Latest episode
AI in FP&A for forecasting, commentary and scenarios on a model your team owns.
We check whether your drivers, your actuals and your hierarchies can carry a model before we put AI on top of them. The price is fixed after that check.
AI in FP&A means a model doing the repeating work of the planning cycle.
AI in FP&A covers the forecast, the first draft of the variance commentary and the scenario branches. A model produces them and a person checks them against the ledger. The tasks that suit it repeat every cycle and have an answer somebody can verify.
The model only works on drivers the business owns. A driver nobody outside finance has agreed to is an assumption. A forecast built on assumptions is one nobody will defend in a board meeting.
Constancia is an EPM consultancy, and we read the planning model before we put AI on it. If the drivers, the actuals or the hierarchies cannot carry a model, that is the first piece of work.
5 tasks in the planning cycle suit a model, and a person signs the fifth.
Awaiting image
- 01
The forecast
The model rolls the forecast forward from the drivers, headcount, pipeline and volume, each month without a rebuild.
- 02
The variance commentary
The model writes the first draft of why actuals differ from plan, by entity and by line, and an analyst edits it.
- 03
The scenarios
A scenario branches off the base case and the downstream effects update themselves, so 3 versions cost the same as 1.
- 04
The reconciliation
The model checks the actuals loaded from each ledger against the plan's hierarchies and lists what does not map.
- 05
The sign-off
A named person in finance signs anything that reaches a number the board reads, and the model records what it used.
The same AI gives 2 different results on 2 different planning models.
A model built on owned drivers
1 forecast, and the sales director stands behind the pipeline number in it.
The drivers have owners outside finance, the actuals load from the ledgers every month, and the hierarchies match the consolidation. The model learns from data that means the same thing every time.
- Headcount from HR, pipeline from the CRM, volume from operations
- 1 set of entities and accounts, the same as the close uses
- A forecast the CFO can trace to its drivers
A model built on spreadsheets
3 forecasts, and nobody can say which one the board should see.
3 workbooks hold 3 versions of the customer list, the actuals arrive by email and the hierarchies were last agreed in 2 different years. The model learns 3 definitions of the same number.
- A driver that is a cell somebody types into on the last Friday
- Actuals that do not tie back to the ledger they came from
- A forecast that is confident, fluent and wrong
AI in FP&A pays back for a planning team that meets 4 conditions.
- 01
More than 1 entity
The plan consolidates across entities, so the same driver has to mean the same thing in each.
- 02
A date in the diary
A budget round, a funding round or a board date that the forecast has to be ready for.
- 03
A planning platform, or a decision to buy one
Abacum (FP&A software) or OneStream (EPM software), or a spreadsheet estate the team has agreed to leave.
- 04
Somebody to own the drivers
A person outside finance who will stand behind headcount, pipeline or volume when the forecast is challenged.
A single-entity business planning in 1 workbook does not need this yet. The spreadsheet is the right tool until the structure outgrows it.
We build AI in FP&A on 2 platforms, and the shape of the group decides which.
OneStream (EPM software)
SensibleAI on the consolidated model
- For a group that plans and closes on the same platform
- Forecasting and agents read the model the consolidation already uses
- 1 set of entities, accounts and periods for the plan and the actuals
Abacum (FP&A software)
Abacum Intelligence on the planning model
- For a finance team whose consolidation is settled and whose plan is not
- Drivers, scenarios and commentary in the same workspace the team plans in
- Live in weeks, and the model is handed back to finance
Constancia is an official partner of both and our consultants are certified on both.
Questions finance teams ask about AI in FP&A.
Is AI taking over FP&A?
No. A model takes the repeating work: rolling the forecast, drafting the commentary, checking the actuals. The judgement about what the numbers mean, and the signature on the board pack, stay with the finance team.
How is AI being used in FP&A?
In 4 places: forecasting from drivers, the first draft of variance commentary, scenario branches that update themselves, and reconciliation of loaded actuals against the plan. Each one works only where the underlying data model is owned.
What is the best AI for FP&A?
The one that reads your planning model directly. On OneStream (EPM software) that is SensibleAI, and on Abacum (FP&A software) it is Abacum Intelligence. A general chatbot fed a spreadsheet export has no way to check its own output against the ledger.
What is the 30% rule in AI?
It is a rule of thumb that circulates online: let a model do about 30% of a task and keep a person on the rest. It is not a standard. Our test is plainer. A model does the tasks that repeat and have a checkable answer, and a person signs anything the board reads.
Is FP&A getting replaced by AI?
The spreadsheet work is. The analyst who used to rebuild the forecast every month now spends that time on why the variance happened. That was always the job.
See more on our
Subscribe on YouTube
Every Sunday, 6am
The Sunday Close.
Our weekly newsletter on AI in finance.
Yours to enjoy with your Sunday morning coffee.
- AI in finance observations from Alex (CEO) and Brad (COO)
- Newest video from our YouTube channel
- AI in finance trends we have spotted each week
Tell us the date you are working to.
The people who would do the work read every message and reply inside 2 working days.