A One-Page Business Case for AI Coding Tools

You have a budget request meeting on the calendar and ninety seconds of the CFO’s attention when it’s your turn. You can try to wing it — “engineering wants an AI coding tool, it’ll make us faster” — or you can walk in with a one-page business case that answers the four questions every CFO asks about any software expense.

This post is the one-page template. Copy the structure, fill in the numbers with your org’s inputs, and put it in front of the CFO before the meeting. The goal is not to sell the concept of AI coding tools — the market has already done that. The goal is to give your CFO the specific cost, specific return, and specific risk framing they need to approve a specific budget number without coming back with six follow-up questions.

The Four Questions

Every software-spend ask faces the same four questions. Answer them on one page. If you need more than one page, you don’t understand the decision well enough to ask for it.

1. What will it cost, total, for twelve months?
2. What is the return, in the same currency as the cost?
3. What is the risk if we say yes, and what is the risk if we say no?
4. Who owns the outcome, and what does success look like in ninety days?

The One-Page Template

Copy the block below. Fill the bracketed placeholders with your org’s numbers. Replace [N], [X], etc. with concrete figures. Everything in italics is guidance; strip the italics before you send.

---

AI Coding Tools — Budget Request

Requestor: [Name, Title] Date: [YYYY-MM-DD] Decision needed by: [YYYY-MM-DD] Amount requested: $[TOTAL] (12 months, [N] seats)

---

1. The cost (keep total of ownership, not sticker price)

Item	12-month cost
Tool licenses, [N] seats × $[per-seat-per-month] × 12	$[A]
Enterprise tier uplift (SSO, DPA, zero-retention)	$[B]
Projected usage overage (30% buffer above baseline)	$[C]
Admin + analytics tooling	$[D]
Security / legal review, one-time	$[E]
Training + enablement, one-time	$[F]
Total	$[TOTAL]

---

2. The return (in dollars, for the same 12 months)

Two calculation paths; pick the one your CFO trusts. Show both if asked.

Path A — Productivity uplift. Conservative assumption: [P]% productivity lift on applicable engineering time (applicable = code authoring and debugging, typically 40–50% of a full-stack engineer’s week). The remainder — meetings, design, review, ops — is unaffected by coding tools.

• [N] engineers × average fully-loaded cost $[E]/year × [applicable %] × [P]% uplift = $[RETURN_A]
• Against $[TOTAL]: ROI multiple = [RETURN_A / TOTAL]×

Path B — Incremental delivery. If the engineering team is capacity-constrained on shipping roadmap items:

• Current shipping rate: [items/quarter]. Projected uplift: [P]% → [additional items/year].
• At [average revenue impact per shipped item], incremental revenue = $[RETURN_B]
• Against $[TOTAL]: ROI multiple = [RETURN_B / TOTAL]×

CFO-defensible numbers: P between 8–15%. Anything above 20% is a red flag and should be supported with pilot data. See ROI calculation guide for the full derivation.

---

3. Risk framing (what happens if yes, and what happens if no — both)

Risks if approved:

• Security/compliance: mitigated by procurement checklist (executed DPA, SOC 2 Type 2, IP indemnification) and internal governance policy. Residual risk Medium, reducing to Low with controls in place. See risk assessment.
• Budget overrun: capped at [N]% overage by contractual clause and internal quotas. Predictable exposure.
• Productivity claim under-delivery: measured quarterly via baseline-adjusted shipping rate; if under-delivered by [X]%, seat count reduces at next renewal.

Risks if declined:

• Competitive: peer companies at comparable headcount are reporting [adoption rate] — falling behind on hiring signal if engineering job listings can’t mention AI tooling.
• Shadow IT: engineers will use free-tier personal accounts anyway, which fails every compliance framework in scope. [Survey / internal evidence if available] shows current shadow-use at approximately [Y]%.
• Recruiting: [senior engineer benchmark data] indicates AI tooling availability is now a factor in offer acceptance at the senior/staff level.

---

4. Ownership and 90-day success criteria

Owner: [VP Eng or designated director] Steering committee: [VP Eng, CISO, Finance partner]

90-day success criteria:

• [X]% of in-scope engineers onboarded, trained, and using the tool weekly.
• Baseline telemetry captured for productivity calculation at renewal.
• Zero Category-A security incidents; ≤[Y] Category-B exceptions reviewed.
• Usage within contracted seat and token budgets.

Tripwires (would trigger early re-evaluation):

• Security incident above a pre-defined severity threshold.
• Productivity metric stalls or regresses vs. baseline after 90 days.
• Vendor material contract or pricing change.

---

Appendix A — Tool selected and evaluation rationale: [one-line summary + link to bakeoff scorecard].

Appendix B — Full governance standard: [link to governance policy].

---

Filling In the Numbers

The placeholders, in the order they appear, with the numbers most commonly wrong:

[TOTAL] — All-in 12-month cost. CFOs hate when “the cost” turns out to be just the license fee. Include overage buffer, enterprise-tier uplift, and the one-time setup spend. Show the breakdown, not just the total.

$[per-seat-per-month]. Use the enterprise-tier quoted price, not the marketing-page price. Enterprise-tier is typically 2–3× public per-seat.

[applicable %] — Share of engineering time AI tools actually affect. Most requests overestimate this. For full-stack engineering, 40–50% of a week is hands-on coding. For senior engineers, closer to 25–30%. For platform/infra, ~35%. Be honest or the CFO will adjust downward more aggressively.

[P]% — Productivity uplift assumption. The single most contested number. Conservative defensible range: 8–15%. Public studies have claimed higher; internal measurements usually land in this range once you account for review overhead and tool switching. Use the conservative end for the business case and let the upside be a pleasant surprise in the renewal cycle.

[E] — Fully-loaded engineer cost. Salary + benefits + overhead, not base salary. For US-based engineers this is typically 1.3–1.5× base; for offshore teams adjust per your finance team’s standard.

[N]% usage overage cap. Negotiate this at procurement (see the contract checklist) and cite the cap here. Without it, the cost line is open-ended, which kills business cases on sight.

[Y] shadow-use rate. This is the hardest number to get because shadow use is by definition unmonitored. Proxies: internal survey asking engineers to self-report, review of network egress data for known AI-tool domains, or a one-time amnesty survey with no attribution. Even a ballpark number (“approximately 30% by self-report”) is more persuasive than no number.

The Meeting Itself

Four patterns for the 90 seconds you’ll actually get:

Lead with the total, not the per-seat price. CFOs think in annual all-in numbers. Per-seat pricing is an engineering abstraction.

Present the ROI path you believe, not both. Offering A and B suggests you’re uncertain which one is real. Pick one, justify it, have the other ready if asked.

Name the tripwires up front. CFOs approve more readily when they see you’ve thought about what would cause you to stop spending. Tripwires are credibility.

Don’t promise the productivity number will be visible in three months. It won’t. Baseline-adjusted uplift requires at least two quarters of data. Promise the measurement system in 90 days; promise the number in 180.

When the Answer Is No

Three recoveries when the ask doesn’t land:

Phase the deployment. “Start with a 10-seat pilot at 10% of the full ask, measure for one quarter, return with data” is a smaller ask that turns the business case into a more concrete one. The pilot program guide is the next read.

Move the cost line. Some CFOs approve $X of software more readily than $X of “AI productivity tools.” Framing the spend as part of the existing developer-tooling budget rather than as a new category can unlock approval without changing the underlying dollars.

Separate the compliance cost. If the enterprise-tier uplift (DPA, SSO, zero retention) is what’s pushing the total into rejection territory, separate it as a line item that the CISO/compliance budget funds, while engineering funds the base license. That reframes the conversation as risk spend, which has a different budget owner.

What This Template Replaces

The “trust us, it works” deck. Engineers pitching AI tooling to finance often rely on external productivity studies (“GitHub Copilot increased task completion by 55%”). These numbers are public, so the CFO has already seen them, and they are not drawn from your organization. They don’t close decisions.

The vibes presentation. “Our engineers love it, morale is up, we need to retain people.” Possibly true, not actionable for a CFO. Convert it into dollars via the retention path only if you have data (e.g., comp benchmarks showing AI-tool availability affects offer acceptance).

The roadmap accelerator pitch. “If we had this, we could ship X and Y this quarter instead of next.” Effective only if Path B above is your chosen path and you have credible throughput data to back it.

Handoffs

• From evaluation (scorecard) — Appendix A of the one-pager is the scorecard output in one sentence.
• To procurement (contract checklist) — once approved, procurement negotiates the clauses that make the cost and risk numbers defensible at renewal.
• To rollout (pilot program guide) — the 90-day success criteria feed directly into the pilot measurement plan.

For the full ROI calculation that underlies Path A and Path B, see AI coding ROI calculation. This post stops at “here’s the one-pager”; that post goes into how to derive the [P]% number defensibly from your own data.