Reviewing a board or management pack before the meeting
Turning a board pack into the five questions worth asking
A prompt that reads a management pack the way a chief financial officer reads it, and gives you the questions rather than a summary.
The video for this one is not recorded yet.
The prompt below is finished and works on its own. Set videoUrl, published and duration in the entry to put the recording here.
The prompt
Reviewing a board or management pack before the meeting
You are an experienced group chief financial officer preparing to sit in a board meeting.
I am going to give you a management pack. Do not summarise it.
Your job is to find the five questions that most need asking, and nothing else.
Work through it in this order and show your working for each step:
1. Reconcile the numbers against each other. List any figure in the narrative that does
not agree with the same figure in the tables, and any total that does not add up.
2. Compare this period against the prior period and against budget. Identify the three
largest variances by absolute value and the three largest by percentage.
3. For each of those variances, state whether the pack explains it. If it does, say
whether the explanation is specific or generic. A generic explanation is one that
would fit any variance in any period.
4. Identify what has been left out. Look specifically for: cash conversion, working
capital movement, customer or supplier concentration, covenant headroom, and anything
that appeared in the prior pack but not in this one.
5. Note any change in presentation between periods: a rebased chart axis, a changed
comparative, a redefined measure, a new adjustment to a headline number.
Then give me exactly five questions.
Rules for the questions:
- Each one must be answerable with a specific number or a specific fact.
- Do not ask anything the pack already answers.
- Rank them so the first is the one I should ask if I only get to ask one.
- After each question, add one line stating what a satisfactory answer would contain,
so I know whether the answer I get is the answer to the question.
If any of your five questions rests on an inference rather than something stated in the
pack, say so explicitly and say what you inferred it from.Board packs are written to be presented, not interrogated. The narrative is prepared by the people whose performance it describes, which is not a criticism of them. It is simply what the document is for.
The reading that matters happens in the gaps: between the commentary and the tables, between this quarter and the last one, and between what is in the pack and what used to be in it. That reading is slow, it is repetitive, and it is exactly the sort of work worth handing to a machine.
How to use it
Paste the prompt, then attach the pack. Attach the previous period’s pack at the same time if you have it, because most of the useful signal is in the comparison.
What to watch for
The step that produces the most value is the fourth one, on what has been left out. It is also the step where a model is most likely to reach. Check every omission it claims before you raise it. The prompt asks the model to flag its own inferences for that reason.
Ranking the questions matters more than it looks. Board time is short and the fifth question usually does not get asked.
If a prompt gets you most of the way there and you want the rest of it built properly, that is the work.