You pay for meeting time and AI time
That is the whole bill. The time we spend planning, writing instructions, reviewing the result, and fixing anything by hand is on us.
An example day
Here is how a working day on your project splits up, and the part of it you pay for.
- Meeting with you30 minBilled
- Writing instructions for the AI1 hNot billed
- AI building your solution50 minBilled
- AI testing and fixing, round after round40 minBilled
- Our review, and anything we fix by hand1 hNot billed
What you pay for
Meeting time
Calls, workshops, demos, and onsite sessions with your team. We bill the length of the meeting.
AI time
The minutes Claude Code, Grok Build, or OpenAI Codex spends building, testing, and fixing your solution. Most work takes several rounds of test and fix, and the AI runs those too. A clock starts when we hand the AI a task and stops when it finishes.
What you do not pay for
- Planning the work and writing the instructions we give the AI.
- Our review of what the AI built.
- Anything we fix by hand.
- Research, learning a new feature, and waiting.
Why we bill this way
AI does most of the building now. Billing the hours a person spends at a keyboard no longer matches the work. Some days a consultant spends an hour getting the instructions right and the AI builds it in ten minutes. Other days it runs the other way.
AI time follows the size of the job, not how fast one person types. You never pay for our learning curve.
We use AI for as much of the work as we can, testing and fixing included. We give our tools the guardrails we learned the hard way: tests, security, accessibility, and how each platform breaks. With those in place, the AI does a better job than we could by hand.
How you check it
We measure every entry. We never estimate it. AI time comes from the tool's own clock. Meeting time is the length of the call. Your client portal shows the hours against each task, day by day.
The tools we use
Claude Code
Anthropic's coding agent, and our default. It writes and changes code, reads how your system is set up, and runs the tests.
Grok Build
xAI's coding agent. It can run several agents at once and rank what they produce, which helps when we want to compare approaches.
OpenAI Codex
OpenAI's coding agent. We use it for a second approach to hard problems and to review work.
We pick the tool for the task. The rate is the same whichever one runs.
Data that cannot go to a third-party AI
These tools send the code and data they work on to their vendor's AI models. If that is not allowed for your business, Private AI Delivery runs the work on models in our own office.
What it costs
One rate card for Salesforce, Odoo, and Mac and iOS work. Buy hours in advance. The more you buy, the lower the rate.
| Hours | Rate | Price | New customer |
|---|---|---|---|
| 4 | $65 / hour | $260 | $221 |
| 25 | $62 / hour | $1,550 | $1,320 |
| 50 | $58 / hour | $2,900 | $2,465 |
| 100 | $55 / hour | $5,500 | $4,675 |
| 200 | $50 / hour | $10,000 | $8,500 |
Hours do not expire while your engagement is open. Two things sit outside the two-line bill: onsite travel time and expenses, and the one-time Private AI Delivery setup if you choose it.
Questions
Do you bill for the time you spend writing instructions for the AI?
No. Planning, writing instructions, and reviewing the result are on us.
What if the AI gets something wrong?
It usually catches the problem itself. The AI tests its own work and fixes what fails, round after round, until the tests pass. Those rounds are AI time. Our review, and anything we fix by hand, is not.
Does this work for a small budget?
Yes. A 4-hour pack is $260, or $221 if you are new to us. Four hours of meetings and AI time goes a lot further than four hours of a person typing.
Is it the same for Salesforce, Odoo, and Apple work?
Yes. Same rate card, same two lines on the bill.
Try It on Something Small
Bring one problem. We will tell you how we would build it and about how much AI time it needs.
Schedule a Call