Four parts: the problem (quantified in dollars), the solution (specific and scoped), the ROI (conservative and time-bound), and risk mitigation (pilot phase). Skip hype. Lead with numbers.
CFOs don't get excited about technology. They get excited about returns. Walk in with "digital transformation" and walk out empty-handed. Walk in with a clear problem, scoped solution, and conservative ROI — walk out with a green light.
Step 1: Quantify the Problem
Make the problem expensive before you mention AI.
Example: Support team spends 3 hours/day on repetitive questions. 3 hrs/day x 5 people = 15 hrs/day. 15 hrs x $35/hr = $525/day. $525 x 260 days = $136,500/year.
Include hidden costs: opportunity cost, error rates, customer impact from slow response.
Step 2: Propose a Specific Solution
Bad: "Implement AI across customer support." Good: "Deploy an AI email responder for Tier 1 tickets on Zendesk, handling 12 question categories representing 65% of volume."
Include: what AI does (and won't), systems involved, who's affected, timeline, total cost.
Step 3: Conservative ROI
Estimate 50% workload handled (conservative). $68,250 in savings. Minus $15,000 implementation + $3,600/year subscriptions.
Year 1: $49,650 net. Payback: Under 3 months.
Add sensitivity analysis: what if results are 25% worse? If numbers still work, finance loves the margin of safety.
Step 4: Risk Mitigation via Pilot
"One team, 30 days. Success: 40% reduction in Tier 1 handling time, 90%+ accuracy. If we hit it, we scale. If not, total exposure: $5,000."
$5,000 bet on a $65K/year saving = easy yes.
One-Page Template
Problem: [Workflow] costs [$X]/year. Solution: Deploy [AI tool] for [function], integrated with [systems]. Investment: [$X] implementation + [$X]/month. Return: [$X] annual savings. [$X] payback period. Risk: 30-day pilot, [$X] total exposure, [criteria]. Ask: Approve [$X] for pilot by [date].
FAQ
CFO skeptical about AI?
Don't sell AI. Sell the business outcome that uses AI.
Hard-to-quantify benefits?
Lead with time saved — easiest to measure, hardest to argue.
Include competitive pressure?
Yes, briefly. "Top 3 competitors deployed similar automation" — one powerful line.
Pilot succeeds but scaling is expensive?
Include scaling costs upfront. Show full picture.
Pilot length?
30 days for simple, 60 for complex.
Who presents?
Person closest to the problem. Technical details from IT/AI partner. Operations leads, finance approves.
Pilot metrics?
Time saved, accuracy rate, adoption rate, one business outcome metric.