Turn a Forrester-Style AI Business Case into a Revenue-Tied Memo Your CFO Won't Reject

The task

You're a founder or operator pitching an internal AI initiative — a copilot, an agent, a workflow bot — to a CFO or board who've been burned by pilots that never touched revenue. You need a two-page memo that names the problem, ties it to a P&L line, and pre-empts the finance objections before they arrive.

Before AI

The manual version is a half-day of context-switching: skim two or three business-case templates, cobble together assumptions in a spreadsheet, guess at payback, and then rewrite the whole thing after your CFO points out you never quantified the baseline. Most founders skip the memo entirely and pitch on vibes — which is exactly why while most organisations are deploying predictive, generative, and agentic AI, few can directly connect those investments to revenue, customer outcomes, or profitability, per Forrester's analysis of why AI products fail before they're built.

The bar the CFO applies is unforgiving. Guidance from finance leaders is that a defensible case must be readable in 10 minutes. Defensible under questioning, and it should open with a problem statement with current-state baseline: Cost, cycle time, error rate — from your pre-deployment data. Most first drafts don't clear that bar.

The workflow

1. Pressure-test the problem statement first

Before you compute a single dollar, force the initiative through a "would Forrester kill this?" screen. If the problem isn't revenue-adjacent, the memo is dead on arrival — and it's better to find out in 30 seconds than after your CFO reads page one.

Paste your raw initiative description into this prompt.

You are a skeptical Forrester analyst reviewing an internal AI initiative pitch. Your job is to decide whether this initiative has a legitimate revenue, retention, or cost-out linkage — or whether it's a "technology in search of a problem."

Read the initiative description that follows. Then produce:

1. **One-sentence problem restatement** in the operator's own domain language (no AI jargon).
2. **Revenue linkage verdict**: DIRECT (touches a booked revenue line), INDIRECT (affects retention, conversion, or CAC), COST-OUT (removes headcount, tooling, or COGS), or NONE (research/curiosity project).
3. **Baseline gap check**: list every current-state metric the pitch is missing (cost per unit, cycle time, error rate, volume, etc.). Be specific about the number you'd need.
4. **Three killer questions** a CFO will ask in the first five minutes. Phrase them the way a CFO actually talks.
5. **Go / Rework / Kill** recommendation with a one-line reason.

Be blunt. If the linkage is NONE, say so and stop. Do not invent numbers that aren't in the input.

Initiative description:

Initiative: "LeadWhisper" — an internal AI agent that listens to our AE discovery calls (via Gong recordings), auto-drafts the follow-up email within 10 minutes of call end, and pushes a suggested next-step task into HubSpot. Rolled out to our 12 AEs at Northwind Analytics (Series B, $18M ARR, avg deal size $42K, sales cycle ~68 days).

Why we think it matters: AEs currently take 1-2 business days to send follow-ups. Anecdotally we lose deals when follow-up slips past 24 hours. Our RevOps lead thinks faster follow-up = higher stage-2 conversion.

Build cost estimate: ~$25K to wire up (contractor, 3 weeks) + ~$800/mo in LLM + Gong API costs. No new headcount.

We have not measured current follow-up time precisely. We do have HubSpot data on stage-2 conversion by rep for the last 4 quarters.

Timeline: want a decision by end of month for Q1 build.

2. Build the revenue math — with assumptions you can defend

If step 1 came back Go or Rework, force the model to construct the financial case using only the numbers in the input, and to flag every assumption it invents. The reason so many AI pitches collapse is the modelling: revenue impact (retention, new capabilities) takes 3–5 years. The first measurable results should appear within the current fiscal year to maintain CFO confidence. Your memo has to reflect that timing honestly.

Using the initiative description and the verdict from the previous step, build the financial case. Output a table with these columns:

| Line item | Value | Source | Confidence (H/M/L) |

Include, at minimum:
- Current-state baseline (cost, time, or conversion metric — whichever the initiative claims to move)
- Target-state assumption (with the % lift or reduction stated explicitly)
- Annual revenue or cost impact (show the arithmetic in one line beneath the table)
- Year-1 build + run cost
- Payback period in months
- Break-even sensitivity: what's the smallest lift that still clears payback within 12 months?

Rules:
- Every "Source" cell must say either "from input" or "ASSUMED — [one-line justification]". Do not fabricate industry benchmarks.
- If a required input is missing, put "NEEDS INPUT" in the value cell and add it to a bulleted "Data to collect before finalising" list beneath the table.
- Keep the whole output under 400 words.

3. Assemble the CFO memo

Now compose the memo itself in the shape a finance reader expects. This is where you convert the analysis into a scannable page — the CFO reads the first paragraph and the risk table, and skims the rest.

Write the final one-page CFO memo using everything above. Use this exact structure and these exact headings:

**TL;DR** (3 bullets max: what we're building, revenue linkage, ask in dollars and months)

**Problem & baseline** (2-3 sentences, quantified)

**Proposed solution** (2-3 sentences, plain English, no model names or vendor pitches)

**Financial case** (paste the table from the previous step; add a one-sentence "what has to be true" summary underneath)

**Risks & how we'll know it's failing** (a small table: Risk | Leading indicator | Kill-switch trigger. Include at least: assumption risk, adoption risk, and data-quality risk.)

**Ask** (dollars, headcount, decision date, and the single next milestone)

Tone: direct, operator-to-operator. No hedging language ("could potentially", "may help to"). No AI-industry buzzwords. If a number is assumed rather than measured, mark it with an asterisk and footnote it at the bottom.

Length ceiling: 500 words including the tables.

Gotchas

• Garbage baseline in, garbage memo out. If step 2 comes back with a wall of "NEEDS INPUT" rows, stop and go collect the numbers before you show anyone the memo. A CFO who spots one made-up baseline will assume the rest are made up too.
• Indirect-revenue linkages get gamed. The model will happily label anything a "retention lever." Pressure-test by asking: can you point to a specific cohort, contract, or renewal date this changes? If not, downgrade the verdict.
• Payback math flatters short cycles. For deal-cycle initiatives longer than ~6 months, a 12-month payback claim is almost always wrong because you won't see the closed-won impact in-year. Add a footnote acknowledging the lag.
• The model over-writes the "Risks" table. It tends to list generic risks (model hallucination, data privacy). Force it to name risks specific to the initiative in the input — reject any risk row that could copy-paste onto a different project.
• Don't skip step 1. If you jump straight to the memo, the model will manufacture a revenue story for a project that doesn't have one. That's the exact failure mode Forrester is describing.

Time saved

A first-draft AI business case memo typically eats 3-4 hours: an hour of template hunting, an hour of spreadsheet work, an hour of writing, and a round of edits after your CFO or head of finance pokes holes. This workflow compresses the first three into roughly 15 minutes of prompting plus 20-30 minutes of you supplying real baseline numbers, and — more importantly — it front-loads the "is this even a real problem" screen so you don't spend hours polishing a memo that should never have been written. Estimate, not measurement; your mileage depends on how clean your baseline data already is.

What we got when we ran it

We executed every prompt above, in order, against the sample input (model: claude-sonnet-4-6). Outputs below are real and unedited apart from length truncation.

Step 1 output

Forrester Analyst Review: "LeadWhisper" --- ### 1. One-Sentence Problem Restatement Northwind's account executives are losing winnable deals because slow post-call follow-up — averaging one to two business days — lets buying momentum decay before prospects reach a second qualified conversation. --- ### 2. Revenue Linkage Verdict **INDIRECT** This does not touch a booked revenue line directly. The mechanism is: faster follow-up → higher stage-2 conversion rate → shorter sales cycle → more deals closed per AE per quarter. That is a conversion and velocity play, which is legitimate but requires the conversion lift to be measurable and material enough to justify the build. --- ### 3. Baseline Gap Check The pitch is missing every number that would make the business case defensible. Specifically: | Missing Metric | Why You Need It | Where to Get It | |---|---|---| | **Current median follow-up time, in hours** | You cannot claim a 10-minute improvement without knowing the baseline. "1-2 business days" is anecdote, not data. | Gong call-end timestamps vs. HubSpot first-email timestamps — this is a pull, not a survey. | | **Stage-2 conversion rate segmented by follow-up speed** | The RevOps hypothesis ("faster = higher conversion") may already be testable in your own HubSpot data. If it isn't true historically, the entire thesis collapses. | 4 quarters of HubSpot data you already said you have. Run the correlation *now*, before committing $25K. | | **Current stage-2 conversion rate (overall and by rep)** | You need a baseline to measure

…output truncated at 250 words.

Step 2 output

LeadWhisper: Financial Case (Draft) | Line Item | Value | Source | Confidence | |---|---|---|---| | Current stage-2 conversion rate | NEEDS INPUT | NEEDS INPUT | L | | Current median follow-up time | NEEDS INPUT | NEEDS INPUT | L | | Discovery calls per AE per month | NEEDS INPUT | NEEDS INPUT | L | | Stage-2 to close rate | NEEDS INPUT | NEEDS INPUT | L | | Average deal value at stage-2 | $42K | From input | M | | Number of AEs in scope | 12 | From input | H | | Target follow-up time | <10 min post-call | From input | H | | Assumed stage-2 conversion lift | 5 percentage points | ASSUMED — midpoint of plausible range; unvalidated; replace with HubSpot correlation output | L | | Incremental deals reaching stage-2 annually | NEEDS INPUT | Depends on call volume × lift | L | | Incremental closed revenue (annual) | NEEDS INPUT | Requires close rate × ACV × lift deals | L | | Year-1 build cost | $25,000 | From input | H | | Year-1 run cost (LLM + Gong API) | $9,600 | From input ($800/mo × 12) | H | | Year-1 total cost | $34,600 | From input | H | | Payback period | NEEDS INPUT | Cannot calculate without revenue impact | L | | Break-even lift (12-month target) | NEEDS INPUT | See sensitivity note

…output truncated at 250 words.

Step 3 output

# LeadWhisper — CFO Decision Memo --- TL;DR - We are building a tool that auto-drafts AE follow-up emails within 10 minutes of a discovery call ending and pushes the next-step task into HubSpot. - Revenue linkage: stage-2 conversion rate. Faster follow-up closes the window where deals go cold. - Ask: $34,600 in Year-1 spend; decision needed by end of month. --- Problem & Baseline AEs take 1–2 business days to send post-discovery follow-ups. HubSpot data shows stage-2 conversion varies meaningfully by rep, but we have not yet pulled the correlation between follow-up speed and conversion — that pull happens this week. At $42K ASD and 12 AEs, every percentage-point of stage-2 conversion we recover is material. --- Proposed Solution Immediately after each discovery call ends, the tool reads the call recording, drafts a follow-up email for AE review, and creates the next-step task in HubSpot — all within 10 minutes. The AE reviews and sends; nothing goes to a prospect without human sign-off. Build time is three weeks using an external contractor; no new headcount. --- Financial Case | Line Item | Value | Source | Confidence | |---|---|---|---| | Current stage-2 conversion rate | NEEDS INPUT | HubSpot pull due this week | L | | Current median follow-up time | NEEDS INPUT | Gong + HubSpot timestamps | L | | Discovery calls / AE / month | NEEDS INPUT | Gong data | L | | Stage-2 to close rate | NEEDS INPUT | HubSpot pull

…output truncated at 250 words.