Sign in
Set up for FP&A teams

AI for insights and analytics

AI for FP&A

AI for FP&A is the analyst's two days of variance work answered in a sentence: what is driving the increase, how much of it is timing, which cost centres to talk to. The model chooses the measure and writes the explanation; a tool builds the bridge from the ledger, the budget and the headcount file, and the lines sum to the variance.

Ask it in your own words

These are the questions FP&A teams ask. Each one maps to a measure our tools compute from your files. The AI model chooses the measure and explains the result; the arithmetic is done by our code, and every total is checked.

The questionThe measure behind itWhat comes back
What is driving my cost increase?Cost bridgeThe change split into headcount, rates and pay, volume, timing, one-offs and other, summing exactly.
Which cost centres are over budget?Cost centres over thresholdCost centres over the threshold with the largest line and the owner.
How much of the variance is timing?Timing in the varianceItems that landed in a different month and will reverse, kept separate from the real variance.
Where is headcount above plan?Headcount against planActual against planned FTE by cost centre and type, and whether it is numbers or rates.
Will the year land on current spend?Run rateTrailing run rate against the full-year budget, by function.
Where is the programme overrun coming from?Programme positionBudget, committed, actual and EAC by workstream, with contingency drawn and the change requests behind the overrun.

A conversation, with the figures

Every figure below was computed by a tool from the rows and checked before it was shown. The lines sum; the percentages match; each line opens to the rows that make it.

Where is the programme overrun coming from?

The estimate at completion is £4.3m over the 100 million budget, split across 3 workstreams that sum to the overrun. Data migration accounts for £2.3m, 53% of it.

Data migration£2.3m53% of the total
Core platform£1.4m33% of the total
Integration£600k14% of the total
The 3 lines sum to£4.3m0 unexplained

What is behind the data migration overrun?

3 items account for £2.1m of the £2.3m overrun on Data migration, 91% of it. Two are approved change requests; the third is contractor rates above the rate card.

Change request CR-014, legacy data cleansing£1.1mapproved; drew contingency
Change request CR-019, second migration wave£700kapproved
Contractor rates above rate card£300kthree vendors; 14% above
These 3 are£2.1m91% of Data migration

What does the programme finance lead put in the steering pack?

The bridge from budget to EAC with the change requests as their own lines, contingency drawn and remaining, and the rate card exceptions with the vendors named. The steering committee decides on the second wave; the pack makes the figures explicit.

Programme finance leadPut the budget-to-EAC bridge in the steering packBefore steering
Workstream lead, Data migrationChallenge the three vendors on rate cardThis fortnight
PMOReport contingency drawn and remaining weeklyWeekly
What is driving my cost increase?Which cost centres are over budget?How much of the variance is timing?Where is headcount above plan?

The measures behind the answers

Each measure has one formula, one source and one meaning. They are computed per cost centre and account and in total, and every one carries an identity that must hold before it is shown.

MeasureFormulaFromWhat it tells you
Variance to budget and forecastActual − budget; actual − latest forecast, by cost centre and accountLedger actuals; budget; forecastWhere the month landed against both plans
Cost bridgeChange = headcount + rates and pay + volume-driven + timing + one-offs + other, summing exactlyLedger; headcount; timing and one-off filesWhat is driving the cost increase
Headcount against planActual FTE − planned FTE, by cost centre and type, with average costHeadcount fileWhere people cost is above plan, and whether it is numbers or rates
Timing in the varianceSum of items landing in a different month from budget that reverseTiming items fileHow much of the variance is not real
One-off itemsSum of items finance has flagged as non-recurringOne-off items fileWhat will not repeat next month
Run rateTrailing three-month actual × 4, against the full-year budgetLedger actuals; budgetWhether the year will land, on current spend
Forecast accuracy and biasActual − forecast by month, signed and absolute, by cost centreForecast versions; ledgerWhich cost centres forecast high or low, consistently
Cost centres over thresholdCost centres more than x percent over budget, with the largest lineLedger; budgetWho to talk to this month
Vendor varianceActual − budget by vendor, with new and lost vendorsLedger with vendorWhich suppliers drove the change
Programme positionBudget, committed, actual and estimate at completion by workstreamProgramme budget; purchase orders; actuals; forecastWhether a large programme will land, and where the overrun sits

What the AI model does, and what our tools do

The AI model chooses the measure

From your question and the measures declared for FP&A teams, the model picks the one that answers it, and the period and comparison the question implies.

Our tools do the arithmetic

Deterministic code reads the rows, computes the measure, and checks the identities below. The same question on the same data gives the same answer, every time.

The AI model explains, and cites

The AI model writes the sentence around the result, naming the cost centre or account behind it. It states no figure that is not in the result, and every figure links to its rows.

The identities that must hold

BridgeEnd − start − Σ lines = 0; a residual is shown as its own line, never hidden
VarianceActual − budget = real variance + timing + one-offs
HeadcountActual FTE = planned FTE + variance; cost = FTE × average cost
ProgrammeEAC = actual + committed + estimate to complete; contingency drawn + remaining = contingency at start

What it reads

The exports FP&A teams already produce. Column names are mapped once and the mapping is reused. A file is the way in; scheduled delivery and connections to your systems come with the plan, and every source is listed here.

  • Ledger actuals
  • Budget
  • Forecast
  • Headcount
  • Timing and one-off files
  • Timing items file
  • One-off items file
  • Ledger with vendor
  • Programme budget
  • Purchase orders
  • Actuals

Who owns each answer

An answer is a list with an owner and a cadence, or it is a chart nobody works.

MeasuresOwnerCadence
Variance and bridgeHead of FP&A; finance business partnersMonthly
Headcount against planBusiness partners with HRMonthly
Programme positionProgramme finance leadMonthly, weekly near gates

Questions FP&A teams ask about AI analytics

What does AI do for FP&A that a spreadsheet cannot?

It removes the two days between the question and the bridge. The analyst still owns the definitions and the judgement; the tool builds the bridge from the files every month, the same way, and the model explains it in plain words with the cost centres named. The figures are the ones the analyst would have computed, because they are computed the same way.

Can the model be trusted with financial figures?

The model never produces one. It chooses which measure answers the question and writes the sentence around a result a tool has already computed and checked. Every bridge asserts that its lines sum to the change before it is shown, and every figure links to the ledger rows it came from.

What files does it need?

Ledger actuals by cost centre and account, the budget and latest forecast phased by month, the headcount file, and the timing and one-off items finance already tracks. For programme cost control, the budget by workstream, purchase orders, actuals and the estimate at completion.

How does it handle a large programme?

Budget, committed, actual and estimate at completion are computed by workstream and work package, the bridge from budget to EAC shows change requests, rate variances and contingency as separate lines, and every figure reconciles to the purchase order and actuals files. A 100 million programme is a set of tables, not a special case.

See it on your data

Bring a few thousand rows. The data map opens next, every column mapped once, and the first question is answered in minutes. Free, in your browser, no account.