Bill for the result or for the effort, now that agents do part of the work
Fixed price vs time and materials is the first decision in every software project: do you sell a result at an agreed price, or the work it takes at an hourly rate? With AI agents in the team, the old rules of thumb shift. Parts of a feature are done in minutes, others take as long as ever, and hours say less and less about what the work is worth. This guide shows who carries which risk in each model, how to calculate a fixed price when agents do part of the work, and how time and materials stays fair when your hours shrink.
Who carries which risk
| Criterion | Fixed price | Time and materials |
|---|---|---|
| Overruns are paid by the client | – | |
| Faster agents raise your earnings | – | |
| Client knows the price upfront | – | |
| Scope can change without renegotiating | –Change request | |
| Fluctuating AI costs are your risk | partlyUnless passed on | |
| Effort must be backed up item by item | – |
Hours no longer measure value
An hourly rate assumes that more hours mean more result. With agents that no longer holds. While you review one change, two agents write tests and migrate an interface. One hour of your time can carry three strands of work, or a single prompt that solved nothing.
So is the billable hour dead? As a measure of value, largely yes. As a measure of effort, no: your hours, agent runtime and AI costs are still what a project costs you. Price by the result where you can describe it, and measure the effort anyway. That is the only way to know whether your price was right.
How to price software development with AI agents
Split the scope
Into parts you can describe and accept one by one. Small parts are easier to estimate and can be invoiced as they are done.
Estimate your hours, not the agents'
Agents shorten the implementation. Specification, review, testing, coordination with the client and deployment stay with you and often make up most of the hours.
Add AI costs and a buffer
AI costs fluctuate with the model and the length of the context. A buffer covers both that and the estimate being wrong.
Check the price against the value
Effort gives you the lower limit, the client's gain the upper one. If agents make you faster, the price does not have to follow your hours down.
Cost it afterwards
Compare the price with the hours actually worked and the AI costs. The effective hourly rate shows whether the price held, and the next quote fits better.
Calculate a fixed price, then check it
Result
Starting values are assumptions. Fixed price = (estimated hours × rate + AI costs) × (1 + buffer). The effective hourly rate after AI costs is the fixed price minus AI costs, per hour actually worked.
Keeping time and materials fair when hours shrink
Time and materials fits where nobody knows the scope yet: prototypes, discovery, legacy code nobody understands. The client pays what the work took. The catch with agents: the faster you get, the less the same result brings in.
Three ways to keep it fair. Set the rate by the result you deliver per hour, not by the hours you used to need. Bill AI work as its own item instead of hiding it in your hours, at cost, with a markup or by usage (the four ways are compared in the guide to AI agency pricing). And show the client what went into the work: working time, AI sessions and the commits that came out of them.
Hybrid models
Most projects do not need a pure model. These combinations shift the risk in steps.
- Discovery on time and materials, then a fixed price: the open questions are cleared by the hour, the build has a price.
- Fixed price in parts: a share of the price per accepted milestone, with a deposit at the start. Lowers your risk without changing the model.
- Time and materials with a cap: billed by effort up to an agreed maximum. The client knows the worst case, you are not paid below your rate.
- Fixed price per part plus time and materials for changes: anything outside the described scope runs by the hour.
What the contract decides
A fixed price usually goes with a promised result: the client accepts it, and only then is the full price due. Time and materials usually goes with a promise of work, not of a result. What counts is what you promise, not how you are paid: a promised result can also be billed by the hour.
Under German law, for example, this is the difference between a contract for work (§ 631 BGB, payment due on acceptance under § 641 BGB) and a service contract (§ 611 BGB). Other jurisdictions draw the line differently. This is not legal advice; as of 10 Oct 2026.
How Time Momentum handles both models
Billing type per project
Time and materials values the working time at an hourly rate. On a fixed price, the agreed price counts and the hours are effort, not a claim.
Value of work by progress
With estimated hours on the order, a fixed price is earned in step with the hours worked, never above the price. You see during the project whether it holds.
Effective hourly rate
Fixed price divided by the hours worked, on the project page next to revenue, costs and margin.
Billing and costingFixed price in parts
Invoice a percentage, an amount or the rest of the fixed price, and deduct deposits already invoiced.
AI work on the invoice
On time and materials, AI work goes on the invoice at cost, with a markup, by token volume or by active runtime.
Costs per project
AI costs, agent runtime, working time and expenses against revenue, as the basis for your next estimate.
AI agency pricing models
Fixed price, retainer, time and materials plus AI costs, or value-based, compared for AI agencies.
Read guideEffective hourly rate calculator
Enter your fee, hours and AI costs, and compare the result with your target rate.
Open toolAgentic engineering vs. vibe coding
Why effort per project matters more, not less, when agents write the code.
Read guide
Frequently asked questions
Fixed price vs time and materials: which is better?
Fixed price when the result can be described and you have built something similar before; time and materials when the scope is open. With AI agents, a fixed price lets you keep the speed, time and materials passes it on to the client.
How do I calculate a fixed price when AI agents do part of the work?
Estimate your own hours for specification, review, testing and coordination, add expected AI costs and a buffer, and check the result against the client's gain. After the project, compare it with the actual effort.
Is the billable hour dead?
As a measure of value, largely. As a measure of effort, no: hours, agent runtime and AI costs are still what a project costs you, and you need them to cost a fixed price.
What is the effective hourly rate?
The fixed price divided by the hours actually worked. It shows what an hour of your work earned on a fixed-price project.
Does Time Momentum support both models?
Yes. Each project is billed on time and materials, at a fixed price or pro bono. Orders, invoices, profitability and client reports are part of Pro; every account starts with 30 days of Pro.
See what your fixed prices were really worth
- 30 days of Pro
- No credit card
- Then continue on Free