Using AI in FP&A: Better Analysis Requires More Than a Good Prompt

Artificial intelligence is quickly becoming an important part of financial planning and

analysis. Used correctly, AI can help finance professionals think through complex problems, improve financial models, write and troubleshoot formulas, identify analytical approaches, and communicate findings more effectively.

However, AI is not a replacement for financial judgment, business knowledge, or careful review. My experience has been that the best results come from combining AI capabilities with an experienced finance professional who understands the business, challenges the output, and knows what questions to ask.

AI rarely produces the best answer on the first attempt

One of the biggest misconceptions about AI is that a single well-written prompt will produce a finished answer.

That is rarely how I use it.

Getting a useful result often requires several iterations. I may begin by describing the business issue, available data, management’s concerns, and the decision the analysis needs to support. I then review the response, identify gaps, correct assumptions, and ask increasingly specific follow-up questions.

A typical process might include:

1. Defining the business question and intended decision.

2. Asking AI to suggest an analytical structure.

3. Challenging the proposed logic and assumptions.

4. Adding business-specific constraints and operating details.

5. Comparing the output with source data and established financial relationships.

6. Revising the analysis until it is accurate, understandable, and actionable.

AI can accelerate the thinking process, but it does not eliminate the need to think.

In fact, experienced judgment may become even more important because AI can produce an answer that sounds convincing while still containing an incorrect assumption, formula, interpretation, or conclusion.

Every AI-assisted analysis must be checked

FP&A work frequently influences hiring, spending, pricing, inventory, financing, and investment decisions. A plausible-looking answer is not sufficient.

I treat AI-generated work as a starting point that must be validated. Depending on the assignment, that review can include:

• Reconciling results to financial statements and source systems

• Testing formulas with known values

• Reviewing signs, units, periods, and accounting classifications

• Comparing totals across supporting schedules

• Performing reasonableness checks

• Stress-testing key assumptions

• Reviewing unusual trends and outliers

• Confirming that the analysis answers the actual business question

The ability to check the work is one reason financial and operational experience remains so valuable. You need to understand what a reasonable result should look like before relying on the output.

The AI tools I use most frequently

ChatGPT Business

I use the paid business version of ChatGPT extensively for thinking through analytical logic, structuring financial models, evaluating alternative approaches, and improving how financial findings are communicated.

Examples include:

• Designing the structure of a forecasting or profitability model

• Identifying important analytical questions

• Thinking through cash-flow and working-capital drivers

• Developing approaches to customer, product, or regional profitability

• Reviewing forecasting assumptions

• Structuring KPI and management-reporting packages

• Brainstorming sensitivities and scenario analyses

• Improving board and management presentations

• Translating complex financial findings into clear action steps

ChatGPT is especially useful as an iterative thinking partner. I can challenge an initial recommendation, add operational context, ask for alternatives, and continue refining the logic.

OpenAI states that information submitted through ChatGPT Business is not used to train its models by default and that business data is encrypted at rest and in transit. Organizations should still review their own policies, workspace settings, client requirements, and intended use before entering confidential information. OpenAI business-data privacy information

Microsoft 365 Copilot

I use Microsoft 365 Copilot frequently for formula development, model improvements, troubleshooting, and working more efficiently within Excel and the Microsoft 365 environment.

Examples include:

• Developing and improving Excel formulas

• Troubleshooting formula errors

• Simplifying complicated calculations

• Suggesting alternative modeling approaches

• Identifying patterns and exceptions

• Organizing model documentation

• Summarizing financial findings

• Improving written explanations and presentations

I particularly value Microsoft 365 Copilot when working with financial models because it operates within the Microsoft 365 environment and its existing permissions and data protections. Microsoft states that prompts, responses, and organizational data accessed through Microsoft Graph are not used to train the foundation models used by Microsoft 365 Copilot.

That does not eliminate the need for responsible data governance. Permissions, file access, retention policies, and company security practices still matter. Copilot can access information a user is already authorized to view, so organizations need to maintain appropriate Microsoft 365 permissions. Microsoft 365 Copilot privacy and security information

Confidentiality must come first

Finance teams work with some of a company’s most sensitive information, including forecasts, payroll, pricing, customer profitability, acquisition plans, banking arrangements, and board materials.

Before using AI with company information, I consider:

• Which AI product and account type are being used

• The provider’s current data-use and retention policies

• The company’s internal AI and information-security policies

• Client contractual and confidentiality requirements

• Whether the information can be anonymized or summarized

• Whether names, account numbers, personal information, or other identifiers should be removed

• Whether the same objective can be accomplished using assumptions or sample data

Using a business-grade AI product is an important starting point, but it is not permission to upload everything. Good judgment and data minimization remain essential.

Where AI creates the most value in FP&A

The greatest benefit is not simply completing the same work faster. AI can help finance professionals explore more possibilities and ask better questions.

For example, instead of stopping after identifying that gross margin declined, an AI-assisted process can help organize questions around:

• Product and customer mix

• Pricing changes

• Material or labor costs

• Freight and fulfillment expenses

• Volume and capacity utilization

• Geographic performance

• One-time versus recurring drivers

The finance professional must still determine which questions are relevant, obtain reliable data, validate the analysis, and convert findings into recommendations.

That combination of technology, financial expertise, and business judgment is where I believe AI provides the most value.

AI-enabled finance still requires experienced leadership

I use AI tools because they help me work more efficiently, evaluate more alternatives, and improve the quality and clarity of financial analysis. I do not use them as a substitute for understanding the business or taking responsibility for the final work.

For companies seeking interim or fractional CFO, FP&A, Controller, or finance-transformation support, the best outcome comes from combining modern analytical tools with hands-on financial leadership.

The objective is not to produce more analysis. It is to produce accurate analysis that helps management make better decisions.