How an engagement runs
A scoping call, an assessment that sets the baseline, one fixed milestone, and a handover to someone on your side. You buy one milestone at a time, and each one is written down before it starts.
From the first call to handover
Five steps. The fifth is optional.
1. Scoping call
You tell us the workflow, the product, the repo, or the data boundary. We tell you which engagement fits, or that none does. "No engagement" is a normal answer on the first call.
- What you get: an entry point (a service line, a named seat, or AI Code Cleanup), or a plain no.
- How long: 30 minutes.
- Done when: both sides know whether there is something to scope.
2. Assessment and baseline
Before anything is built, we measure where you are. For a workflow that means volume, cycle time, and cost per transaction. For a repo it means what the code does, what depends on it, and what it costs to keep. For a regulated deployment it means the data boundary and the hardware the workload needs. The baseline is the number the milestone is measured against. How we set that number is in calculating AI ROI.
- What you get: the baseline, the number the milestone should move, and a written scope.
- What we need: access to the process, the repo, or the data. Without it there is no baseline, and we do not start.
- Done when: you can say what the milestone will be measured against.
3. One fixed milestone
One milestone, with a written scope and a time box. The time box depends on the offer and is agreed before work starts. The weeks below are methodology targets, calibrated to your baseline, not results from past clients.
- Business Process AI: first measurable financial impact in 4–8 weeks, on one workflow.
- AI Pivot Squad: the first milestone is structured to reach first value in 4 weeks. A product rebuild is a sequence of milestones, and 4 weeks is the first of them.
- Sovereign AI: sizing happens in the assessment, and the real workload is benchmarked before anyone buys hardware. The milestone ends with the model and the workflow running inside your boundary.
- AI Code Cleanup: a $10,000 audit, a scored keep, repair, or delete report in 7 days. The repair is quoted separately, once the report exists.
4. Handover
The code lands in your repository. A named owner on your side runs the system, with documentation and runbooks written for that person, who is trained on it during the milestone. Thirty days of hypercare follow go-live. That is not a retainer.
- What you leave with: the running system, the baseline and the measured result, documentation and runbooks, and a named owner.
- Done when: the system runs without us.
5. A next milestone, if there is one
Some clients stop after one milestone. Others scope a second: another workflow, the next replaced path in a rebuild, or a named seat such as AI Ops to watch the model's behavior and cost after launch. Each one is a new fixed milestone with its own scope. There is no open-ended retainer.
Who does the work
A senior practitioner leads every engagement, so the person who scopes the work is the person who builds it. We take a small number of clients per quarter. There is no junior bench and no handoff from a sales team to a delivery team.
Mikhail Zenin (production AI engineering) and Alexey Kichin (platform architecture) are named on the About page. The full catalog is on the services page.
What we will not do
We are explicit about the work we turn down. It is easier to be useful inside a clearly defined frame.
No PoC theater
We do not build pilots whose purpose is to justify the next pilot. If a workstream cannot ship to production, we do not start it.
No headcount-multiplier billing
Our economics do not improve when we put more bodies on your account. We staff for the work, not for the invoice.
No rip-and-replace
Your existing systems are not the enemy. We integrate with them. The bar for replacing infrastructure is a P&L case, not an architectural preference.
No model-of-the-month
We choose the model that meets the spec at the lowest defensible cost and stick with it. Re-platforming because a new model shipped last week is somebody else's hobby.
No vendor lock-in we cannot defend
Where lock-in is the right call (and sometimes it is), we say so and show the math. Where it isn't, we design for portability from day one. Our sovereign AI on-premise service exists precisely for clients where this matters most.
Common questions
How long does an engagement take?
One milestone at a time, with the time box written into the scope before work starts. The time box depends on the offer. Business Process AI targets first measurable impact in 4–8 weeks, set against the baseline. A Squad's first milestone is structured for 4 weeks. The AI Code Cleanup audit takes 7 days. A product rebuild is a sequence of milestones, not one long project.
How are engagements priced?
Each milestone has a fixed scope, written down before it starts. There is no open-ended retainer and no billing by headcount. We publish two prices. The AI Code Cleanup audit is $10,000 for a scored report in 7 days, and the repair is a separate quote. Priced AI Features costs $10,000 to set up, then $2,000 per AI feature.
What do you need from us before you start?
Access to the process, the repo, or the data, depending on the offer, and someone on your side who will own the result. Without access we cannot set a baseline, and without a baseline we do not start.
What if our data is not clean?
It usually isn't. The assessment measures data quality along with everything else. It is not a condition for starting. If the data cannot support a baseline, you hear that before any build is scoped.
Are you a boutique AI consultancy, as an alternative to a Big 4 firm?
Yes. A senior practitioner leads every engagement, so the person who scopes the work also builds it, and there is no analyst pyramid underneath. We ship production code rather than decks, and each engagement is designed to end.
Are you a hands-on AI implementation firm, or a strategy consultancy?
We implement. The leave-behind is a running workflow or a running system, and we turn down a strategy narrative with no production system behind it. Product engineering is AI Pivot Squad. A workflow on the operation you already run is Business Process AI. Not sure which fits? Start at AI implementation.
What happens after handover?
The code is in your repository with its documentation, and a named owner on your side runs the system. That person is trained on it during the milestone. Thirty days of hypercare follow go-live. After that, more work is a new fixed milestone, or an AI Ops seat if you want someone watching the model's behavior and cost. There is no ongoing retainer.
Why don't you publish case studies?
We do not publish client names or results. What we do publish is what you can check before you sign: the audit price and the Priced AI Features price list, the time box for each offer, what you leave with, and the work we turn down.
Book a scoping call
The first call is 30 minutes. We use it to name the workflow, the product, the repo, or the boundary — or to say there is no engagement.
Book a scoping call