AI in Forecasting: A Finance Leader’s View

AI in Forecasting: A Finance Leader’s View

AI is increasingly used in financial forecasting — to model scenarios, spot patterns in historical data, and produce projections faster than a manual process could. Some organisations report meaningful gains in forecast accuracy, and the shift from periodic, backward-looking budgeting to continuous, forward-looking planning is among the largest productivity opportunities available to finance today. For the finance leader, though, the important question is not how to build these models. It is how to interpret, challenge and govern the forecasts they produce. This guide takes the oversight view.

It covers what AI-assisted forecasting is genuinely good at and where it fails; why the finance leader’s job is to challenge rather than accept; the model-risk and validation obligations that apply; the confident-error problem that makes AI forecasts deceptive; and how to govern forecasting across the function. The mechanics of building a model sit with the FP&A team; the judgement to trust or challenge its output sits with the leader.

What AI-assisted forecasting is good at — and what it is not

AI can process more data, faster, and surface relationships a manual model might miss. Machine-learning approaches can ingest hundreds of variables — revenue drivers, macroeconomic signals, operational metrics — and produce rolling forecasts that update continuously rather than quarterly. Some organisations using AI in financial planning report forecast-accuracy improvements of up to 40%. That figure should be read as a vendor-and-case claim rather than a guarantee, but the direction is real: done well, AI-assisted forecasting is faster and can be more responsive than the annual-cycle model it replaces.

But a forecast is a claim about the future, and AI carries specific limitations the finance leader must keep in view:

• It extrapolates from the past. AI-assisted forecasts lean on historical patterns and can perform well in backtesting yet miss a change the data has not seen. A model that fits history well is not the same as one that predicts the future well.
• It can be confidently wrong. The defining risk of AI forecasting is not that it errs but that it errs with conviction — presenting a flawed projection with apparent precision. In one widely-discussed live test in front of finance professionals, an AI tool built a model with an unbalanced balance sheet, could not locate the error, and suggested the imbalance was acceptable to present anyway.
• It does not know the business context. A pipeline change, a lost customer, a market shift the model has not been told about — these are the finance leader’s to bring.

The finance leader’s job is to challenge, not to accept

The value a CFO adds to an AI-assisted forecast is scrutiny. Where does this projection assume the past continues? What would have to be true for it to hold? What does it not account for? A finance leader who takes an AI forecast at face value has added nothing; one who interrogates it turns a model output into a considered view the board can rely on. This is the same discipline good finance leaders have always applied to forecasts — AI raises the premium on it, precisely because the output looks more authoritative than it may be.

A striking implication has emerged from those testing these tools seriously: AI-assisted forecasting demands stronger modelling skills to audit the output, not weaker ones. When errors hide inside formulas the finance team did not write, the ability to spot them becomes more valuable, not less. The finance leader should resist any assumption that AI lets the function get away with thinner modelling capability — the opposite is true.

Model risk and validation

Where AI-assisted forecasting informs material decisions — and especially in regulated firms — model risk becomes a formal obligation rather than a good idea. AI models can drift, produce biased outputs, or fail silently, and validation is a supervisory expectation, not an optional extra. The PRA’s model-risk-management expectations under SS1/23 and, for firms with EU exposure, the EU AI Act impose specific obligations on high-risk AI systems used in financial contexts. A finance leader relying on AI-assisted forecasting for anything material should ensure the model is validated, its assumptions documented, and its limitations understood — and should treat concentration risk seriously where forecasting, fraud detection and reporting all depend on the same underlying provider.

Governing forecasting across the function

Beyond any single forecast, the finance leader owns how AI-assisted forecasting is used across the function: which tools are approved, how their outputs are reviewed before they inform a decision, and how assumptions are made visible rather than buried in a model. A forecast whose logic cannot be explained is a governance problem, however sophisticated the model. Given that a large majority of business spreadsheets are estimated to contain material errors even before AI is involved, a disciplined review standard for AI-assisted forecasts is not bureaucracy — it is the minimum a finance leader should insist on.

Communicating AI-assisted forecasts to the board

When a forecast informed by AI goes to the board or to investors, the finance leader owns how its reliability is represented. Presenting an AI projection as more certain than it is exposes the business; presenting it with honest ranges and clear assumptions builds trust. Static budgeting is increasingly described as a fiduciary risk in a volatile environment — but so is over-trusting a forecast because it came from a sophisticated tool. The CFO is accountable for that framing regardless of how the number was produced.

The rolling-forecast shift — and what it asks of the leader

The most consequential change AI brings to forecasting is not accuracy but cadence. Traditional FP&A locked finance into annual or quarterly cycles; AI-enabled forecasting allows rolling forecasts that update monthly, weekly, or in some functions continuously. Advanced teams now run hundreds of scenarios in minutes rather than building one base case and a couple of sensitivities by hand. For a finance leader this is a genuine capability gain — the function can respond to a shock, a policy change or a pipeline shift almost as it happens.

But faster forecasts multiply the finance leader’s review burden rather than removing it. A forecast that updates weekly is a forecast whose assumptions need checking weekly, and a function running hundreds of scenarios needs the judgement to know which few actually matter. The leader’s task shifts from producing the forecast to governing a forecasting system — setting which scenarios are worth attention, which assumptions are load-bearing, and which outputs are reliable enough to act on. Speed without that judgement produces more forecasts, not better decisions.

For UK finance leaders specifically

The rolling-forecast capability is particularly valuable against the UK’s policy cadence. Events such as the annual Budget can reshape a company’s financial outlook overnight — a change to allowances, National Insurance, or the National Living Wage feeds directly into cost and cash models. An AI-enabled forecasting function can remodel quickly against such a change; the finance leader’s role is to ensure the remodelling reflects the actual policy detail correctly, which is a matter of judgement and technical knowledge that the tool does not supply.

What this means for finance leadership

AI does not remove the need for forecasting judgement — it concentrates it. The finance leader who can use AI to forecast faster while holding the scrutiny, the context, the model-risk discipline and the honest communication is applying exactly the oversight AI adoption calls for. It is one of the capabilities FD Capital assesses when placing senior finance leaders into businesses relying on AI-assisted planning.

Call 020 3287 9501 or email recruitment@fdcapital.co.uk to discuss a finance leadership appointment where governing AI-assisted forecasting is part of the brief.

FD Capital — CFO and Finance Director Recruitment

Fellow of the ICAEW | Placing finance leaders who bring judgement, not just tools, to planning and forecasting, since 2018. We recruit permanent, interim and fractional finance leaders across the UK. 4,600+ network. 160+ placements. Shortlists in 3–7 working days.

This guide is general information for finance leaders and does not constitute legal, regulatory or professional advice. Businesses should take their own advice on their specific circumstances. Regulatory positions described are current as at mid-2026 and are developing; readers should check the FCA’s latest publications.