How to explain your AI spend to your CFO

The conversation you need to have

Your CFO sees a line item that is growing 20% per quarter with no clear connection to revenue. That is a red flag in any budget review. Here is how to reframe it.

Frame 1: Unit economics

Stop talking about total spend. Start talking about cost per output.

• -"We spend $0.04 per code review. A human code review takes 30 minutes at $75/hour. The AI review costs 0.1% of the human alternative."
• -"We spend $0.002 per classification call. We process 50,000 support tickets per month. Manual classification would require 2 additional FTEs."

Unit economics turn a scary total number into a defensible per-unit cost that your CFO can compare against alternatives.

Frame 2: Attribution

Finance wants to know where the money goes. "We spend $15,000/month on AI" is a problem. "Engineering spends $8,000 on code generation, Support spends $4,000 on ticket routing, and Product spends $3,000 on content" is a budget.

Per-team attribution makes AI spend manageable. Each team owns their line item and can justify or optimize it independently.

Frame 3: Reduction trajectory

Show that costs are going down per unit even as usage goes up. This is the graph finance loves:

• -Q1: 100,000 calls, $12,000 total, $0.12/call
• -Q2: 150,000 calls, $13,500 total, $0.09/call
• -Q3 (projected): 200,000 calls, $14,000 total, $0.07/call

Usage up 100%. Total spend up 17%. Cost per call down 42%. That is an efficiency story, not a cost problem.

What to bring to the meeting

1. Current spend by team and model tier. Show that you know exactly where every dollar goes. 2. Cost per API call trend. Demonstrate that optimization is working and cost per unit is declining. 3. Projected savings from optimization. Quantify the impact of model routing, caching, and prompt optimization in dollar terms. 4. Comparison to alternatives. Show what these tasks would cost with human labor or without AI tooling at all.

The one-liner

If you need a single sentence for the exec summary: "We spend $X per month on AI that does the work of Y FTEs, and the cost per unit is declining Z% per quarter through optimization."

Build your executive cost summary at /teardown, or start tracking with the free dashboard.