8 MIN · WORKFLOW 1 OF 10
Stop Asking AI Vague Finance Questions
AI does not know what ‘good analysis’ means unless you define the job. Give it a role, the business context, the actual data, a precise task, checks and the output shape. This turns a plausible-sounding answer into work you can inspect. In finance, the checks matter most: require formulas, source references, missing-data flags and a clear split between fact and interpretation.
“Analyse this balance sheet.”
AI must guess the job, period, checks and output.“Compare FY25 and FY26. Calculate key movements.”
The task is clearer, but the quality controls are missing.Role + context + data + task + checks + output.
The answer becomes traceable, reviewable and reusable.THE SIX-PART FINANCE PROMPT
ROLE: Act as a careful FP&A analyst supporting a monthly management review. CONTEXT: You are analysing a mid-sized Indian consumer-products company. Management wants insight, not a restatement of the numbers. Currency is ₹ lakh. DATA: Use only the figures I paste below. Treat blanks as missing, not zero. TASK: Calculate period-on-period changes, identify the three most decision-relevant movements, explain likely business drivers as hypotheses, and list the evidence needed to confirm each hypothesis. CHECKS: Recalculate totals, show formulas, flag inconsistent units and distinguish fact from inference. Do not invent missing figures. Show every formula used, separate facts from interpretations, and end with a verification checklist. State that this is decision support, not accounting or investment advice. OUTPUT: 1) five-line executive summary, 2) calculation table, 3) three hypotheses with confidence labels, 4) questions for management, 5) verification checklist.
Copy this realistic sample into the prompt. Keep the units and period labels intact.
Weak prompt: “Analyse this balance sheet.” Better prompt: “Compare FY25 and FY26, calculate key movements and explain the largest changes.” Professional prompt: use the six-part prompt above, then paste a clean table with units and period labels.
Does your output find these signals?
Your wording may differ. Recalculate the numbers yourself before accepting the conclusion.
How a finance professional takes this further
Lock the source of truth before prompting. Export a clean range from Excel, retain the original file, specify units once and reconcile totals independently. A professional never treats a fluent explanation as evidence.
Take a finance question you asked recently. Rewrite it with all six parts. Circle the one check that would catch the costliest possible error.
SAVE FOUR MORE
Prompts for the same workflow
Missing-data detector
Review the pasted finance data before analysing it. List missing periods, blank cells, mixed units, broken totals, duplicate rows and ambiguous labels. Do not begin analysis until the data-quality issues are summarised.
Assumption register
Create an assumption register for this analysis with columns: assumption, why needed, evidence available, confidence, sensitivity and owner who should confirm it.
Calculation verifier
Independently recalculate every percentage, subtotal and ratio in the draft below. Show the formula, inputs, your result and whether the draft is correct. Do not rewrite the narrative.
Finance follow-up coach
Read this AI-generated analysis and propose five follow-up questions that would reduce uncertainty most. Rank them by decision impact and explain what evidence would answer each.