Monthly Reports
The standing executive report should fit on one page and answer four questions:
- Total spend (token costs, platform fees, infrastructure, fine-tuning, egress, professional services).
- Breakdown by team, business unit, and feature.
- Trend versus prior 3, 6, 12 months.
- Variance from budget at the agent and aggregate level.
Plus unit economics: cost per conversation, cost per resolved case, cost per qualified lead, cost per generated draft. Unit economics are the only numbers that survive when the volume changes.
Sample structure:
April 2026 AI Spend
────────────────────────────────
Total $487,400 (+8.2% MoM, +14.1% vs budget)
By function:
CX (Customer Agent) $214,000 $0.42/resolved (target $0.50)
Sales (Prospecting) $98,200 $1.18/qualified (target $1.00)
Marketing (Content) $67,500 $4.10/draft
Internal (Copilot) $107,700 $89/seat/mo
Top 3 cost movers vs March:
GPT-5 model upgrade +$42K
Holiday campaign +$18K
Knowledge refresh +$11K
Quarterly Deep Dives
Quarterly cadence supports investment decisions. ROI analysis per AI line item: revenue lifted, cost avoided, productivity recaptured. Attribution methodology stated explicitly so the next quarter compares like-for-like. Kill list: features whose unit economics or business outcomes underperform after a fair pilot window. Reinvestment plan: where the saved budget goes.
The FinOps Foundation’s AI cost framework (released October 2025) gives a reference taxonomy — use it to keep terminology stable across quarters.
Executive-Level Framing
Same underlying data, three audience-appropriate cuts:
- CFO: unit economics, total spend trajectory, multi-year forecast, vendor concentration risk, contractual exposure.
- COO: operational impact, throughput per agent, escalation rates, SLA performance, capacity planning.
- CMO: marketing ROI, content velocity, campaign attribution, audience-level performance.
- CIO: vendor stack, technical debt, build-vs-buy, security posture, AI Act conformity status.
Don’t hand a raw data dump to any of them. Distill to decision-ready summaries: three numbers, three findings, three recommendations.
Decisions the Reports Should Drive
A budget report that doesn’t change a decision is theater. Each cycle should produce concrete actions:
- Which AI features to scale (positive unit economics, capacity headroom).
- Which to cut (negative unit economics after a fair window, no path to improvement).
- Where to invest more (high ROI, capacity-constrained).
- Which vendors to consolidate (overlap, pricing arbitrage opportunity).
- What contract clauses to renegotiate (volume tier breakpoints crossed, model upgrades requiring new pricing).
Track decisions in the next report so the loop closes — “in March we decided X; here is the result in April.”
Common Failure Modes
- Reporting only token cost, missing platform fees and labor that dominate total cost.
- No unit economics; bills look big but no one knows whether they’re efficient.
- Reports go to executives who don’t have decision rights over AI spend.
- Cadence too slow (quarterly only) — costs move monthly and surprise quarterly.
What to Do This Week
Build the one-page monthly template. Review it with your CFO before next month-end. The first iteration is wrong; the third is useful.