Why the report matters more than the dashboard
Your client won’t log into a dashboard. They will read a one-page document that arrives with the invoice. That page is where they decide whether the AI you built is a cost to cut or an asset to expand — so it should answer their questions, in their order, in plain money.
The seven numbers
- AI spend this month. The real figure, priced from the provider’s own token counts — not an estimate.
- What it would have cost without your optimizations. The baseline. Without it, savings are just a claim.
- What was saved, and how. The difference, split by mechanism — answers served from cache, calls moved to a cheaper model — so the client sees the work, not just the total.
- Volume. Requests this month. It explains spend changes before anyone asks: “spend up 20% on 35% more requests” is good news.
- Breakdown by feature or project. Which part of their product the money goes to.
- Model mix. Spend by model. It shows that premium models are used where they earn it.
- Next month. One or two concrete recommendations with an expected effect, and anything that needs their decision.
Presentation that works
- Lead with the saving and the percentage, then the spend — in that order.
- One page. Tables for breakdowns, sentences for meaning.
- Your agency’s name and logo, not your tools’ — the client is buying your judgement.
- Keep your own costs and fees off it if you rebill with a markup; the report is about their AI, not your supplier invoices.
- Same format every month, so the trend reads at a glance.
Common mistakes
- Projected savings presented as real. “Could save” is a recommendation; only savings measured on actual requests belong in the headline.
- Token counts without money. Clients think in dollars; tokens belong in an appendix.
- One lump for all clients. If you run several clients on one provider account, a report built from the provider invoice can’t be split fairly. Meter per client at the request level. See passing LLM costs to clients.
Producing it without a spreadsheet
ModelGate generates this report per client from the gateway’s own records: AI spend, what the traffic would have cost without it, savings and the share from cache, breakdowns by project and by model, for any month. It prints to PDF or downloads as CSV, carries your agency’s name, logo and colour, and leaves ModelGate’s fee off. You can also give the client a read-only login to see it themselves. The recommendation for next month is still yours to write — that’s the part they pay you for. ModelGate for agencies.
Keep reading
- How to price AI work for clients when model costs move every month — Retainer, cost-plus or a share of savings — and the token math that decides your margin.
- Passing LLM API costs through to clients: three models that work — The client's own provider account, rebilling with a markup, or a bundled fee — and what each needs.
- ModelGate for agencies — per-client ledgers, caps, reports and billing.