AI in Financial Modelling
Executive Summary
Key Takeaways
- ✓ AI in financial modelling means applying machine learning and generative AI within the modelling process itself, driver identification, formula assistance, scenario generation, and narrative drafting, distinct from AI applied elsewhere in the finance function.
- ✓ The calculated number in a financial model should remain the output of the model's own structured formula logic, not a value generated directly by a language model, regardless of how the model was built or documented.
- ✓ Generative AI is well suited to accelerating the mechanical construction of a model, boilerplate formula patterns, formatting, first-draft schedule structure, where a modeller then verifies the underlying logic.
- ✓ Machine learning is well suited to identifying candidate drivers and relationships in historical data that inform a model's assumptions, but the resulting assumption set still requires the same commercial judgement review any assumption set requires.
- ✓ The line between AI-assisted model construction and AI-generated model output is the single most important distinction in this guide, and is the same reliability distinction addressed for audit on the AI Financial Model Audit page.
Objective¶
This guide defines AI in financial modelling as a distinct practice from artificial intelligence in finance broadly, within AI Financial Modelling & Artificial Intelligence in Finance.
The Central Distinction¶
AI in financial modelling covers three genuinely different applications within the modelling process: driver identification (machine learning surfacing candidate relationships in historical data), construction assistance (generative AI drafting formula patterns, structure, and formatting), and narrative drafting (generative AI summarising a model's output into commentary). None of these should be confused with a fourth thing this guide explicitly excludes: AI directly calculating the model's output. The calculated number in a financial model should remain the product of the model's own structured, traceable formula logic, verifiable independently of how it was constructed.
Where AI-Assisted Construction Adds Genuine Value¶
Driver identification. Machine learning can surface candidate relationships in historical data, which cost lines correlate with which volume driver, that a modeller would otherwise need to identify manually. The output is a candidate list requiring the same commercial judgement any driver selection requires; it is an input to modelling, not a substitute for it.
Formula and structure assistance. Generative AI can accelerate mechanical, well-defined construction tasks, drafting a boilerplate depreciation schedule, formatting a summary tab, proposing a first-draft structure for a new schedule. A modeller then reviews and verifies the underlying formula logic before relying on it, exactly as they would review formulas written by a junior analyst.
Scenario generation. Generative AI can help generate scenario variations, a set of plausible sensitivity ranges around a central case, that a modeller then evaluates and refines rather than accepts uncritically.
Narrative drafting. Generative AI can draft commentary around a model's calculated output for a board pack or investment memo, provided the underlying numbers being narrated are themselves the verified output of the structured model, not numbers the AI itself produced.
Where the Line Must Hold¶
A financial model's calculated output, the number a lender, investor, or board relies on, should be the product of the model's own auditable formula logic. Using a language model to estimate or generate that number directly, rather than to assist in building the structure that calculates it, forfeits the traceability and repeatability a material financial decision typically requires. This is the same distinction drawn between deterministic and generative approaches on the AI Financial Model Audit page, applied here to construction rather than audit.
Common Construction Pitfalls¶
Allowing AI-drafted formulas into a model unverified. Accepting a generative AI's first-draft formula without independently verifying its logic reintroduces exactly the structural risk a careful build process is designed to prevent.
Treating a machine learning driver suggestion as a finished assumption. A statistically identified relationship in historical data is a candidate for further commercial judgement, not a ready-to-use assumption, since historical correlation does not establish forward-looking causation or continued relevance.
Blurring construction assistance with output generation. Using a language model to estimate a number the model itself should calculate collapses the distinction this guide is built around, and removes the traceability a structured model provides.
Recommended Practices¶
- Use AI to accelerate mechanical construction and drafting tasks, not to generate the model's calculated output directly.
- Review and verify any AI-assisted formula or structure exactly as you would review work from a human analyst, before relying on it.
- Treat machine learning driver suggestions as candidates for judgement, not finished assumptions.
- Keep the model's calculation logic traceable and independently auditable regardless of how it was constructed.
Continue Reading¶
Related Pillars¶
Related Technical Guides¶
How OXXON tests thisRun a free structural check with FMAE
Frequently Asked Questions
What is AI in financial modelling?
The application of machine learning and generative AI techniques within the financial modelling process itself, driver identification from historical data, formula and structure construction assistance, scenario generation, and narrative drafting around a model's output.
Does AI in financial modelling mean the AI calculates the model's numbers?
No, and this is the central distinction this guide makes. The calculated number should remain the output of the model's own structured, auditable formula logic. AI supports the construction and narration of the model; it should not be the mechanism producing the calculated result itself.
What is generative AI best used for in model construction?
Accelerating mechanical, well-defined construction tasks, boilerplate formula patterns, schedule formatting, first-draft structure, where a modeller then reviews and verifies the underlying logic before relying on it.
What is machine learning best used for in model construction?
Identifying candidate drivers and relationships in historical data that inform a model's assumptions, though the resulting assumption set still requires the same commercial judgement review any assumption set requires, whether AI-assisted or not.
How does this relate to AI financial model audit?
The same reliability distinction applies. Just as an AI financial model audit distinguishes deterministic rule-based methodology from generative commentary, AI-assisted model construction should distinguish AI support for building and narrating a model from AI generation of the model's calculated output itself.
Can a model built with AI assistance still be audited normally?
Yes. A model's formulas and structure can and should be audited on their own merits regardless of how they were constructed, since a structural audit tests the calculation logic itself, not the process used to write it.
Related Articles
AI Financial Modelling & Artificial Intelligence in Finance
AI financial modelling is the application of machine learning and generative AI techniques within the financial modelling process itself, driver identification, construction assistance, scenario generation, and narrative drafting, while artificial intelligence in finance is the broader application of those same technique categories across the finance function generally. This page is the hub for the Knowledge Centre's AI financial modelling content: the foundational distinction between machine learning, natural language processing, and generative AI; how AI accelerates modelling construction without replacing the auditable calculation layer beneath it; a staged framework for adopting AI reliably; enterprise applications across FP&A, forecasting, valuation, and investment analysis; governance and risk practice; and the institutional best practice synthesis this domain builds toward.
Artificial Intelligence in Finance
Artificial intelligence in finance spans a wide range of techniques, machine learning, natural language processing, and generative AI, applied across a wide range of finance functions, financial modelling, FP&A, risk management, treasury, and audit. This guide sets out the main categories of AI technique in practical finance use today, the finance functions each is best suited to, and the foundational distinction between AI applied to raw data (prediction, classification) and AI applied to language and reasoning (generation, summarisation), as the entry point for the more specific guides in this domain.
Generative AI in Financial Modelling
Generative AI, large language models applied to drafting and language tasks, has a specific and bounded role in financial modelling: accelerating structure, formatting, and narrative drafting, not producing verified numerical output. This guide sets out that role in detail, the specific failure modes generative AI introduces into a modelling workflow, hallucinated figures, plausible-but-incorrect formula logic, and unverifiable citations, and the concrete review practices that contain each failure mode.
AI-Assisted Financial Analysis
AI-assisted financial analysis works best when structured as a defined workflow rather than an ad hoc use of a chat tool: decomposing an analysis into discrete tasks, assigning each task to the approach best suited to it (AI-assisted or human-led), and placing a human verification checkpoint at each point where AI output feeds into a conclusion. This guide sets out that workflow structure and the checkpoint discipline that keeps it reliable.
What Is an AI Financial Model Audit?
An AI financial model audit is a financial model audit performed by an automated engine rather than a human reviewer working manually. Not every application of AI to financial models works the same way, and the distinction between approaches is not a marketing detail — it is the difference between an audit whose findings are repeatable and explainable, and one whose findings may not be. This page defines what an AI financial model audit is, the specific distinction between deterministic, rule-based audit and general-purpose generative AI review, and why that distinction determines whether an automated tool's output is suitable to support a material financial decision.