MM/FINIMO Send me a message
MM/FINIMO 0 of 10 answered

A self-assessment for finance teams

The FP&A AI Reality Check

Ten questions about how your team closes, reviews and signs off numbers. You get a score, three AI use cases you can start with safely, and what to fix first.

Answer for the team as it works today, not as the process document says. One question at a time, a few minutes in all. Your answers stay in this browser. If you ask for your next-step tool by email, we only receive your email and your maturity level (1 to 3).

Start the check

By Martim Pinto, MMFinimo. Ten years in FP&A, financial systems and controlling.

Your result

What your team can start with

What controlled output looks like

One month of variance commentary, with the controls showing

This is the control layer a draft has to pass through before it reaches anyone. It runs on Helix Consumer Goods, a fictitious company, for September 2026.

commentary_2026-09.mdSTATUS: DRAFT — NOT APPROVED
Source: helix_pnl_24m.csv (sha256 d5c41192e998) · thresholds |5.0%| or |100,000 EUR| · generated 2026-11-13T00:40:30
1

CC100 Digital Commerce — actual 7,733,233 EUR, 258,767 EUR below forecast (-3.2%). Driver: [to be confirmed by cost-centre owner]. Reviewer decides whether this is timing, volume, price or one-off.

lineage: 2026-09-CC100 → row 102 of helix_pnl_24m.csv; actual − forecast; variance -258,767 EUR (-3.2%) breaches threshold (|5.0%| or |100,000 EUR|).

2

CC200 Wholesale — flagged as DATA ISSUE, not a business variance. Actual shows 0 against a forecast of 2,808,000 EUR. Action: data owner to confirm the ERP load before any commentary is written.

lineage: 2026-09-CC200 → row 103 of helix_pnl_24m.csv; actual − forecast; actual = 0 with forecast > 0: missing load or mapping. Do not comment on the business.

3

CC300 Marketing — within threshold (-2.7%). No commentary required.

lineage: 2026-09-CC300 → row 104 of helix_pnl_24m.csv; actual − forecast; variance -32,693 EUR (-2.7%) within threshold.

Two more lines within threshold (CC400 Logistics, CC500 G&A) are left out here.
  1. 1
    Material on amount, not on percentage. At -3.2% the line is under the 5% threshold, but 258,767 EUR is over the 100,000 EUR one. The driver is left blank for the cost-centre owner.
  2. 2
    A zero is treated as a data problem. Wholesale shows no actuals against a 2.8m EUR forecast. The line is flagged and gets no narrative until the ERP load is confirmed.
  3. 3
    Every line names its source row. A reviewer can open the exact row of the exact file version, identified by its hash.
  4. 4
    Nothing is approved until a person approves it. The control log below has an empty approver column and every status reads DRAFT.

The control log behind it

output_idcost_centerclassificationsource_rowsource_hashapproverstatus
2026-09-CC100CC100 Digital CommerceMATERIAL102d5c41192e998(empty)DRAFT
2026-09-CC200CC200 WholesaleDATA_ISSUE103d5c41192e998(empty)DRAFT
2026-09-CC300CC300 MarketingIMMATERIAL104d5c41192e998(empty)DRAFT
2026-09-CC400CC400 LogisticsIMMATERIAL105d5c41192e998(empty)DRAFT
2026-09-CC500CC500 G&AIMMATERIAL106d5c41192e998(empty)DRAFT

Synthetic figures. Helix Consumer Goods is a fictitious company. These lines came from a deterministic script, run on 13 November 2026. No AI model wrote them: an AI step may only reword a line between draft and approval, and never after.

Free series

AI in FP&A, the controlled way

Short, practical tools that show finance teams where AI really helps, without losing lineage, approvals or the controller's trust. Follow me on LinkedIn for the next one.