Your First AI Financial Analysis
Today you stop reading about AI and start using it. You're going to take a profit and loss statement, paste it into ChatGPT, and extract meaningful insights — variance commentary, trend analysis, and actionable recommendations — in minutes instead of hours.
This is the skill you'll use more than any other. Whether you're a bookkeeper preparing month-end reports, an analyst building board packs, or a CFO reviewing divisional performance, the ability to turn raw financial data into professional commentary at speed is transformative.
The anatomy of a good finance prompt
AI is only as good as what you ask it. A vague prompt gives you a vague answer. A structured prompt gives you board-ready output.
Here's the framework for finance prompts — the CRAF method:
Context — Tell the AI who you are and what the data represents. "I'm a management accountant reviewing Q3 results for a mid-size retail company."
Role — Tell the AI what role to play. "Act as a senior financial analyst writing commentary for the board."
Ask — Be specific about what you want. "Write variance commentary for each line item where the variance exceeds 5%."
Format — Specify the output format. "Use bullet points with the variance percentage and a one-sentence explanation for each."
A prompt using all four elements will consistently outperform a prompt that just says "analyse this data."
Your first analysis — step by step
Here's a sample P&L you can paste into ChatGPT right now. In practice, you'd use your own anonymised data.
Sample Q3 P&L — Retail Division
| Line Item | Actual | Budget | Variance |
|-----------|--------|--------|----------|
| Revenue | £2,450,000 | £2,600,000 | (£150,000) |
| Cost of Goods Sold | £1,470,000 | £1,430,000 | (£40,000) |
| Gross Profit | £980,000 | £1,170,000 | (£190,000) |
| Staff Costs | £320,000 | £340,000 | £20,000 |
| Marketing | £85,000 | £110,000 | £25,000 |
| Rent & Utilities | £95,000 | £92,000 | (£3,000) |
| Operating Profit | £480,000 | £628,000 | (£148,000) |
Now try this prompt: "I'm a management accountant at a mid-size retail company. Here are our Q3 actuals vs budget. Act as a senior financial analyst and write variance commentary for the board pack. For each line item, state the variance in pounds and percentage, explain the likely cause, and flag any items that need management attention. Use a professional but accessible tone."
Refining your output — the iteration loop
Your first AI output is a draft, not a final product. The real skill is in the iteration:
Round 1: Generate — Paste data and prompt. Get the initial analysis.
Round 2: Refine — Ask follow-up questions. "The revenue shortfall was actually due to a delayed product launch, not footfall. Can you update the commentary?" AI will adjust instantly.
Round 3: Extend — Ask for more. "Now add a forward-looking section with three recommendations for Q4." The AI builds on the context already established.
Round 4: Format — Ask for the output in your preferred format. "Reformat this as a table with columns for Line Item, Variance, RAG Status, and Commentary."
This loop — generate, refine, extend, format — is the workflow you'll use daily. Each round takes seconds, and the output gets sharper every time.
Common mistakes to avoid
Mistake 1: Pasting raw data without context. Don't just paste numbers. Tell the AI what industry, what time period, what level of detail you need. Context transforms the output.
Mistake 2: Accepting the first output. The first draft is rarely perfect. Iterate. Correct. Refine. That's where the value is.
Mistake 3: Including confidential details. Remember yesterday's rules — anonymise client names, remove account numbers, use aggregated data where possible.
Mistake 4: Skipping verification. AI can invent plausible-sounding explanations. If it says "the variance is due to seasonal trends," verify that's actually true for your business. Your job is to validate, not to transcribe.