board member, list what they asked before and the themes they came back to. Predict the three questions each is most likely to ask at this gate, given the gaps you found. For each, say which slide or source answers it, or write “no answer in the pack”. Base every prediction on a line in the minutes or the pack, and say which one. Rank all the questions by how much damage a weak answer would do.
Asking for the line behind each prediction makes it harder for the tool to invent a plausible-sounding board. Keep the questions you can’t yet answer, and drop any that rest on nothing in the documents.
Check your organisation’s policy before you paste a gate pack, a cost report or board minutes into an AI tool. Minutes record what named people said, so an assistant inside your organisation’s own Microsoft 365 or Google Workspace is a better place for this than a personal account.
Put the Answers on One Page Before the Gate Review
Finish with one entry per question, ranked, so the whole set fits on a page you can take into the room:
Question: How much contingency has been drawn since Gate 2? Asked by: Finance Director (Gate 1 and Gate 2 minutes) Answer: $0.26m of $0.40m, for data migration rework and an extra design sprint Proof: cost report, 20 October Raise it ourselves: yes, on the budget slide
Now fill in the same five lines for each question on your own list. Fix the pack where you can, and for each gap you can’t close in time, decide whether to raise it first. A gap you name on slide three, with a plan beside it, shows the board you’re in control of it. The same gap found by the board on slide eight becomes the thing they remember.
The criteria and minutes for your next gate already exist. An hour with both, a week out, turns the board’s likeliest questions into a page you’ve answered.
What did the board ask at your last gate review that you’d have liked to see coming two days earlier?
Frequently Asked Questions
How can I use AI to prepare for a gate review? Have it check your pack against each gate criterion as a yes-or-no question, quoting the evidence, then predict the board’s questions from the minutes of past gates. Prepare answers for the questions your pack can’t answer.
What should I give AI to check a gate pack? The gate criteria, the pack or its storyline, the source documents the pack draws on, and the minutes of the last two or three gate reviews.
Can AI predict what a steering committee will ask? It can make a useful forecast when each prediction rests on something a board member asked before or a gap in the pack. Treat the result as a rehearsal, and ask for the line behind every prediction.
What did the board ask at your last gate review that you’d have liked to see coming two days earlier?
Yes, AI helped me to write this :)