How AI automation agencies typically bill
The build phase is usually fixed-price or milestone-based; the ongoing part is where agencies most often leave money on the table by not billing for it explicitly.
- Discovery & use-case mapping: a fixed fee, often the first paid engagement with a new client
- Agent/workflow build: milestone-based, tied to specific integrations or capabilities
- Monitoring & maintenance retainer: a recurring fee for keeping the system working as APIs and models change
Common invoice line items
Line items should separate the build from the integrations and from the ongoing maintenance, since these are genuinely different work with different billing logic.
- AI Discovery & Use-Case Mapping — $2,500
- Agent/Workflow Build — $6,000
- CRM/API Integration — $3,000
- Monthly Monitoring & Maintenance Retainer — $1,500/month
What counts as maintenance vs. a new project
The most common billing dispute in this space is what's covered by the maintenance retainer versus what's a new, separately quoted project — a client asking for "one more automation" often assumes it's covered. State explicitly in the agreement what maintenance includes (monitoring, minor prompt/config adjustments, API-change fixes) and what isn't (new workflows, new integrations).
Passing through usage costs
Many AI agencies pass through model or API usage costs (tokens, compute) to the client, separate from the management fee. Itemizing this separately on the invoice, rather than folding it into the retainer, keeps the retainer price stable even as usage costs fluctuate month to month.
Frequently asked questions
Should AI maintenance be billed separately from the build?
Yes. The initial build is typically a fixed-price or milestone project; ongoing monitoring and maintenance (as models and APIs change) should be a separate recurring retainer, stated explicitly so it isn't assumed to be free ongoing support.
How should usage/API costs be invoiced?
As a separate line item from the management fee, passed through at cost or with a stated markup, so the retainer price stays predictable even as usage varies.
What's the biggest billing mistake AI automation agencies make?
Not billing for ongoing maintenance at all — treating post-launch monitoring and fixes as free support, which is unsustainable once a client has more than a couple of live automations to maintain.