Early bird Book a demo

Guides

AI variance commentary for FP&A teams

AI variance commentary explains why actuals moved against budget, forecast and last year, in words a CFO can sign. Here is what good commentary says, how AI drafts it from your close, a worked example and five checks before it reaches the board.

By Wouter Born, founder of finstory · Updated 12 October 2026

The short answer

Variance commentary is the sentence next to the number: what moved, by how much, against what, and why. AI variance commentary hands the first draft of that sentence to AI. It reads the closed numbers, picks the variances worth explaining, splits each one into its drivers and drafts the words. When the ledger can’t show the reason, it asks the person who knows.

FP&A stays the author. The team checks the draft, adds the business context and signs it off. What changes is where the hours go: less time assembling, more time on the questions the board will ask.

Variance commentary is one layer of AI financial storytelling. The story decides which variances matter and in what order; the commentary explains each one.

What good variance commentary says

Most commentary fails the same way: it restates the table. “Revenue was below budget due to lower sales” tells the reader nothing the number didn’t already say.

Good commentary answers four questions in two or three sentences:

  1. What moved, against what? Name the line, the size and the comparison: budget, forecast or last year.
  2. What drove it? Split the movement into drivers: volume, price, mix, timing, one-offs or a specific cost line.
  3. Why? The business reason behind the driver. This is the part the ledger rarely holds.
  4. So what? Is it timing or trend, and does it change the outlook?

If a sentence doesn’t answer one of these, the board doesn’t need it.

How AI drafts it from your close

  1. It reads governed reports, not raw exports. The draft starts from your P&L against budget, by entity, location or department, on data that loaded and reconciled.
  2. It picks what matters. Variances are ranked by size and by how much they change the picture. The noise stays out.
  3. It splits each variance into drivers. It opens the line one level at a time, by store, channel, account or cost group, until the movement has a shape.
  4. It quotes, never recalculates. Every figure in the draft comes straight from a report. The AI writes the words; the numbers stay the numbers in your reports.
  5. It asks when the data runs out. When the drivers show what happened but not why, it asks the owner a short question and keeps the answer with the figure it explains.
  6. It writes in your format. Headline first, then the drivers, then the reason and what it means for the rest of the year.

A worked example

Born2Cycle is our demo company, a 14-store bicycle retailer on NetSuite. On its May close, EBITDA landed at $1.32M, $354K ahead of budget. One store didn’t follow the group.

The usual line, written at midnight:

Bend was behind budget due to lower sales and higher costs.

The AI’s first pass, from the reports:

Bend Flagship is −$62K against budget on operating income: revenue −$37K, cost of goods +$13K and inbound freight +$10K. Bend ships the web orders, so this may be an online story. What drove it?

The CFO’s answer, in one line:

June’s road-bike range landed early, so its freight hit May. Timing, not trend.

The commentary the board reads:

Bend is $62K behind budget: timing, not trend. Revenue came in −$37K against budget, with cost of goods +$13K and inbound freight +$10K. The freight belongs to June’s road-bike range, which landed early, so it doesn’t change the outlook for the store.

The AI found the variance and its drivers in seconds. Only the CFO knew the reason. Good commentary needs both, and the answer stays pinned to the figure it explains, so next month’s draft starts from it. You can watch this exact exchange in the live story on our home page, and read a complete board story from the September close.

Five checks before AI commentary reaches the board

  1. Every number traces back. Click a figure in the commentary and land on the report and ledger lines behind it.
  2. Direction is in words. “Ahead of budget” and “behind last year”, never a bare plus or minus on a cost line that the reader has to decode.
  3. No invented reasons. If the draft gives a cause the data can’t show, and nobody told it, delete it.
  4. The threshold is agreed. Finance decides which variances get a sentence, and the AI applies that rule the same way every month.
  5. Context is attributed. An answer from a budget holder is marked as theirs, so the board knows who said it.

Where finstory fits

finstory is board reporting AI built for this job. It connects to NetSuite and QuickBooks Online, keeps actuals, budgets and forecasts in one governed model, and drafts the commentary and the board story with you in Claude, page by page. Every figure is quoted from your reports, and when the ledger can’t explain a move, finstory asks.

Questions CFOs ask

What is variance commentary?

The written explanation next to a variance in a management or board report. It says what moved against budget, forecast or last year, by how much, and why, so the reader gets a few findings instead of a table of differences.

Can AI explain why a variance happened?

Partly. AI can split a variance into its drivers, such as volume, price, mix, timing and cost lines, straight from your reports. The business reason, like a deal that slipped or a shipment that landed early, is often not in the ledger. A good tool asks the person who knows instead of guessing.

Which variances should get commentary?

The ones that change the board's view of the year. Agree a threshold with your CFO, for example a percentage and an amount against budget, and comment on everything above it. Group or leave out the rest.

Wouter Born founded finstory after more than 25 years in finance software.