A Board-Pack Narrative Template That Remembers Last Month

Writing the variance narrative is the part of a board pack that gets done last and fastest. This post gives you a working template for drafting it with AI, built around one idea most templates miss: the story should remember what you said last month.

It works with Claude or ChatGPT under a business account (see our data rules post), and it does not need any special software.

Why most AI commentary reads flat

AI commentary from a set of numbers alone tends to describe rather than explain: "Revenue rose against budget. Operating expenses were above plan." The board already sees that in the table. What they pay a CFO for is cause and consequence, and cause usually lives outside the ledger: a delayed hire, a customer that paid late, a price change you discussed with the CEO in last month's meeting.

So the template splits the work three ways:

1. Numbers: computed outside the AI, in your spreadsheet or reporting tool.
2. Causes: tagged by the person who knows the client.
3. Narrative: drafted by AI from the numbers, the causes and last month's notes, then reviewed.

What you need for each client

• This month's closed P&L, balance sheet and cash flow, with budget and prior month columns.
• The current forecast or reforecast, with its key assumptions written as short sentences.
• A client memory file: last month's narrative, the actions agreed at the last meeting and any open questions. Keep it as one short document per client and update it after every meeting.
• Your house style notes: tone, length, words you avoid, how you format figures.

The memory file is the part people skip and the part that makes the draft useful.

Step 1: compute the variances yourself

Do not ask the AI to do arithmetic on your behalf. Build a small table in your spreadsheet or reporting tool:

Sort by the size of the variance and keep the top five to eight lines. Everything below that threshold is noise for a board.

Step 2: tag the causes

For each line in the table, the analyst or CFO who knows the client adds one short cause tag: "timing: customer paid in next month," "hiring delayed to Q3," "one-off legal fee," "price increase effective [date]." If the cause is unknown, write "unknown, ask CEO." That tag becomes a board question rather than a guess.

This step takes a few minutes and is the difference between a narrative and a description.

Step 3: the drafting prompt

Paste the table, the tags, the forecast assumptions and the memory file, then use a prompt along these lines:

You are drafting the monthly board narrative for [client name], a [one-line description of the business]. Use only the figures in the variance table. Do not calculate new figures and do not round differently from the table. If a figure you need is missing, write [MISSING] instead.

Write:

1. An executive summary of 150 to 250 words: the three things the board most needs to know this month.
2. A short section for each variance line: what moved, the tagged cause, and what it means for the forecast.
3. A paragraph that reports progress on each action in the client memory file, saying plainly which ones are done, late or not started.
4. Three to five questions the board is likely to ask, with the answer if it is in the material and "to confirm" if not.

Follow the house style notes. Write for a CEO who reads the summary and skims the rest.

A Claude consulting firm publishes a similar flow (load statements, run a variance prompt against budget, write a 200 to 300 word executive summary, then section commentary, then review) and recommends a CFO review of 15 to 30 minutes per pack (ClaudeImplementation, updated Apr 2026). Their time-saving estimates are their own and unmeasured, so we don't repeat them here.

Step 4: the review checklist

Every draft gets checked by a CFO before the client sees it. Fathom reports that in its own survey of 20 presenters, all of them reviewed AI commentary before sharing it externally (Fathom, vendor figure). Make it quick with a fixed checklist:

• Every figure in the text matches the variance table.
• No cause is stated that wasn't tagged by a person.
• Every [MISSING] has been filled or removed.
• Last month's actions are reported accurately.
• Nothing in the text would surprise the CEO in the meeting.
• Tone matches the client: some boards want blunt, some want gentle.

Step 5: close the loop

After the board meeting, update the client memory file with the new actions and anything the board asked that you couldn't answer. That file is next month's starting point. Over a few months the drafts get noticeably more specific, because the memory gets richer.

Common failures and fixes

• The draft invents a cause. Your prompt allowed it. Add "Only use causes from the tags; if a variance has no tag, say the cause is to be confirmed."
• The numbers drift. The AI recalculated or rounded. Keep the "use only figures in the table" line and check figures first in review.
• Every month reads the same. The memory file is thin. Add the actual questions the board asked and the CEO's own words where you can.
• It's too long. Set word limits per section and say who reads it.

Where a reporting tool's built-in AI fits

If you use Fathom Pro, its AI Commentary Writer is included in your plan (Fathom pricing, fetched Oct 2026). Try it on the same client and compare. The built-in tool sees your ledger data well. What it can't see is your memory file, the forecast assumptions held in a separate tool, or the cause tags only your team knows. Many firms will end up using both: the tool for the figures and charts, this template for the story. We compare the tools in what each reporting tool's AI covers.

Measure it

For two closes, note how many minutes your CFO spends editing each draft and roughly how much of the draft text survives. If edit time is falling and most of the text survives, roll the template to more clients. If not, fix the memory file and the cause tags before blaming the model.

For the wider picture of how firms are using AI, start with how fractional CFOs use AI.

---

Want a second pair of eyes on your monthly cycle? Lunan offers a free conversation: we ask how your close-to-board-pack month runs, which tools you use and how you bill, then send a short written note on where drafting would save the most time and where your current tools already cover it. Start a free conversation.