Boards rarely ask which software was used to type a SteerCo pack. They asked whether the numbers were right and who would answer for them. AI does not change those two questions. It changes how easy it is to get the first one wrong while sounding sure. So the disclosure is not an apology. It is a statement of who checked what. This is governance practice from a practitioner, not legal advice; your organisation’s own policy on AI use comes first.
Why hiding it does not work
Someone on the board already uses these tools. They recognise the cadence. If they ask and you hedge, you have lost the pack. If they find out later, you have lost the next six. The only safe position is the one you would be happy to have read back to you in a review: “yes, the narrative was drafted with an AI tool from our plan and RAID log, I checked every figure against the source, and I sign it.”
Note what that sentence does not say. It does not say the tool is accurate. It does not say the tool approved anything. It says a named person did the checking and takes the pack as their own.
The one-line disclosure
Put it on the first or last page, in the same place every month. A form that works:
Narrative sections of this pack were drafted with an AI tool from the programme plan, RAID log and budget tracker frozen on 24 October. All figures, dates and ratings were checked against those sources by [name], [role], on 27 October. Two consistency exceptions were accepted and are listed on page 9. This pack is issued by [name] and the decisions in it are owned by [sponsor].
Three sentences. What the tool did. What a person did, against which source, on which date. Who signs and who owns. If your organisation has a policy that requires more, follow it. If it has none, this is a good start, and you should raise the gap with whoever owns policy.
The three questions the board actually has
- Can I rely on these numbers? Answer with the check, not the tool: “every figure was checked against the frozen source by a named person”. The method for doing that is in What checks should I run before I issue a project report?
- What did the tool get wrong, and what did you do about it? This is the question that builds trust if you answer it well. “The draft carried an older go-live date from the previous plan. The check caught it and it was corrected to 14 March.” A board that hears this once stops worrying about the tool and starts trusting the process.
- Who do I hold to account? A name. Yours, for the pack. The sponsor’s, for the decisions. Never “the system”.
Who is accountable, in plain terms
- The project manager who issues the pack owns every figure, date and rating in it, drafted by a tool or not. Issuing is the act of taking ownership.
- The sponsor or chair owns the decisions the pack asks for and the ones it records.
- Whoever accepted an exception owns that exception: the choice to issue with a known mismatch, and the reason given. That should be a named role with authority, not whoever happened to be at the keyboard.
- The tool owns nothing. It is a drafting aid and, if it is a good one, a checking aid. It cannot be questioned by a board, cannot be disciplined and cannot resign. Anything that shifts accountability onto it has shifted it nowhere.
This is the same principle as the one in Can AI make up numbers in a project report?: the tool is confident, so the person has to be careful.
Keep a record you could hand over
Six months on, a review or an audit may ask how the March pack was produced. You want to be able to hand over, without searching: the source files as frozen, the issued pack, the list of what the checks found, what was fixed, what was accepted and by whom with the reason, and the disclosure line. If your tool keeps that record for you, good. If it does not, keep it yourself in a folder per period. The record is what turns “we used AI” from a confession into a controlled process.
A real example
A programme director I worked with added the disclosure line to her monthly pack in the first month her team used an AI drafting tool. The chair read it aloud and asked question two: what did it get wrong? She had the answer ready: a budget forecast rounded to the wrong line, caught at check and corrected. The chair said “good” and moved on. Nobody on that board asked about AI again for a year. The line, and the ready answer, were the whole of the governance.
Where PMOpack fits
PMOpack is honest that AI drafts the narrative in the pack. What it adds is the record a board can be shown. Fixed checks, which are not AI, run across the pack at generation time and flag where artefacts in the same period disagree. A critical finding blocks publish until a person fixes the data or accepts it with a written reason. Accepting is an act by a workspace owner or admin, and the acceptance is stored with the name and the reason and cannot be edited later. That is the exception list for your disclosure line, kept for you. The published pack carries a hash that shows the file matches the record; it does not show that the facts are true, and PMOpack does not approve or assure a report. A person issues the pack. A person is accountable for it. PMOpack makes what that person flagged, fixed and accepted visible.
Questions people ask
- Do I have to tell the steering committee that AI helped write the pack?
- Your organisation's policy decides. As a matter of good governance, yes: one plain line in the pack costs nothing and removes the risk of the board finding out later and wondering what else was not said.
- Who is accountable if an AI-drafted figure in the pack is wrong?
- The person who issued the pack. The same person who would be accountable if they had typed it. A tool cannot be accountable to a board; a named person can.
- What should an AI disclosure line in a board pack say?
- Three things: what the tool drafted, what a named person checked and against which source, and who signs. One or two sentences on the first or last page.
- Should the board see what the checks found?
- They should see any open exception: a finding a person chose to accept rather than fix, with the written reason and the name. Findings that were fixed before issue can stay in the record.
- Does a tool's audit record make the PM less accountable?
- No. It makes the PM's judgement visible: what was flagged, what was fixed, what was accepted and why. That protects a careful PM. It does not replace one.