Forecast Agent

Fifty-one percent of CFOs name forecast accuracy as one of their five top priorities (Gartner CFO Survey 2025). Yet most FP&A teams still work with point forecasts in spreadsheets - a single number that is already outdated on the day it is produced. The problem lies not in the calculation. It lies in the missing separation between what an algorithm can deliver and what requires human judgement.

Forecast accuracy fails on processes, not models

When a forecast misses its mark, the instinctive reaction is to question the model. More variables, more complex regressions, more data points. But KPMG research shows that companies with high data quality are 3.5 times more likely to produce reliable forecasts than companies with better models but worse data.

The bottleneck runs deeper: in most finance departments, the forecast process mixes mechanical work with judgement calls. The same person who cleanses historical data is expected, in the same step, to define growth assumptions. The result is projections that are neither truly data-driven nor strategically grounded - but a compromise of both.

A Forecast Agent solves this by decomposition. Data preparation, trend extrapolation, scenario calculation and sensitivity analysis run rule-based. Defining assumptions, assessing plausibility and strategic interpretation stay with the CFO or Head of FP&A. Not because automation would be impossible there, but because these steps demand business knowledge that no model can replicate.

Scenario analysis replaces the deceptive point forecast

A revenue forecast of EUR 142 million (USD 156 million) conveys a precision that does not exist. Reality is a range - depending on market developments, exchange rates, raw material prices and dozens of other variables. According to an EPM Channel study, only 42 percent of companies use a rolling forecast, even though companies with rolling forecasts achieve 25 percent more accurate projections.

The Forecast Agent calculates three fully modelled scenarios: best case, base case, worst case. Each scenario rests on a defined assumption set - not on optimism or pessimism, but on parameterised variables. Exchange rate assumption plus two percent, raw material costs minus five percent, sales volume constant: that yields a concrete scenario with traceable logic.

For corporate steering, this represents a step change in decision quality. The board no longer debates a single number but discusses ranges and their probabilities. The question shifts from “Is the forecast right?” to “Under which conditions do we land in the worst case - and what do we do then?”

Sensitivity analysis identifies the one variable that matters

Not all assumptions carry equal weight. A forecast may hinge on ten variables, but often a single one dominates the outcome corridor. For an export-oriented manufacturer, the EUR/USD rate can produce more earnings volatility than all internal cost levers combined.

The Forecast Agent’s sensitivity analysis systematically calculates how much the result changes when each variable shifts by a defined percentage. The output is a ranking: which assumption has the greatest leverage? Where does hedging pay off? Where is monitoring sufficient?

This transparency changes board conversations. Instead of twenty slides of detailed figures, a clear statement enters the room: “Our EBIT reacts three times more strongly to the copper price than to personnel costs. That is where our risk sits - and where our action option lies.”

The Decision Layer separates calculation from strategic judgement

The Forecast Agent operates at Decision Layer tiers 1 and 2. Five of the eight decision steps are rule-based: data preparation, trend extrapolation, scenario calculation, sensitivity analysis and budget comparison. Three steps remain with the human: defining assumptions, checking plausibility, strategic assessment.

This architecture is deliberately conservative. For listed companies, forecasts flow into the outlook section of the annual report. The assumptions behind each scenario must be traceable - not only for management but also for statutory auditors and the supervisory board. The Decision Layer documents for every forecast cycle which assumptions were defined by whom, which data served as the basis and how the scenarios were calculated.

The result for FP&A teams: instead of spending weeks on data cleansing and spreadsheet maintenance, they invest their time where they actually create value - in interpreting the numbers and delivering strategic recommendations to the board.