Priced AI Features
AI features at a price you can count.
For the owner, COO or CFO of a 50–500 person B2B company who wants a number before a call. One setup measures the workflow, builds the harness, and connects your systems. After that, each AI feature has a fixed price and a 1-week time box.
The price list
Prices are in US dollars. You pay for setup once, then buy features one at a time.
| What you buy | Price (USD) | Time box |
|---|---|---|
| Setup — 3 weeks, 2 systems connected | $10,000 | 3 weeks, 2 systems included |
| Each extra system | $3,000 | Added to the setup |
| Each feature — ships within 1 week | $2,000 | 1 week per feature |
Worked example. Setup plus 3 features costs $16,000. The first feature goes live about 4 weeks after kickoff: 3 weeks of setup, then 1 week for the feature.
What counts as a feature
One AI action on one step of a workflow, in systems that are already connected.
The action is one of five: extract, classify, draft, route, or answer. Each feature ships with its eval, its logs and its docs. We write the eval before the feature, and the feature is accepted when the eval passes on the cases you agreed. If a step needs more than one action, we split it into more features before work starts.
Counts as one feature
- Extract invoice lines into the ERP.
- Triage a support queue.
- Draft a reply to a quote request.
Not a feature
- A new system connection. That is an extra system, $3,000.
- A new data source. We scope it on the call.
- A chatbot demo. We do not build one.
What setup leaves you
3 weeks of work that every later feature builds on. You pay for it once.
A baseline
How the workflow runs today, measured, so each feature has a before to compare against.
The harness
An eval runner, logs, a deploy path, a named owner on your side, and a runbook that owner can follow.
Connections
Up to 2 of your systems connected. Each system after that is $3,000.
At the end you own the code, the harness, the evals and the runbook.
Why a feature costs $2,000
Most of the cost of an AI feature sits around the model: reaching the data, testing the output, logging what it did, and deploying it without breaking the workflow. Setup pays for that once. After it, a feature is one action against systems that are already connected, tested by an eval we wrote first. That is a week of work with a known shape, which is why it can carry a fixed price.
If a request does not fit that shape, we split it into features or price it as a new connection. You see which before we start.
Who leads it
A senior practitioner leads every engagement, from the baseline to the last feature.
Common questions
They do different jobs. The audit scores a codebase that was generated faster than anyone reviewed it, and says what to keep, repair, or delete. This setup builds a harness inside a working business so that new AI features can be added at a fixed price. If your problem is a generated repo, start with the audit.
We split it into more features before work starts. You see the count and approve it first.
Yes. You own the harness and the connections. Build on them yourselves, or stop there. There is no retainer to cancel.
No. We turn the work down when we cannot get access to the systems the workflow runs in, when nobody can describe the workflow step by step, or when the ask is a chatbot demo. In those cases a fixed price would be a guess.
Bring one workflow.
A 30-minute scoping call is enough to name the systems, count the features, and give you the total.