How FP&A Teams Are Using AI for Forecasting

Forecasting is at the heart of what FP&A does, and it is one of the areas being most actively reshaped by AI. The traditional forecasting process — gathering data, building and maintaining models, generating forecasts, analysing the results — involves a great deal of work that AI can accelerate, and AI also offers genuinely new forecasting capabilities that go beyond what traditional methods provide. For FP&A teams, this is a significant opportunity: to forecast faster, to explore more, and potentially to forecast better, provided the tools are used with judgement and the forecasting retains the human understanding and rigour that good forecasting requires. The FP&A team that adopts AI for forecasting well becomes more capable; the one that adopts it carelessly, or that abdicates the judgement to the tools, does not.

This guide is written for FP&A professionals who want to understand how AI is being used in forecasting and how to use it well. It covers where AI helps in the forecasting process, the new forecasting capabilities AI offers, the risks and the limits of AI in forecasting, how to adopt it sensibly, and what it means for the FP&A role. It is practical rather than speculative, focused on how AI is actually used in forecasting rather than on grand claims. It connects to the broader use of AI across finance covered in our hub on AI in finance. The aim is a clear, practical understanding of how FP&A teams capture the genuine benefits of AI in forecasting while maintaining the rigour and judgement that the discipline requires.

Where AI Helps in the Forecasting Process

AI helps across the forecasting process in ways that accelerate the work and free the FP&A team for higher-value analysis. In the data work that underpins forecasting — gathering, cleaning and preparing the data — AI and automation can do much of the routine handling, reducing the time the FP&A team spends on data preparation rather than analysis. In the building and maintenance of forecasts, AI can assist with the construction and updating of models, accelerating the mechanical work. In the generation of forecasts, AI can produce forecasts from the data, including through statistical and machine-learning methods that can identify patterns and relationships in the data that inform the forecast.

AI also helps in the analysis of forecasts and the exploration of scenarios. It can assist in analysing the forecast results, identifying the drivers and the patterns, and it can accelerate the generation and analysis of scenarios, making it faster to explore the range of possibilities. And it can help with the commentary and communication of the forecast, drafting the narrative that explains the forecast for the FP&A team to review and refine. Across these applications, the pattern is the consistent one of AI in finance: the tool accelerates the routine and the mechanical, freeing the FP&A team to focus on the judgement, the assumptions, the analysis and the insight that add the value. The FP&A team that applies AI across the forecasting process this way forecasts faster and frees capacity for the higher-value work, which is the immediate benefit AI brings to forecasting.

The New Forecasting Capabilities AI Offers

Beyond accelerating the traditional process, AI offers genuinely new forecasting capabilities that go beyond conventional methods, and FP&A teams are increasingly exploring them. Machine-learning forecasting methods can identify complex patterns and relationships in data that traditional methods may miss, potentially producing more accurate forecasts where the underlying patterns are complex and the data supports the methods. These methods can incorporate a wider range of data and detect non-obvious relationships, offering forecasting capabilities that go beyond the conventional driver-based and statistical approaches in certain circumstances.

AI also enables more extensive scenario analysis and exploration, because it can generate and analyse a wider range of scenarios faster, allowing the FP&A team to explore the possibility space more thoroughly than manual methods permit. And it can potentially incorporate more data sources and signals into the forecast — external data, leading indicators, a wider range of inputs — than traditional forecasting typically uses. These new capabilities are genuine, but they should be approached with appropriate judgement: they are powerful in the right circumstances but not universally superior, and they bring their own risks and limits, discussed below. The FP&A team that explores the new capabilities AI offers — the machine-learning methods, the more extensive scenario exploration, the wider data — can enhance its forecasting where these genuinely add value, while recognising that they are tools to be applied with judgement rather than a wholesale replacement for forecasting understanding. Understanding both the acceleration of the traditional process and the new capabilities is the full picture of how AI is used in forecasting.

The Risks and Limits of AI in Forecasting

An honest account of AI in forecasting must address its risks and limits, because forecasting is consequential and the FP&A team remains accountable for the forecasts regardless of the tools. The most fundamental risk is the uncritical reliance on AI-generated forecasts — treating a forecast produced by an AI method as authoritative without understanding and scrutinising it. AI forecasting methods, particularly the more complex machine-learning ones, can produce forecasts whose basis is opaque, and a forecast that is not understood cannot be properly judged or relied upon. The FP&A team must understand and scrutinise the AI-generated forecasts, applying judgement to whether they are sensible, rather than accepting them because they came from a sophisticated method.

A related limit is that AI forecasting methods depend on the data and the patterns in it, and they can fail when the future differs from the past in ways the data does not capture — a structural change, an unprecedented event, a shift the historical patterns do not reflect. Forecasting fundamentally involves judgement about the future that pure extrapolation from historical data, however sophisticated, cannot fully provide, and the human understanding of the business and its context remains essential. A further risk is the loss of forecasting understanding if the FP&A team relies on AI without engaging with the forecasting, becoming a passer-on of AI forecasts rather than a team that genuinely understands the forecast. And the data and confidentiality risks that attend all AI use in finance apply. The FP&A team that understands these risks and limits — the need to scrutinise AI forecasts, the dependence on data and the limits when the future differs from the past, the need to retain forecasting understanding, the data risks — uses AI in forecasting soundly; one that relies on it uncritically risks forecasts that fail when it matters. Managing these risks is what allows the benefits to be captured safely.

Adopting AI for Forecasting Sensibly

Adopting AI for forecasting well is a matter of judicious application that captures the benefits while managing the risks. The sensible approach uses AI to accelerate the forecasting process and to augment the forecasting capability, while retaining the human judgement, understanding and rigour at the centre of the forecasting. AI handles the data work, assists with the models, accelerates the scenarios, and offers additional forecasting methods, while the FP&A team owns the assumptions, scrutinises the forecasts, applies the judgement about the future, and retains the understanding of the business that good forecasting requires. This combination — AI accelerating and augmenting, the human judging and understanding — captures the benefits without surrendering the rigour.

The adoption should be incremental and evidence-led, starting with the applications where the benefit is clear and the risk is contained — the data work, the acceleration of the routine, the scenario exploration — and extending to the more advanced forecasting methods with appropriate care, testing them against the FP&A team’s understanding and the alternatives rather than adopting them wholesale. Where AI forecasting methods are used, they should be understood and validated, not treated as black boxes whose output is accepted. And the FP&A team should develop the skills to use the tools well and to judge their output, because the value depends on using them skilfully and applying judgement to them. The FP&A team that adopts AI for forecasting this way — judiciously, incrementally, with the human judgement retained and the methods understood — captures the genuine benefits while managing the risks, which is what sensible adoption requires.

What It Means for the FP&A Role

AI is changing forecasting and the FP&A role, and the change rewards the FP&A professionals who embrace the tools while retaining the judgement and understanding that forecasting requires. The mechanical aspects of forecasting — the data work, the model maintenance, the routine generation — are increasingly accelerated by AI, freeing the FP&A team for the judgement, the analysis, the scenario thinking, and the business partnering that add the most value. The FP&A role shifts further toward the high-value work — the understanding of the business, the judgement about the future, the analysis and the insight, the partnering with the business — as the mechanical forecasting work is automated. This is, on balance, a positive shift, moving the FP&A professional toward the more valuable and more interesting work.

This shift rewards the FP&A professionals who develop the judgement, the business understanding, and the analytical and partnering capabilities that the evolving role demands, alongside the skill to use the AI tools well. The ability to operate the tools is necessary but not sufficient; the value lies increasingly in the judgement to use them well, to scrutinise their output, to apply the human understanding the tools cannot provide, and to turn the forecasting into insight and partnership. The FP&A professional who develops these capabilities, and who embraces the tools that free the time for them, positions themselves well for a role that is becoming more analytical, more judgement-based, and more valuable. AI is not replacing the FP&A professional but augmenting them, and the FP&A professional who navigates this well — using the tools, retaining the judgement, focusing on the high-value work — becomes more capable and more valuable. This is the opportunity AI represents for FP&A, and the FP&A team that embraces it thoughtfully forecasts faster, explores more, and frees itself for the work that genuinely adds value.

Verification and the Human in the Loop

The single most important discipline in using AI for forecasting is keeping the human firmly in control of the judgement, because AI can produce forecasts and analysis that are fluent and plausible but wrong, and the FP&A professional remains accountable for the forecast regardless of the tools used to produce it. AI may assist in building a forecast, generating a first-pass analysis, or accelerating the routine work, but the assumptions, the judgement about what is reasonable, and the responsibility for the output must remain with the FP&A professional. An AI-generated forecast accepted without scrutiny may embed errors or implausible assumptions that flow through to the decisions the forecast informs, which is exactly the risk the human in the loop guards against.

This means using AI as an assistant that accelerates and supports the forecasting while the FP&A professional verifies and owns the result, rather than as an oracle whose output is taken on trust. The efficiency comes from the AI doing the routine work quickly — the data handling, the first-pass analysis, the drafting — not from skipping the judgement and verification that sound forecasting requires. The FP&A professional who maintains this discipline captures the efficiency of AI while preserving the rigour of the forecasting; one who abdicates the judgement to the tool risks producing forecasts that are fast but unreliable. The human in the loop is what allows AI to enhance forecasting rather than undermine it, and maintaining that control is the foundation of using AI for forecasting responsibly, as covered more broadly in our guidance on AI in the finance function.