AI Sensitivity Analysis
Executive Summary
Key Takeaways
- ✓ Machine learning can help rank which drivers most influence a model's output across historical data, directing modeller attention to the variables most worth testing rigorously.
- ✓ Generative AI can draft commentary explaining sensitivity results, which should be checked against the actual tested ranges and output figures before being finalised.
- ✓ The specific sensitivity ranges tested against each driver should remain a modeller's defined, documented, and auditable input, not a range an AI model generates on its own.
- ✓ A driver ranking based on historical influence is a useful prioritisation tool, not a determination that lower-ranked drivers can be safely ignored, since a driver's historical influence may not capture a genuinely novel forward-looking risk.
- ✓ The distinction between AI helping identify what to test and AI determining the ranges being tested is the central discipline this guide sets out.
Objective¶
This guide sets out AI-assisted sensitivity analysis practice within AI for FP&A and the broader AI Financial Modelling & Artificial Intelligence in Finance domain.
Where Machine Learning Adds Value: Driver Ranking¶
Machine learning can analyse historical data to rank which drivers have historically most influenced a model's output, directing a modeller's attention toward the variables most worth testing rigorously rather than spreading testing effort evenly across every possible input. This ranking is a prioritisation aid, not a determination of what to test, since a driver's historical influence does not guarantee it captures every genuinely novel forward-looking risk.
Where the Line Holds: Range-Setting Remains a Modeller's Input¶
The specific sensitivity ranges tested against each driver, how far up and down a given assumption is flexed, should remain a modeller's defined, documented, and auditable input. This is the same principle applied throughout this domain, addressed in AI in Financial Modelling: AI can accelerate identifying what to prioritise, but the specific parameters of the analysis itself should remain a transparent, defensible modelling decision.
Drafting Sensitivity Commentary¶
Generative AI can draft commentary explaining the results of a sensitivity analysis, which output moved most in response to which driver, following the same drafting-with-verification pattern set out in AI for Financial Analysts. This commentary should be checked against the actual tested ranges and resulting output figures before being finalised, since a fluent narrative can misstate the direction or magnitude of a sensitivity result while still reading plausibly.
Common Construction Pitfalls¶
Treating a low driver ranking as evidence a variable is safe to ignore. A ranking based on historical influence does not capture every forward-looking risk a lower-ranked driver could still carry in a materially changed future scenario.
Allowing AI to set sensitivity ranges without modeller review. A range set without a documented, defensible modelling rationale weakens the auditability a sensitivity analysis is meant to provide.
Publishing AI-drafted commentary without checking it against tested output. Sensitivity commentary that misdescribes which driver moved output the most undermines confidence in the analysis, even if the underlying calculations were correct.
Recommended Practices¶
- Use machine learning driver ranking to prioritise testing effort, not to determine what can be skipped.
- Set and document sensitivity ranges as an explicit modeller decision, regardless of AI-assisted ranking input.
- Verify AI-drafted sensitivity commentary against the actual tested ranges and output before finalising it.
Continue Reading¶
Related Pillars¶
Related Technical Guides¶
How OXXON tests thisRun a free structural check with FMAE
Frequently Asked Questions
How does machine learning support sensitivity analysis?
By helping rank which drivers most influence a model's output based on patterns in historical data, directing a modeller's attention to the variables most worth testing rigorously, rather than determining the specific ranges to test.
Should the sensitivity ranges tested be generated by AI?
No. The specific ranges tested against each driver should remain a modeller's defined, documented, and auditable input, informed by AI-assisted driver ranking but set through the modeller's own judgement about plausible variation.
Does a low driver ranking mean a variable can be ignored in sensitivity testing?
No. A driver ranking based on historical influence is a prioritisation tool, not a determination that lower-ranked drivers are safe to ignore, since historical influence may not capture a genuinely novel forward-looking risk a lower-ranked driver could still carry.
How should AI-drafted sensitivity commentary be handled?
Checked against the actual tested ranges and resulting output figures before being finalised, following the same number tie-out discipline applied to any AI-drafted commentary across this domain.
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.
AI for FP&A
FP&A is one of the finance functions where AI adoption has progressed furthest, spanning budgeting, forecasting, variance analysis, and management reporting. This guide maps AI application across the FP&A cycle, distinguishes tasks where machine learning-driven prediction is well matched from tasks better suited to generative AI drafting, and sets the frame for the more specific budgeting, forecasting, and analysis guides that follow it in this domain.
AI Scenario Planning
AI-assisted scenario planning uses generative AI to draft a wider range of plausible scenario variations around a base case than a modeller might generate manually, accelerating the ideation stage of scenario construction. This guide sets out that role, why an AI-drafted scenario must still be checked for internal consistency before use, and why probability weighting across scenarios remains a judgement exercise no AI tool performs on its own.
AI Forecasting Models
AI forecasting models use machine learning to predict a future value from patterns learned in historical data, complementing traditional driver-based forecasting rather than replacing it. This guide sets out the training data requirements a machine learning forecast depends on, how its accuracy should be measured and monitored over time, and the specific forecasting tasks where a machine learning approach adds genuine value over a traditional driver-based model.
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.