Billing for AI work
Bill for the work, not just the hours.
AI cut your delivery time and your invoice followed it down. WhoWorked records what a deliverable actually took across humans and agents, so you can choose how to price the AI half and show a client the evidence behind the number.
- Works with time and materials contracts
- Model cost on every session
- Client-ready disclosure exports
One agent session, four prices
48m at $175
A set price per task
$6.21 of model spend, marked up
Tracked, invoiced at zero
You decide per client. The record supports whichever you pick
The question behind the invoice
Faster delivery turns into a smaller invoice unless you can show the work.
On a time and materials contract, an agent that saves nine hours removes nine billable hours. The work still happened, the deliverable still shipped, and the line item shrank.
What holds the price is a record of what the delivery took: human hours, agent sessions, model spend and the output that resulted.
On the invoice call
“Six hours on this section. Did anyone actually spend six hours on it?”
The honest answer is usually no, and saying so costs you the line unless you can show what the delivery took instead.
The conversation you are already having
“If the AI wrote it in twenty minutes, why is this line six hours at your senior rate?”
Answering from the delivery record moves the conversation onto what was produced. Answering from memory usually ends in a discount.
Where billing breaks
Three gaps between what you delivered and what you can invoice.
Hours no longer proxy for value
The clock was a reasonable stand-in for effort while people did all the work. It stopped being one the week an agent started producing part of the deliverable.
Model spend arrives detached
Token cost lands on a provider bill at month end with no client, project or deliverable attached, so it never reaches the invoice it belongs to.
No policy for the AI half
Bill it as hours, at a fixed rate, as a pass-through with margin, or not at all. Most firms have not decided, because nothing in their stack made it a decision.
What the record makes possible
Put the delivery cost and the invoice on the same deliverable.
Human cost, agent session cost and model spend sit against what you billed. Margin becomes a number on the deliverable rather than something finance assembles at the end of the quarter, and the disclosure a client asks for comes out of the same record.
- Choose a billing policy per client and apply it to agent sessions.
- Carry model and token cost through to the project it belongs to.
- Read gross margin per deliverable across human and AI cost.
- Export a client-ready disclosure the day procurement asks for one.
Portal migration, March
Invoice line
Billed
$2,010
Delivery cost
$742
Illustrative: 63% gross margin on one deliverable
Price one engagement from the record instead of the clock.
Bring a single project in, connect the agents running on it, and see the delivery cost against what you billed. 30 days, no card.
From session to invoice line.
01
The session carries its cost
An agent reports the model it used, the tokens it spent and what the run cost, against the project it worked on.
02
Your policy decides the price
Billed as hours, at a fixed rate, as cost plus margin, or held internally at zero. The same record supports whichever you pick, per client.
03
The client sees the evidence
Pro exports contributors, sessions, model spend and the assisted share behind a deliverable, in a form you can send without editing it first.
FAQ
Start counting all the work.
30-day free trial. Bring the time history you already have.