draftkit

Posted on Jul 23

10 AI Prompts I Use for FP&A (Budgets, Forecasts, Variance Analysis, Board Decks)

#ai #productivity #finance #prompts

I spent three years in FP&A before AI was useful. The job hasn't changed — you still own the budget cycle, the monthly close review, the quarterly board deck, and the annual operating plan. What changed is the writing around the numbers. Every variance needs a narrative. Every forecast needs assumptions documented. Every board meeting needs a summary a non-finance executive can read in 90 seconds.

These 10 prompts handle that writing layer. I've tested each one at least three times against real financial scenarios. They don't replace your model — they replace the hour you spend explaining what the model already told you.

Testing method: I run each prompt 3 times with the same inputs. If the output varies wildly, the prompt is too vague. I generalize it by removing specifics and re-testing on a different business. I read it aloud — if it sounds like a template, I cut the generic parts.

1. The Budget Memo (Forces Prioritization, Not Just Numbers)

Role: You are an FP&A manager writing a budget memo for the leadership team.
Context: I'm giving you next year's department budget request with the prior-year actuals and the variance justification. The audience is the CFO and department heads who will push back on line items.
Constraints:

Start with the ONE number that matters most (total, delta vs. prior year, or % of revenue — your call based on what's most defensible).
Group line items into max 4 categories. Do not list individual line items under $5K.
For each category, state the driver in one sentence (headcount, software, contractor, travel).
Include a "What we'd cut first" section — 3 items ranked by impact-to-cost ratio. This pre-empts the inevitable CFO question.
Do NOT use the words "strategic investment," "capability building," or "operational excellence."
Max 400 words.

Example input:

Prior year actuals: Engineering $1.2M, Sales $800K, G&A $400K
Next year request: Engineering $1.5M (+25%), Sales $950K (+19%), G&A $420K (+5%)
Key drivers: Eng hiring 3 senior engineers, Sales adding 2 AEs + SDR tools, G&A is mostly rent + accounting software.

Why it works: The "what we'd cut first" section is the unlock. CFOs ask "where can we trim?" in every budget meeting. Pre-answering it shifts the conversation from whether to cut to which cut is acceptable.

2. The Rolling Forecast Update (Assumptions Before Numbers)

Role: You are an FP&A analyst updating the rolling 12-month forecast.
Context: I'm providing the current forecast, the actuals through this month, and the changes to key assumptions (pipeline coverage, churn signals, hiring delays).
Constraints:

Section 1: Assumptions changed this month — list each as "Assumption → Old value → New value → Reason." No numbers in the forecast yet.
Section 2: Forecast impact — for each changed assumption, state the directional impact (+/-) and the magnitude range (e.g., "+$40K to +$60K to Q4 revenue"). Ranges, not point estimates.
Section 3: Confidence level — High / Medium / Low for each change, with a one-line basis.
Do NOT produce a revised P&L. That's the model's job, not the prompt's.
Flag any assumption that contradicts another (e.g., raising the close rate while lowering average deal size — are we sure both move in opposite directions?).

Why it works: Forecasting discussions derail when people argue about the output number instead of the input assumption. This prompt forces the conversation to assumptions first, numbers second.

3. Variance Explanation (The Monthly Close Narrative)

Role: You are writing the variance analysis section of the monthly financial review.
Context: I'm providing actual vs. budget for each line item, with the variance amount and whether it's favorable or unfavorable.
Constraints:

For each material variance (>5% of budget OR >$10K, whichever is lower):
- What: The line item, the variance amount, and direction.
- Why: The driver — ONE sentence. "Timing" is banned unless you specify what shifted.
- So what: The impact on the full-year forecast (will it reverse, persist, or worsen?).
Sort by absolute dollar impact, largest first.
Do NOT explain immaterial variances. Group them as "Below threshold — no action."
End with "Top 3 to watch next month" — the variances most likely to recur or compound.

Why it works: The "timing" ban is critical. "Timing" is the most overused variance explanation in FP&A — it means nothing without specifics. Forcing specificity either produces a real explanation or exposes that nobody knows the actual driver.

4. Board-Ready KPI Summary (90 Seconds, Not 9 Minutes)

Role: You are preparing the KPI summary slide for the upcoming board meeting.
Context: I'm providing 8-12 KPIs with this month's value, prior month, prior year, and target.
Constraints:

Output is a single slide — max 6 KPIs. You choose which 6 matter most for a board audience (growth, unit economics, runway, not operational minutiae).
For each KPI: the number, the delta vs. prior period (with direction arrow), and a 6-word max annotation.
The annotation must be a takeaway, not a description. "Revenue $4.2M ↑12%" is a description. "Revenue beat target, driven by enterprise expansion" is a takeaway.
Group into max 3 sections (e.g., Growth, Efficiency, Health).
Do NOT include any KPI the board has never seen before without a one-line definition. If you must include a new metric, define it inline.
End with "One number the board will ask about:" — predict the single metric that will generate the most discussion and pre-answer it.

Why it works: Board members read slides in 10 seconds. The annotation constraint forces you to pre-digest each KPI into a takeaway, which means the board meeting discusses the implication instead of spending 5 minutes just understanding what the numbers are.

5. Investor Update (The Monthly Email VCs Actually Read)

Role: You are writing the monthly investor update email.
Context: I'm providing the key metrics, wins, challenges, and asks for this month.
Constraints:

Subject line: "Month Year Update — [Company]" — no clever subject lines.
Section 1: TL;DR — 2 sentences max. What happened, what's next.
Section 2: Metrics — table format. Metric / This month / Last month / Change. Max 5 metrics.
Section 3: Highlights — max 3 bullet points, each starting with a verb.
Section 4: Lowlights — max 3 bullet points. This section must NOT be shorter than Highlights. If it is, you're hiding something.
Section 5: Asks — specific requests (intros, hiring, advice). Each ask is one line with the context embedded.
Do NOT use "continuing to," "making progress on," or "excited about." These are filler.
The email must be readable in under 90 seconds. If it's longer, cut.

Why it works: The Lowlights ≥ Highlights constraint is the most important one. Investors pattern-match for founders who only share good news — it signals either naivety or dishonesty. Equal-weighted lowlights build trust and actually generate more helpful responses.

6. The Scenario Analysis Memo (Best Case / Base Case / Worst Case)

Role: You are writing a scenario analysis memo for the leadership team.
Context: I'm providing the base case forecast, the key uncertainties (demand, pricing, hiring, churn), and the time horizon.
Constraints:

Define 3 scenarios: Upside, Base, Downside. Do NOT call them "optimistic / realistic / pessimistic" — those words carry emotional weight that biases the reader.
For each scenario:
- Trigger: What has to be true for this scenario to play out? (2-3 conditions)
- Impact: Revenue range, expense range, cash runway impact.
- Probability: Your estimate with a one-line basis. Must sum to ~100% across all three.
Do NOT produce full P&Ls for each scenario. Ranges only.
End with "Monitoring signals" — the 3 metrics that would tell you which scenario is materializing, and the threshold for each that would trigger a re-forecast.

Why it works: The trigger conditions are the unlock. Most scenario analyses just present three number sets with vague labels. Specifying what would have to be true makes the scenarios testable — you can check each month whether the upside or downside triggers are firing.

7. The Headcount Plan Narrative (For the Hiring Committee)

Role: You are writing the headcount plan justification for the hiring committee.
Context: I'm providing the requested hires (role, level, team, target start date), the business rationale, and the budget impact.
Constraints:

Section 1: Summary table — Role / Team / Q-hire / Annual cost / Priority (P0/P1/P2).
Section 2: For each P0 role (must-haves), write a 2-sentence justification: (a) what this person unblocks, (b) what happens if we delay by one quarter.
Section 3: Phasing — which roles can start in Q1 vs. Q2 vs. Q3. Group by dependency, not by team.
Section 4: Risk — the 2 roles where we have the highest risk of not filling (scarce talent, comp gap, etc.).
Do NOT justify headcount with "we need to scale" or "to support growth." State the specific bottleneck or output that's currently capacity-constrained.

Why it works: The "what happens if we delay by one quarter" constraint forces honest prioritization. If the answer is "nothing really breaks," it's not a P0. This prompt surfaces which hires are actually urgent vs. which are nice-to-have.

8. The Cash Runway Memo (The One That Keeps You Up at Night)

Role: You are writing the cash runway analysis for the CEO and board.
Context: I'm providing current cash balance, monthly burn rate, revenue forecast, and planned fundraising timeline.
Constraints:

Section 1: The number. "Current runway: X months at current burn (date zero: [month/year])." One sentence. No hedging.
Section 2: Burn trend — is burn accelerating, flat, or declining? Show the last 3 months and the forecasted next 3.
Section 3: Sensitivity — what happens to runway if revenue is 20% below forecast? What if burn is 15% above plan?
Section 4: Levers — 3 concrete cost reduction actions ranked by impact, each with the monthly savings estimate and the implementation timeline.
Section 5: Trigger — the specific cash balance or runway month at which we MUST start executing cost reductions. Not "when we get concerned" — a number.
Do NOT say "we are well-capitalized." If the runway is under 18 months, say so plainly.

Why it works: The trigger number is the unlock. Runway memos without a trigger are just status reports. A specific "at $2.5M cash balance, we begin hiring freeze" gives the CEO a decision point, not a anxiety-inducing data dump.

9. The Pricing Analysis Summary (When Sales Wants to Discount)

Role: You are writing the pricing/discount analysis for the pricing committee.
Context: I'm providing the standard price, the proposed discounted price, the deal size, the customer segment, and the competitive context.
Constraints:

Section 1: The deal economics — standard price, proposed price, discount %, annual contract value at both prices, and the lifetime value delta.
Section 2: Precedent — have we done this discount before? For which customers? What happened (did they renew at full price or did the discount stick)?
Section 3: Cannibalization risk — if other customers in this segment learn about this price, what's the exposure? Estimate the number of accounts that would demand the same discount.
Section 4: Recommendation — Approve / Reject / Approve with conditions. If "with conditions," specify them (contract length, payment terms, case-study requirement).
Do NOT recommend approval based on "strategic value" or "logo acquisition" without quantifying it.

Why it works: The cannibalization section is the most valuable. Discounts rarely stay isolated — they become the new benchmark. This prompt forces the pricing committee to think about the second-order effect, not just the single deal.

10. The Annual Operating Plan Kickoff Memo (Setting the Frame Before the Grind)

Role: You are writing the AOP kickoff memo to department heads.
Context: I'm providing the company-level targets (revenue, growth, margin), the timeline, the key assumptions, and the process for department submissions.
Constraints:

Section 1: The targets — revenue, gross margin, operating income, headcount. Each with a one-line "why this number" (market context, board expectation, prior-year trajectory).
Section 2: Key assumptions — the 5 assumptions that will most impact the plan (pricing, churn, hiring velocity, market growth, FX). State each with the current assumption and the range of uncertainty.
Section 3: Timeline — submission dates, review dates, lock date. Max 5 dates. No Gantt charts in a memo.
Section 4: What's different this year — the 2-3 things that changed from last year's process. If nothing changed, say "Same process as last year" and explain why stability is intentional.
Do NOT tell department heads to "be ambitious but realistic." That's meaningless. Tell them the specific tension to resolve (e.g., "We're targeting 30% growth but capping headcount at 15% — your plan needs to show how output per head improves").

Why it works: The "what's different this year" section prevents autopilot budgeting. If the process is identical to last year, people will copy-paste last year's submission. Forcing an explicit statement of either change or intentional stability resets the framing.

How These Prompts Fit Together

These 10 prompts cover the full FP&A communication cycle:

Prompt	When	Audience
Budget memo	Annual	Leadership team
Rolling forecast	Monthly	FP&A + CFO
Variance analysis	Monthly close	Department heads
KPI summary	Board meetings	Board
Investor update	Monthly	Investors
Scenario analysis	Quarterly	Leadership team
Headcount plan	Hiring cycles	Hiring committee
Cash runway	Monthly	CEO + board
Pricing analysis	Deal reviews	Pricing committee
AOP kickoff	Annual	Department heads

The pattern across all of them: force the takeaway before the data. Every FP&A document has the same failure mode — a wall of numbers with no interpretation. These prompts front-load the "so what" and push the numbers to a supporting role.

The Testing Method (So You Can Adapt These)

Run 3 times with the same input. If outputs diverge significantly, the prompt has too much degrees of freedom — tighten the constraints.
Generalize by testing on a different business model. A prompt that works for SaaS but breaks for e-commerce is too specific.
Read aloud. If it sounds like a Mad Libs template, cut the generic scaffolding. The prompt should produce output that sounds like you wrote it, not like a form was filled in.
Cut anything generic. "Provide a comprehensive analysis" is not a constraint — it's filler. Every constraint should be something the output would get wrong without it.

If you found these useful, I've packaged 50+ prompts for startup operations — finance, hiring, sales ops, and growth — into a complete system:

Developer Product-Launch Prompt Pack — 7 prompts for shipping products faster ($9)
Developer Productivity Prompt Library — 30 prompts for engineering workflows ($49)
SaaS Marketing Copy Prompt Pack — 25 prompts for marketing that converts ($49)
AI Startup Operations Prompt System — 50+ prompts + 10 workflows for running a startup with AI ($199)
AI Business Transformation Playbook — 100+ prompts + 12 workflows, the complete system ($999)