AI in financial modelling
Our approach to AI
We advocate using AI selectively to help modellers understand, review and improve financial models—not to replace their judgement.
Effective use requires modellers to understand the calculations, give clear instructions and critically assess the results.
Our approach combines the structure of the FAST Standard, project-specific instructions and human review of AI recommendations and changes.
The FAST Standard
The FAST Standard builds upon modelling best practice with the aim of making financial models easy to read and understand.
The FAST Standard gives models a consistent, explicit structure.
When an AI agent interprets or modifies a model, these conventions provide clear patterns to follow and reduce the number of construction decisions it must make.
Our aim is to make AI-assisted work easier to direct and check.
Elements of the FAST Standard that are of value when working with AI are:
- 01Separating inputs, workings and outputs
- 02Short, simple and consistent formulas
- 03Use of colour coding
- 04In-built checks and alerts
- 05Minimising the use of macros
- 06Detailed units for every line item
- 07Unique and meaningful labels for every line item – no line items called ‘Total’
- 08Modelling timing using flags to separate the ‘when’ from the ‘how much’
- 09Modular and explicit calculations
- 10Using a consistent sheet structure as the major building block of the model
These conventions give the modeller a clearer basis for explaining a task to AI and checking its response – for example, by making units, timing and intermediate calculations explicit.
Download a
financial model
Learn more about
the FAST Standard
Context, boundaries, rules and skills
We give AI the project context, modelling conventions and instructions it needs for each task. We specify which worksheets or calculations are in scope, what may be changed and the output required. Alongside the FAST Standard, we provide our AI agents with the following:
Project brief - explaining the type and purpose of the model, the key stakeholders and their requirements, the sector and wider commercial context. This information is documented on the model’s User Guide sheet.
General rules - rules that apply to all workbooks, such as formatting rules, formula construction rules and workbook structure rules.
Specific rules - rules that apply to the current project, such as descriptions of worksheets and worksheet structure.
Skills - reusable instructions for completing defined tasks, such as a technical review.
Applications of AI
Modellers must remain responsible for the work. AI-generated explanations and review findings need to be checked against the model and relevant source information. We review AI-generated changes and test the resulting calculations before accepting them into the model.
Working with financial models
We use AI when operating and extracting information from financial models:
- Model familiarisation – understanding the model’s structure and calculation flow, from a whole-model overview to individual schedules such as fixed assets.
- Interrogating and explaining formulas – explaining how a model or calculation works, including the relevant assumptions, precedent calculations and references.
- Running scenarios and sensitivities – running predefined cases, storing and reviewing results.
Reviewing financial models
We use AI to support technical review and provide a second pair of eyes on analytical tasks:
- Technical review – for example identifying inconsistent formulas and embedded constants
- Analytical review – investigating unexpected trends and relationships in financial statements
- Analysing change – investigating movements in outputs between model versions or scenarios, and reviewing sensitivity results.
- Document review – reviewing the model against source documents, for example, the Credit Agreement
- Formula review – testing whether formulas perform the intended calculation, both with current inputs and when assumptions change.
Building and modifying models
The modeller defines the model’s structure and the scope of each AI task. We use AI to support the following work:
- Building calculation blocks – creating individual blocks or groups of blocks within a module.
- Improving formulas – suggesting simpler formulas and checking that they preserve the intended calculation.
- Working with macros – creating, explaining and debugging macros.
- Creating output presentations – creating dashboards and presentations.
Our emphasis is on sound modelling skills and informed use of AI: knowing where it can help, giving it a well-defined task and checking its work.