A hybrid team is people and AI agents delivering client work together: a developer directs Claude Code or Codex, reviews what it produces, and ships it. Hybrid teams change the economics of hourly billing, and they make value-based pricing more attractive. What they do not change is the need to track time. It just changes what the time is for.
The problem with hourly billing for a hybrid team
Hourly billing charges for effort. When an agent takes over part of the effort, the same result takes fewer billable hours, so the agency is paid less for delivering faster.
Billing the agent's hours as well recovers part of that, at a lower rate. It does not recover all of it, and it invites a question nobody wants on an invoice: why are we paying for the machine's time at all?
Value-based pricing avoids the question. The client pays for the outcome, and how the work was divided between people and agents is the agency's business. As hourly, fixed fee or value-based explains, it suits a narrow band of work: a measurable commercial outcome and a senior buyer. Hybrid delivery widens that band, because the gap between what the work is worth and what it costs to deliver is growing.
On a value-priced job, time measures cost, not price
When the invoice no longer comes from the hours, the hours become the cost side of the ledger. Without them a value-based price is a guess you can never check.
For a hybrid team there are three kinds of cost to keep apart:
- People's hours, at each person's cost per hour. This is still most of the cost of most jobs.
- Agent hours, at an agent cost rate. Kept apart from people's hours so nobody's utilization is inflated by their agents.
- Token cost, what the model provider charged, which runs alongside the agent hours.
Add them up against the fee and you have the margin. Divide the fee by the people's hours and you have the effective hourly rate, which is the number that tells you whether the price was right.
A worked example
An agency rebuilds a client's checkout flow for a value-based fee of $30,000, agreed against the conversion improvement the client expects.
Two developers deliver it over five weeks with coding agents. The agency tracks everything:
| Hours | Cost per hour | Cost | |
|---|---|---|---|
| People's time | 120 | $70 | $8,400 |
| Agent time | 40 | $12 | $480 |
| Delivery cost | $8,880 |
The margin is $30,000 − $8,880 = $21,120, or 70.4%. The effective hourly rate on people's time is $30,000 ÷ 120 = $250.
Now price the same job hourly, with people at $150 an hour and agent time at $60:
| Hours | Rate | Billed | |
|---|---|---|---|
| People's time | 120 | $150 | $18,000 |
| Agent time | 40 | $60 | $2,400 |
| Hourly invoice | $20,400 |
The work and the result are identical. The value-based price earns $9,600 more, and the faster the team gets, the wider that gap becomes. That is the case for value-based pricing on hybrid teams. The tracked hours are what let you make the case with numbers, and what tell you, on the next job, whether the price still holds.
The agent's estimate is evidence for the next proposal
Value-based pricing is easier to defend when you know what the work would have taken without agents. Some agents can say. In Time Trakkr, an agent can record its own estimate of how long a piece of work would have taken a person by hand, and the Agent work report adds those estimates up beside the time the agent actually took.
Suppose the sessions on the checkout job that gave an estimate come to 140 hours by hand against the agent's 40. That is a useful signal about how much of the job's value came from the way it was delivered.
Treat it as a signal, not a fact. Nothing checks the agent's guess, sessions that gave no estimate are left out, and it should never appear in front of a client as a claim of hours saved.
Track it without billing it
On a value-priced engagement, agent time should be recorded and costed, but never billed. In Time Trakkr an admin sets an agent rate rule for that client or project with Track but never bill ticked. Agent time under that rule is never billable, still carries its cost rate, and still counts toward the project's budget and profitability.
The same approach works for a mixed agreement, which is common in practice: a value-based fee for the core outcome, plus hourly billing for change requests. Map the change-request work to its own project with an ordinary agent rate, and its agent time lands on its own invoice lines, labeled as agent time.
What tracking will not tell you
Tracking tells you what the work cost. It does not tell you what the work was worth: that comes from the client's numbers and the conversation with the buyer, and no timesheet can supply it. Token cost in Time Trakkr is an estimate at list price, never what the provider actually charged. And an agent hour measures time spent, not value delivered: an hour that went down a dead end costs the same as one that shipped.
Common questions
Should we stop tracking time when we move to value-based pricing?
No. The time stops determining the invoice and starts determining whether the price was right. The agencies that stop tracking are the ones that find out years later that a service line was losing money.
Should agent hours count toward a developer's utilization?
No. A developer who runs agents in parallel would show more hours than the day has. Keep agent time apart from people's time, and count it toward the project instead. What is AI agent time tracking? explains the split.
How do we tell a value-based client about agents?
Tell them how the work will be delivered, as you would any method. The fee is for the outcome, so agent hours do not belong on the invoice, but a client who discovers agents were used and was never told will reasonably ask what else they were not told.
Can we compare value-priced jobs with hourly ones?
Yes, through effective hourly rate and margin on people's hours, which are the same calculation whatever the pricing model. Realization rate vs utilization covers where billable work loses its value on hourly jobs.


