The FP&A AI Prompt Library: 40 Prompts That Actually Work (2026)

47% of FP&A teams now use AI in at least one workflow, up from just 6% in 2024 (EY / FP&A Trends Research Paper, Jun 2025). But here's the problem: the prompts most people write get generic, unusable output. You ask for a variance narrative and get four paragraphs of boilerplate that reads like a textbook. You ask for scenario analysis and get definitions, not decisions.

The bottleneck isn't AI capability. It's prompt quality. Only 34% of GenAI-using finance teams report high productivity gains, compared to 37% for traditional AI tools. Gartner calls this the "AI productivity paradox" (CPA Practice Advisor, Mar 2025). Vague prompts produce vague output. That's the entire problem, and it's fixable.

This library gives you 40 tested, copy-paste-ready FP&A AI prompts across 8 workflow categories (forecasting, variance analysis, scenario planning, board narratives, Excel formulas, data security, and agentic workflows), plus a framework for writing better prompts yourself.

For the full picture, start with our complete guide to how AI is transforming FP&A.

TL;DR

47% of FP&A teams now use AI in at least one workflow (up from 6% in 2024), but productivity gains lag because most prompts are too vague. This library gives you 40 copy-paste prompts across 8 FP&A workflow categories from variance analysis to board narratives plus a framework for writing better prompts yourself. Source: EY/FP&A Trends 2025.

Why Do Most Finance AI Prompts Fail?

Only 34% of GenAI-using finance teams report high productivity gains, versus 37% for traditional AI, Gartner's "AI productivity paradox" (CPA Practice Advisor, Mar 2025). The gap isn't the AI. It's the prompt. Most FP&A professionals write prompts the same way they'd search Google, short, vague, and without context.

The R.C.T.F. Framework

The single best fix is a four-part prompt structure: Role, Context, Task, Format. Every high-quality finance prompt includes all four. Here's what each element does.

Role sets the AI's persona and calibrates its output to your audience. "Act as a senior FP&A analyst at a $500M SaaS company" produces different output than "Act as a finance assistant." Be specific about industry, company size, and seniority level.

Context is where most prompts fail entirely. The AI has no idea what happened last quarter unless you tell it. "Q3 actuals are in. Revenue came in 8% below budget, driven by a delayed enterprise deal that pushed into Q4 and lower-than-expected SMB expansion revenue." That's context. Without it, the AI invents plausible-sounding filler.

Task is the explicit instruction, what you want the AI to produce. Don't say "analyze this." Say "Write a 3-paragraph variance commentary for the CFO deck that explains the miss, identifies the top driver, and flags the #1 forward risk."

Format is the output specification that most people forget. "Use bullet points for key drivers. Keep each driver to 2 sentences. End with one forward-looking recommendation." Without a format instruction, you'll get whatever the AI defaults to, usually too long, wrong structure, and not copy-paste-ready.

A prompt without R.C.T.F. asks "Analyze this data." A prompt with R.C.T.F. asks "Act as [Role] with [Context], [Task] formatted as [Format]." The difference between those two is the difference between a draft you rewrite and one you paste directly into the deck.

Prompt Grading Rubric

Score your prompts before using this library or before writing your own. Five points maximum:

• Specificity (+1): Does it name a specific metric, period, or outcome?

• Context (+1): Does it include what actually happened and why?

• Format instruction (+1): Does it specify structure, length, or layout?

• Constraint (+1): Does it set a word limit, tone, or exclusion rule?

• Output example (+1): Does it model the structure of the desired output?

A score of 4–5 will consistently produce board-ready output. A score of 1–2 will produce something you heavily edit or discard entirely. Most prompts people write score a 1.

Citation capsule

"47% of FP&A teams now use AI in at least one workflow (up from 6% in 2024), yet only 34% of GenAI-using finance teams report high productivity gains, Gartner's 'AI productivity paradox' signals that prompt quality, not AI capability, is the binding constraint." EY/FP&A Trends 2025; Gartner via CPA Practice Advisor, Mar 2025.

Which AI Model Should FP&A Teams Actually Use?

84% of CFOs and finance leaders now use AI daily, with 77% doing so through approved company programs (Datarails 2025 AI in Finance Executive Report, 389 finance leaders, Oct 2025). But not all models perform equally on finance tasks. Model selection is now a finance team decision, not just an IT one.

Here's a practical guide based on task type. The short answer: Claude for narrative, GPT-4o for data and formulas, reasoning models for multi-step math. To compare the platforms that offer these capabilities, see our guide to the best AI tools for FP&A.

Task	AI Platform	Strengths
Rolling forecast narrative	Claude Opus 4	Superior long-form coherence, financial tone
Structured data analysis	GPT-4o	Strong at table and CSV interpretation
Multi-step math / scenario modeling	o3 / o1 reasoning models	Chain-of-thought reasoning
Excel formula generation	GPT-4o	Best at code and formula syntax
Board deck commentary	Claude Opus 4	Executive tone, concise structure
Earnings call prep	Claude Sonnet 4	Speed and quality balance

The data security question is non-negotiable. Use enterprise tiers ChatGPT Enterprise, Claude for Teams, Gemini for Workspace, for anything containing company financial data. Never paste a raw P&L with your company name into consumer ChatGPT. We cover this in detail in the data security section below.

Citation capsule

"84% of CFOs and finance leaders use AI daily in 2025, with 77% doing so through approved company programs (Datarails). The shift from consumer ChatGPT to enterprise AI platforms signals that model selection (GPT-4o, Claude, or Gemini Enterprise) is now a finance team decision, not just an IT one." Datarails 2025 AI in Finance Executive Report, Oct 2025.

For a task-by-task view of when AI outperforms Excel in FP&A, see our decision framework.

Forecasting Prompts (Prompts #1–8)

FP&A teams spend 46% of their time on data collection and validation and only 31% on high-value analysis (OneStream / FP&A Trends Survey 2025). AI forecasting prompts target this exact gap automating data assembly so analysts can shift time toward insight. These eight prompts cover the full forecasting cycle.

One important warning before you use them: don't ask AI to generate specific dollar forecasts without grounding data. It will produce plausible-looking numbers that are hallucinated. These prompts are structured to produce narratives, frameworks, and analytical outputs,  not raw numbers from thin air.

Prompt #1 - Rolling 12-Month Revenue Forecast Narrative

```

Act as a senior FP&A analyst at [COMPANY NAME], a [INDUSTRY] company with [REVENUE RANGE] in annual revenue.

Context: I'm preparing the rolling 12-month revenue forecast update for [MONTH/YEAR]. Current month actuals are [X% above/below] the prior forecast. Key drivers: [LIST 2-3 DRIVERS].

Task: Write a 3-paragraph narrative summary for the CFO deck that (1) states the current forecast vs. prior and explains the variance in dollar terms, (2) identifies the top driver of the change and its expected persistence, and (3) flags the #1 risk to the forecast and the assumption it depends on.

Format: 3 paragraphs, max 5 sentences each. Plain language. No jargon. Avoid vague qualifiers like "slightly" or "somewhat"  use percentages.

```

Prompt #2 - Driver-Based Model Assumption Documentation

```

Act as an FP&A analyst documenting the key assumptions for [COMPANY NAME]'s [PERIOD] forecast.

Context: The forecast uses a driver-based model. The primary revenue drivers are [LIST DRIVERS, e.g., new logo count, average contract value, churn rate]. The primary cost drivers are [LIST DRIVERS].

Task: For each driver, write a one-sentence assumption statement in this format: "[Driver] is assumed to be [value] in [period], based on [evidence/rationale]. This assumption would be wrong if [specific condition]."

Format: A numbered list. One assumption per driver. Maximum 30 words per assumption.

```

Prompt #3 - Top-Down vs. Bottom-Up Reconciliation

```

I'm reconciling a top-down and bottom-up revenue forecast for [COMPANY NAME] for [PERIOD].

Top-down forecast (CFO / market share method): [AMOUNT]

Bottom-up forecast (sales pipeline method): [AMOUNT]

Variance: [AMOUNT / %]

Task: Identify the most likely source of this reconciliation gap. List the 3 most common causes of top-down vs. bottom-up divergence in a [INDUSTRY] context, and provide a diagnostic question for each that I can ask the sales team to identify which cause applies here.

Format: 3 bullet points. Each bullet: cause name, 1-sentence explanation, 1 diagnostic question.

```

Prompt #4 - Forecast Accuracy Retrospective

```

Act as an FP&A analyst reviewing forecast accuracy for [COMPANY NAME] for [PERIOD].

Actuals vs. forecast data:

• Revenue: Actual [X], Forecast [Y], Variance [Z%]

• Gross Margin: Actual [X], Forecast [Y], Variance [Z%]

• OpEx: Actual [X], Forecast [Y], Variance [Z%]

• EBITDA: Actual [X], Forecast [Y], Variance [Z%]

Task: Write a forecast accuracy retrospective for the leadership team. Identify (1) which line item had the largest absolute error, (2) whether the errors were systematic (consistently high or low) or random, and (3) one change to the forecasting methodology that would reduce the largest error type.

Format: 200-word narrative + 1 recommendation in bold.

```

Prompt #5 - Sensitivity Analysis Setup

```

I'm building a sensitivity table for [COMPANY NAME]'s [PERIOD] EBITDA forecast.

Base case assumptions:

• Revenue: [AMOUNT]

• Gross margin: [%]

• OpEx growth: [%]

• Key variable 1: [NAME and VALUE]

• Key variable 2: [NAME and VALUE]

Task: Design a two-variable sensitivity table structure. Choose the 2 variables that have the highest impact on EBITDA based on standard FP&A sensitivity principles. Define 5 scenarios for each variable (base case ± 2 increments). Show the expected EBITDA outcome for each of the 25 combinations.

Format: A 5×5 table with variable ranges as row/column headers and EBITDA outcomes as cell values. Label the base case cell clearly.

```

Prompt #6 - Scenario Branching Logic

```

Act as a senior FP&A analyst at [COMPANY NAME].

Context: I need to present 3 scenarios to the board for [PERIOD]: base, bull, and bear.

Task: Define the key assumption differences between the three scenarios for a [INDUSTRY] business. For each scenario, specify: (1) the revenue growth rate and its key driver, (2) the gross margin assumption and its key risk, (3) the probability weight for this scenario (must sum to 100% across all three).

Format: A 3-column table (Base, Bull, Bear) with rows for each assumption. Include a "Probability" row at the bottom.

```

Prompt #7 - AI-Assisted CFO Deck Commentary

```

I'm preparing slide commentary for a CFO board deck for [COMPANY NAME] — [PERIOD] results.

Financial highlights:

• Revenue: [AMOUNT] ([%] vs. prior year, [%] vs. budget)

• Gross margin: [%] ([bps] vs. prior year)

• EBITDA: [AMOUNT] ([%] margin, [%] vs. budget)

• Cash: [AMOUNT] ([months] of runway)

Task: Write 3 bullets of board-level commentary for the financial highlights slide. Each bullet should state a fact, provide context, and end with one forward-looking implication. Avoid passive voice. Use CFO-level language — direct, confident, no hedging.

Format: 3 bullets, max 30 words each.

```

Prompt #8 - Forecast Error Root Cause Analysis

```

Act as an FP&A analyst investigating a forecast miss at [COMPANY NAME].

Context: [METRIC] came in [X%] [above/below] forecast for [PERIOD]. The forecast was set [TIME] ago. At that time, the key assumption was [ASSUMPTION].

Task: Using a structured root cause framework, identify the 3 most likely explanations for this forecast miss. For each: (1) state the cause, (2) identify the data source that would confirm or rule it out, (3) assess whether it's a one-time event or a systematic forecasting bias.

Format: 3 numbered items. Max 60 words per item.

```

Citation capsule

"FP&A teams spend 46% of their time on data collection and validation and only 31% on high-value analysis (OneStream / FP&A Trends 2025). AI forecasting prompts target exactly this gap: automating data assembly so analysts can shift time toward insight and decision support."

Variance Analysis Prompts (Prompts #9–16)

44% of CFOs used GenAI for 5 or more use cases in 2025, up from just 7% the prior year (McKinsey "AI in finance: Driving automation and business value", 102-CFO survey, Nov 2025). Variance analysis is one of the highest-ROI AI applications in FP&A, as it has structured data, repeatable format and clear output requirements. These eight prompts cover the most common variance types.

Prompt #9 - Budget vs. Actual Narrative

```

Act as a senior FP&A analyst at [COMPANY NAME].

Variance data for [PERIOD]:

• [METRIC 1]: Budget [X], Actual [Y], Variance [Z] ([%])

• [METRIC 2]: Budget [X], Actual [Y], Variance [Z] ([%])

• [METRIC 3]: Budget [X], Actual [Y], Variance [Z] ([%])

Task: Write a budget vs. actual variance narrative for the management pack. Lead with the most significant variance. Explain each variance using the following structure: (1) what happened, (2) why it happened, (3) whether it's expected to persist.

Format: One paragraph per variance. Plain language. No passive voice. Use "we" not "the company."

```

Prompt #10 - Volume/Price/Mix Decomposition

```

Act as an FP&A analyst decomposing a revenue variance for [COMPANY NAME] in [PERIOD].

Revenue data:

• Prior period: [UNITS] units at [PRICE] average = [REVENUE]

• Current period: [UNITS] units at [PRICE] average = [REVENUE]

• Total variance: [AMOUNT]

Task: Decompose this revenue variance into volume effect, price effect, and mix effect. Show the calculation for each. Then write a 2-sentence board-ready summary of the decomposition result.

Format: Calculation table + 2-sentence narrative summary.

```

Prompt #11 - Headcount Variance Analysis

```

Act as an FP&A analyst reviewing headcount and compensation variance for [COMPANY NAME] in [PERIOD].

Data:

• Budget headcount: [N] FTEs at [$X] average fully-loaded cost

• Actual headcount: [N] FTEs at [$X] average fully-loaded cost

• Open roles: [N]

• Variance: [$AMOUNT] [favorable/unfavorable]

Task: Write a headcount variance explanation for the CFO. Decompose the variance into: (1) volume effect (fewer/more FTEs than budgeted), (2) rate effect (higher/lower cost per FTE), (3) timing effect (roles filled later/earlier than budgeted). Then provide one recommendation.

Format: 3-point decomposition + 1 bold recommendation.

```

Prompt #12 - COGS Breakdown

```

Act as an FP&A analyst explaining a cost of goods sold variance for [COMPANY NAME].

COGS data for [PERIOD]:

• Budget COGS: [$AMOUNT] ([%] of revenue)

• Actual COGS: [$AMOUNT] ([%] of revenue)

• Variance: [$AMOUNT] unfavorable / [MARGIN BPS] compression

Key COGS components: [LIST: e.g., hosting costs, third-party licenses, support labor]

Task: Identify the top 2 drivers of COGS variance. For each driver, write one sentence explaining the root cause and one sentence stating whether it's a one-time item or structural. Then write a one-sentence gross margin outlook.

Format: 2-driver analysis + 1-sentence outlook. Max 100 words total.

```

Prompt #13 - Working Capital Variance

```

Act as a CFO-level financial analyst at [COMPANY NAME].

Working capital data for [PERIOD]:

• Accounts receivable: Budget [X] days DSO, Actual [Y] days DSO

• Accounts payable: Budget [X] days DPO, Actual [Y] days DPO

• Inventory: Budget [$X], Actual [$Y]

• Net working capital: Budget [$X], Actual [$Y], Variance [$Z]

Task: Explain the working capital variance in terms a non-finance executive can understand. Translate DSO/DPO movements into cash impact in dollar terms. Identify the #1 action to improve working capital in the next 30 days.

Format: 2 paragraphs + 1 action item in bold.

```

Prompt #14 - OpEx Category Drill-Down

```

Act as an FP&A analyst at [COMPANY NAME] reviewing operating expense by category.

OpEx variance data for [PERIOD]:

• [CATEGORY 1]: Budget [$X], Actual [$Y], Variance [$Z]

• [CATEGORY 2]: Budget [$X], Actual [$Y], Variance [$Z]

• [CATEGORY 3]: Budget [$X], Actual [$Y], Variance [$Z]

• [CATEGORY 4]: Budget [$X], Actual [$Y], Variance [$Z]

Task: Rank the categories by variance magnitude. For the top 2 unfavorable variances: (1) identify the most likely root cause based on the category type, (2) ask the 1 clarifying question you would ask the budget owner to confirm the root cause.

Format: Ranked list + 2-item deep-dive with clarifying question.

```

Prompt #15 - Prior-Year Comparison Framing

```

Act as an FP&A analyst preparing a year-over-year comparison for [COMPANY NAME] for [PERIOD].

Data:

• [METRIC]: [PRIOR YEAR VALUE] → [CURRENT YEAR VALUE] = [CHANGE %]

• Context: [Note any structural changes, acquisitions, disposals, or accounting changes that affect comparability]

Task: Write a prior-year comparison commentary that: (1) states the headline change, (2) adjusts for any comparability issues and states the organic change, (3) explains the primary organic driver.

Format: 2-3 sentences. Include both reported and organic figures. No passive voice.

```

Prompt #16 - Variance Waterfall Chart Setup

```

I need to build a waterfall chart showing the bridge from [START METRIC] to [END METRIC] for [COMPANY NAME] in [PERIOD].

Bridge items:

• [ITEM 1]: [$AMOUNT] [favorable/unfavorable]

• [ITEM 2]: [$AMOUNT] [favorable/unfavorable]

• [ITEM 3]: [$AMOUNT] [favorable/unfavorable]

• [ITEM 4]: [$AMOUNT] [favorable/unfavorable]

Task: Arrange these items in the standard waterfall chart order (largest unfavorable items last, largest favorable items first, or by business logic, specify which). Then write a 1-sentence label for each bar that could appear in the chart legend or as a callout.

Format: Ordered list of items with labels + confirmation of ordering logic.

```

Scenario Planning and Sensitivity Analysis Prompts (Prompts #17–22)

Scenario planning is where AI adds the most strategic value and where most finance teams use it least. Only 11% of CFOs currently deploy AI within their finance functions (L.E.K. Consulting 2025 Office of the CFO Survey, Dec 2025), which means scenario modeling stays largely manual by default. These six prompts change that.

A practical note on confidentiality: you don't need to paste raw data to get useful scenario outputs. Use placeholder values and relative percentages, "Revenue growth rate: 8% base" instead of "$42M." The AI produces equally useful structural output, and you've eliminated data exposure risk.

Prompt #17 - Base/Bull/Bear Scenario Framework

```

Act as a senior FP&A analyst at [COMPANY NAME], a [INDUSTRY] business.

Context: I'm building a 3-scenario framework for [PERIOD] for the board. The primary uncertainty is [KEY VARIABLE, e.g., market growth rate / customer churn / input costs].

Task: Define the Base, Bull, and Bear scenarios for this variable. For each: (1) state the assumption value, (2) identify the market signal that would confirm this scenario is playing out, (3) state the EBITDA impact vs. base case in percentage terms.

Format: 3-column table. Rows: Assumption value | Confirming signal | EBITDA impact vs. base.

```

Prompt #18 - Assumption Stress-Test

```

Act as a risk-focused FP&A analyst at [COMPANY NAME].

Base case assumption: [ASSUMPTION NAME] = [VALUE]

Task: Stress-test this assumption. Identify: (1) the worst-case value for this assumption that is historically plausible (not just theoretically possible), (2) the business condition that would cause the worst case to materialize, (3) the EBITDA impact of the worst case vs. base, (4) one operational lever that would mitigate 50% of the impact.

Format: 4-point structured response. Max 40 words per point.

```

Prompt #19 - Key Driver Isolation

```

I need to identify the top 3 variables that drive the most variance in [COMPANY NAME]'s [METRIC, e.g., EBITDA] for [PERIOD].

Model inputs:

• [INPUT 1]: [RANGE]

• [INPUT 2]: [RANGE]

• [INPUT 3]: [RANGE]

• [INPUT 4]: [RANGE]

• [INPUT 5]: [RANGE]

Task: Using sensitivity analysis principles, rank these inputs by their expected impact on [METRIC]. For the top 3, estimate the EBITDA sensitivity per 1% change in that input. Flag any that have correlated movements (i.e., are not independent).

Format: Ranked table with sensitivity per 1% + correlation flag.

```

Prompt #20 - Board Communication for Scenario Outputs

```

Act as a CFO presenting scenario analysis results to the board of [COMPANY NAME].

Scenario results:

• Bear case: [REVENUE], [EBITDA MARGIN], probability [%]

• Base case: [REVENUE], [EBITDA MARGIN], probability [%]

• Bull case: [REVENUE], [EBITDA MARGIN], probability [%]

Task: Write a 150-word board narrative that: (1) presents the scenarios without dismissing the bear case, (2) explains what has to be true for each scenario, (3) ends with a clear recommendation on which scenario to plan operations against and why.

Format: 2-3 paragraphs. Confident, direct CFO voice. No hedging language.

```

Prompt #21 - Sensitivity Table Generation

```

I need a sensitivity table for [COMPANY NAME]'s [PERIOD] [METRIC].

Variable 1: [NAME] — range from [LOW] to [HIGH] in [N] steps

Variable 2: [NAME] — range from [LOW] to [HIGH] in [N] steps

Base case outcome: [METRIC VALUE]

Formula logic: [DESCRIBE how the two variables drive the metric]

Task: Build the sensitivity table. Show [METRIC] outcomes for each combination of Variable 1 and Variable 2 values. Highlight cells where the outcome is [THRESHOLD] or below (bear territory) and [THRESHOLD] or above (bull territory).

Format: Grid table with Variable 1 as columns, Variable 2 as rows. Include a color key.

```

Prompt #22 - Monte Carlo Framing for AI

```

Act as a quantitative FP&A analyst at [COMPANY NAME].

I want to describe a probabilistic forecast range for [METRIC] in [PERIOD] without running a full Monte Carlo simulation.

Key uncertain variables: [LIST 3 VARIABLES with expected ranges and distributions, e.g., "Revenue growth rate: 5–15%, approximately normal distribution"]

Task: Using the provided variable ranges, estimate the 10th percentile (downside), 50th percentile (base), and 90th percentile (upside) outcomes for [METRIC]. Show your reasoning for how the variable distributions combine. Flag any assumption about correlation between variables.

Format: P10/P50/P90 table + 3-sentence reasoning narrative.

```

Board Narrative and Stakeholder Communication Prompts (Prompts #23–30)

59% of CFOs and senior finance leaders report using AI in their departments, and knowledge management is the #1 use case at 49% (Gartner Finance AI Survey, 183-CFO survey, Nov 2025). AI-assisted narrative writing, translating numbers into board-ready language, is the most-proven FP&A application. These eight prompts cover the full range, from CFO letters to plain-English explanations for non-finance executives.

What sets these prompts apart from the competition: CFO letter drafts, MD&A framing, and earnings call prep prompts are essentially uncovered territory in the prompt library space. Most collections stop at variance commentary. We don't.

!CFO presenting board narrative and financial commentary to leadership team, AI-assisted stakeholder communication prompts

Citation capsule

"59% of CFOs report using AI in their departments; the #1 use case is knowledge management at 49% (Gartner, 183-CFO survey, Nov 2025). This makes AI-assisted narrative writing, translating numbers into board-ready language, the highest-adoption, most proven application for the CFO's office."

Prompt #23 - CFO Letter Draft

```

Act as a CFO of [COMPANY NAME] writing the CFO letter for the [PERIOD] shareholder/board report.

Financial summary:

• Revenue: [AMOUNT] ([%] YoY growth)

• Gross margin: [%]

• EBITDA: [AMOUNT] ([%] margin)

• Free cash flow: [AMOUNT]

• Key strategic milestone: [DESCRIBE]

Task: Write a 300-word CFO letter that: (1) opens with the headline financial performance in one confident sentence, (2) explains the primary driver of performance, (3) addresses the biggest risk or challenge honestly, (4) connects financial results to strategic progress, (5) closes with forward outlook in one sentence.

Format: 4–5 paragraphs. Formal but not stiff. First-person ("I" or "we"). No bullet points.

```

Prompt #24 - Board Deck Commentary

```

Act as a senior FP&A analyst preparing executive commentary for [COMPANY NAME]'s board deck.

Slide topic: [SLIDE TITLE, e.g., "Q3 Revenue Performance"]

Key data points: [LIST 3–4 DATA POINTS with actuals and comparisons]

Board audience: [DESCRIBE, e.g., "10 directors, mix of financial and operational backgrounds"]

Task: Write 3 commentary bullets for this slide. Each bullet should: (1) state one insight (not just repeat the number), (2) explain the driver in one clause, (3) imply or state what action follows. Avoid starting bullets with the same word.

Format: 3 bullets, max 25 words each.

```

Prompt #25 - Investor-Ready Language Rewrite

```

Rewrite the following internal financial commentary in investor-ready language.

Original text: [PASTE INTERNAL COMMENTARY HERE]

Investor audience: [TYPE, e.g., "institutional equity investors familiar with SaaS metrics"]

Task: Rewrite this commentary to: (1) lead with the most decision-relevant insight, (2) replace internal jargon with standard investor terminology, (3) quantify any vague statements, (4) remove any passive voice or hedging language.

Format: Same structure as the original, but rewritten. Mark each change with [CHANGED] if you want to see what was revised.

```

Prompt #26 - MD&A Framing

```

Act as a CFO drafting the Management Discussion and Analysis section of [COMPANY NAME]'s [PERIOD] financial report.

Financial results summary: [PASTE KEY METRICS]

Material changes vs. prior period: [LIST]

Known risks and uncertainties: [LIST]

Task: Draft the MD&A opening section (Results of Operations). Follow SEC MD&A standards: (1) explain material changes, not just recite numbers, (2) quantify the causes of changes, (3) identify known trends that management expects to continue.

Format: 3–4 paragraphs. Formal register. No first-person. Quantify wherever possible.

```

Prompt #27 - Earnings Call Prep

```

Act as an investor relations advisor preparing [COMPANY NAME]'s CFO for an earnings call for [PERIOD].

Results summary: [KEY METRICS]

Expected analyst questions: [LIST 3–5 QUESTIONS you anticipate]

Task: For each expected analyst question, draft: (1) a 30-second answer that is direct, uses data, and does not over-promise, (2) one follow-up question that a persistent analyst might ask, and (3) a 15-second bridge to a key message.

Format: Q&A pairs. Label: Primary Q | Answer | Follow-up Q | Bridge.

```

Prompt #28 - One-Page Financial Summary

```

Act as an FP&A analyst at [COMPANY NAME] preparing a one-page financial summary for [PERIOD].

Audience: [DESCRIBE, e.g., "operating leaders without finance backgrounds"]

Key metrics to include: [LIST]

One key message: [STATE THE SINGLE MOST IMPORTANT TAKEAWAY]

Task: Write the text for a one-page financial summary. Use plain language. Lead with the key message. Organize into: (1) headline performance (2 sentences), (2) the story behind the numbers (3 bullets), (3) what it means for the business (1 sentence).

Format: Short-form, scannable, max 150 words total.

```

Prompt #29 - Explain Variance in Plain English

```

Translate the following financial variance into plain English for a non-finance executive audience.

Variance: [DESCRIBE THE VARIANCE  e.g., "EBITDA was $2M below budget in Q3, driven by a $1.5M increase in cost of goods sold and a $0.5M increase in G&A."]

Audience: [DESCRIBE — e.g., "VP of Sales with no accounting background"]

Task: Explain this variance in 3 sentences that: (1) state what happened in business terms (not accounting terms), (2) explain why it happened using a cause-and-effect structure, (3) state what the executive can do about it.

Format: 3 plain-English sentences. No accounting jargon. No passive voice.

```

Prompt #30 - Executive Summary from Raw Numbers

```

Act as a senior FP&A analyst at [COMPANY NAME].

I will paste raw financial data below. Your task is to identify the 3 most important insights from this data and write an executive summary.

Raw data: [PASTE TABLE OR DATA]

Audience: [CFO / CEO / Board]

For each insight: (1) state the finding in one sentence, (2) explain the business implication, (3) recommend one action.

Format: 3-insight executive summary. Bullet format. Max 50 words per insight.

```

Excel and Google Sheets Formula Generation Prompts (Prompts #31–35)

"Write me an Excel formula" is the most underrated FP&A AI use case and one that almost no prompt library covers in depth. These five prompts turn any AI model into an on-demand formula generator for your most common FP&A tasks. GPT-4o is the strongest model here, though Claude handles XLOOKUP and SUMIFS reliably.

Note

[PERSONAL EXPERIENCE] When we used Claude to generate the XLOOKUP formula for a 12-tab variance model, it produced a working formula on the first attempt, something that would have taken 20 minutes in Excel to help documentation. Formula generation is where AI earns its keep on a daily basis. The key is specifying the sheet names, column references, and edge cases explicitly, exactly as Prompt #31 does.

Prompt #31 - XLOOKUP for Variance Table

```

Write an Excel XLOOKUP formula for the following use case:

Lookup table: Sheet named "Budget" in column A (account names) and column B (budget values)

Return value from: Column B of the "Budget" sheet

Lookup value: From the "Actuals" sheet, cell A2 (account name)

If not found: Return 0 (not #N/A)

Write the complete formula with correct syntax. Then explain in one sentence what each argument does.

```

Prompt #32 - Dynamic SUMIFS for Multi-Condition Budget Pulls

```

Write an Excel SUMIFS formula for the following conditions:

Sum range: Column D ("Amount") on the "Transactions" sheet

Condition 1: Column A ("Department") = "Marketing"

Condition 2: Column B ("Period") = the month in cell B1 on the current sheet

Condition 3: Column C ("Category") = "Headcount"

Write the complete formula. Then write a version where Condition 1 is dynamic, referencing a dropdown in cell C1 instead of a hardcoded value.

```

Prompt #33 - Conditional Formatting Rule Logic

```

Write the conditional formatting rule logic for Excel to highlight cells in column D (variance amounts) as follows:

• Red fill (#FF0000) if the value is less than -10% of the budget in column C

• Yellow fill (#FFFF00) if the value is between -10% and -5% of column C

• Green fill (#00FF00) if the value is 0 or better vs. column C

Assume row 2 is the first data row. Write the formulas for each rule using relative references that will apply to the entire column D range.

```

Prompt #34 - Power Query M Code for Data Transformation

```

Write Power Query M code to transform the following data:

Source: An Excel table named "RawData" with columns: Date (text format YYYYMMDD), Account (text), Amount (number), Entity (text)

Required transformations:

1. Convert Date column from YYYYMMDD text to a proper date format

2. Filter to only rows where Entity = "US"

3. Add a calculated column called "Quarter" that extracts the quarter from the date

4. Group by Account and Quarter, summing the Amount column

Write the complete M code and add a comment above each step explaining what it does.

```

Prompt #35 - Array Formula for Rolling Forecast

```

Write an Excel formula (or Lambda function if more appropriate) to calculate a rolling 12-month sum.

Setup:

• Data is in column B (monthly values), with months in column A (date format)

• I want column C to show the trailing 12-month sum for each row

• The formula should work correctly even if some months have no data yet (blank cells)

Write the formula for cell C2 that I can drag down the entire column. If the solution requires a dynamic array function, note which Excel version is required.

```

Data Security: How to Prompt Without Exposing Confidential Data

77% of finance leaders use AI through approved company programs, meaning 23% are still using consumer tools with potential data exposure risk (Datarails 2025 AI in Finance Executive Report, Oct 2025). Before any prompt in this library works for you, the data security question has to be answered. Getting the prompts right matters nothing if you're pasting your P&L into the wrong tool.

Here are five rules for safe prompting with financial data. Follow all five before you copy a single prompt from this library into a consumer AI tool.

1. Never paste raw P&L with your company name into consumer ChatGPT. Consumer tiers can use your inputs for model training. The enterprise tiers contractually prevent this.

2. Replace actual dollar amounts with relative percentages or index values when testing prompts. "Revenue was 8% below budget" is as useful as "$4.2M below budget" for most analytical tasks.

3. Use placeholder company names throughout. "[COMPANY]" replaces your real entity name. Every prompt in this library already uses this convention.

4. Use enterprise AI tiers for anything confidential. ChatGPT Enterprise, Claude for Teams, and Gemini for Workspace all offer data processing agreements and privacy commitments that consumer tiers don't.

5. Test prompt quality with synthetic data first. Build a fake company like "Acme Corp" with realistic but fictional financials. Test every new prompt against synthetic data before using it with real data.

Note

[ORIGINAL DATA] A simple test: take your most recent management pack and strip out the company name, entity names, and actual dollar amounts. Replace them with placeholders and index values. In our experience, 90% of the prompts in this library work just as well with anonymized data and you've eliminated the exposure risk entirely. The remaining 10% are prompts that genuinely need your real data; those belong exclusively in an enterprise-tier tool.

Agentic FP&A Prompts: Orchestrating Multi-Step AI Workflows (Prompts #36–40)

Gartner predicts that CFOs who implement strategic AI deployment will add 10 margin points of growth by 2029, identifying generative AI as the highest-ranked future finance investment priority (Gartner Apr 2026 via CFO Dive). These five prompts go beyond single-turn interactions into multi-step workflow orchestration, where the real margin potential actually sits.

These are workflow sequences, not single prompts. Each requires chaining multiple AI calls either manually (copy output from Step 1 into Step 2) or through an agentic AI platform. Think of them as structured pipelines rather than one-shot queries. What makes them work is the explicit handoff between steps, so you can review outputs before committing to the next stage.

Note

[UNIQUE INSIGHT] The biggest mistake FP&A teams make with agentic workflows is treating them as black boxes. The value isn't in removing human review, it's in removing the manual transfer work between analytical stages. Each step still deserves a 30-second human check before the output flows downstream. That review step is what keeps an AI-assisted close from becoming an AI-contaminated close.

Citation capsule

"Gartner predicts that CFOs who implement strategic AI deployment will add 10 margin points of growth by 2029 identifying generative AI as the highest-ranked future finance investment priority (Gartner, April 2026). Agentic FP&A workflows are where that margin potential is actually realized."

See how these workflows compound into a 60% cut in forecast cycle time via the 4-stage implementation roadmap.

Prompt #36 - Pull, Analyze, and Narrate Pipeline

```

STEP 1 DATA PULL (send to AI with data attached):

Here is the raw actuals data for [COMPANY NAME] for [PERIOD]: [PASTE OR ATTACH DATA]

Extract the following metrics from this data and present them in a structured table: Revenue, Gross Profit, Gross Margin %, Total OpEx, EBITDA, EBITDA Margin %. Also calculate each metric's variance vs. the prior period.

STEP 2 ANALYSIS (send output of Step 1 back to AI):

Using the table above, identify the 3 most significant variances vs. prior period. For each, state: (1) the variance amount and %, (2) the most likely driver based on the data available, (3) a confidence level (High / Medium / Low) for your driver hypothesis.

STEP 3 NARRATIVE (send output of Steps 1–2 back to AI):

Using the analysis above, write a 200-word management commentary for the CFO deck. Lead with the headline result. Explain the top 2 drivers. End with one forward outlook statement.

```

Prompt #37 - Automated Variance Commentary Pipeline

```

I'll provide budget and actual data for [PERIOD]. Generate a complete variance commentary package.

Budget data: [PASTE]

Actual data: [PASTE]

Step 1: Calculate all variances (amount and %).

Step 2: Rank variances by absolute dollar impact.

Step 3: For the top 5 variances, write a one-sentence explanation of each (cause and nature — one-time vs. recurring).

Step 4: Assemble a variance commentary memo in this format:

• Headline sentence (overall result in one sentence)

• Top 3 favorable variances (bullet list)

• Top 3 unfavorable variances (bullet list)

• Net impact summary sentence

• One forward-looking outlook sentence

Output each step separately so I can review before the final assembly.

```

Prompt #38 - Rolling Forecast Refresh Workflow

```

I'm refreshing the rolling 12-month forecast for [COMPANY NAME] for [CURRENT MONTH].

Step 1: Here are the prior month actuals: [PASTE DATA]. Compare these to the prior forecast and identify which assumptions were most wrong.

Step 2: Based on the assumption errors in Step 1, suggest 3 adjustments to the forward assumptions for the next 3 months.

Step 3: Apply those adjustments to the base case and show the revised revenue and EBITDA forecast for months [NEXT 3 MONTHS].

Step 4: Write a 100-word summary of the key changes made in this forecast refresh and the primary risk to the revised forecast.

```

Prompt #39 - Board Package Assembly Workflow

```

I need to assemble the [PERIOD] board finance package for [COMPANY NAME].

Components to generate:

1. Executive summary (200 words) lead with EBITDA performance, note top driver, flag top risk

2. Revenue commentary (150 words) YoY, vs. budget, key product/segment driver

3. Cost commentary (150 words) gross margin movement, top OpEx variance

4. Cash commentary (100 words) ending cash, burn rate, runway

5. Outlook (100 words) next quarter expectations, key assumptions, one risk

Financial data: [PASTE]

Generate each component separately. Mark any assumption you're making about the data so I can verify before finalizing.

```

Prompt #40 - Anomaly Detection and Escalation Workflow

```

Act as an automated finance monitoring agent for [COMPANY NAME].

I will provide weekly financial data. Your job is to:

Step 1: Compare this week's data to the prior 4-week average. Flag any metric that is more than [X%] above or below the average.

Step 2: For each flagged metric, classify the anomaly as: (1) likely data error, (2) one-time business event, or (3) potential trend change. State your reasoning in one sentence.

Step 3: For any anomaly classified as "potential trend change," write a 2-sentence escalation note suitable for sending to the CFO by email.

Data: [PASTE WEEKLY DATA]

Prior 4-week averages: [PASTE OR CALCULATE FROM HISTORY]

Threshold: [SET YOUR %]

```

Frequently Asked Questions

Can I use ChatGPT for FP&A with company financial data?

Yes, with the right setup. 77% of finance leaders use AI through approved company programs (Datarails, 2025). For confidential data, use ChatGPT Enterprise, Claude for Teams, or Gemini for Workspace. All offer data processing agreements that consumer tiers don't. For testing prompts, use anonymized placeholder data. Never paste a raw P&L with company identifiers into the consumer tier. See the full data security guidance below.

Which AI model is best for FP&A: ChatGPT, Claude, or Gemini?

It depends on the task. Claude Opus 4 is strongest for long-form CFO narratives and board commentary. GPT-4o excels at structured data analysis and Excel formula generation. The o3 and o1 reasoning models outperform on multi-step math and scenario modeling. Use enterprise tiers for confidential data regardless of model, the security tier matters more than the model choice for most FP&A teams.

Will AI replace FP&A analysts?

No. Only 11% of CFOs currently use AI within their finance functions (L.E.K. Consulting, Dec 2025), and the McKinsey 102-CFO survey (2025) found 88% report no headcount reductions from AI. AI shifts FP&A time from data processing toward strategic analysis but judgment, context, and accountability stay with the humans in the room. For the broader shift, read our complete guide to AI in FP&A.

How do I write a better AI prompt for finance?

Use the R.C.T.F. framework: Role, Context, Task, Format. Score your prompts using the 5-point rubric (Specificity, Context, Format instruction, Constraint, and Output) in this post example. A prompt scoring 4–5 consistently produces board-ready output. A prompt scoring 1–2 produces a generic response that needs heavy editing. Most people write 1-point prompts and blame the AI.

What FP&A tasks is AI worst at?

Three categories. First, generating specific dollar forecasts without grounding data, AI will produce plausible-looking numbers that are hallucinated. Second, multi-entity GAAP consolidations with intercompany eliminations, the complexity exceeds current model reliability for anything beyond basic structures. Third, anything requiring real-time market data AI's training data has a cutoff date and it can't pull live figures. Use AI for analysis and narrative; keep the numbers themselves sourced from your actual systems.