AI Architect
Decide where the model is allowed to act.
For CTOs and heads of platform choosing how AI sits in the stack. This seat decides what is an agent, what stays a workflow, where data is allowed to go, and which choices you will still want in 18 months.
Everyone wants an agent. I need someone who will tell us what should not be one.
What the AI Architect owns
System shape
RAG, workflow, agent, fine-tune, or none of these.
Boundaries
Data boundaries, model routing, and latency and cost ceilings.
The architecture pack
A pack your security review and your next hire can both read.
The people who design it write it. Average 15 years. There is no junior bench to rent. The engagement has to survive an architecture board and a security review.
Rebuild the product you run. Staff it as a milestone, not a req.
For CIOs and CTOs who are ready to hire in order to rebuild the product they already run. The AI Architect decides what stays, what is replaced, and where a model is allowed to act. A squad then ships that replacement on fixed milestones.
I need to hire a team to rebuild this. The current product can't carry what we're asking of it.
The first deliverable is the boundary.
What is replaced, what stays a workflow, where data may go, and which choices still hold in 18 months. Existing users stay on the current product until a path is actually replaced.
The people who design it write it.
Average 15 years. The engagement has to survive an architecture board and a security review.
The milestone does not wait on the req.
Hiring can stay open for the person who operates the system afterward. The engagement leaves a production system, an architecture pack, and an internal owner, and then it ends.
A full product rebuild is a sequence of milestones. Four weeks is the shape of a first decision and a first replaced path. It is not the shape of the whole rebuild.
This job starts here, then continues on AI Pivot Squad on fixed milestones. Add a Harness Engineer when model behavior is part of the rebuild. Use AI Code Cleanup only when the current tree was generated and nobody owns it. A legacy product is a rebuild. Business Process AI is the wrong shelf when the job is the technical product.
Proof pattern for this seat
Alexey Kichin's platform background is the proof pattern for this seat: enterprise architecture across banking and FinTech, including six years as Enterprise Architect for Emerging Markets at Deutsche Bank. On this site he remains Senior AI Platforms Practitioner. Staff the role from that depth.
Where this seat sits
Sovereign AI
Lead seat when the decision is the data boundary.
AI Pivot Squad
The architect decides the shape. The squad builds it.
Business Process AI
The architect draws the boundary before a workflow agent is inserted into a process.
Common questions
Yes. That is the job. An enterprise architect who has not shipped a model, or an engineer who picks a framework first, is a different hire.
No. His title on the About page stays Senior AI Platforms Practitioner. His background is the proof pattern for the seat.
No. It is one named person, a fixed milestone, and an architecture pack. The engagement is designed to end. Hiring for the person who operates it afterward can stay open. The milestone does not wait on that req.
Draw the boundary before the build.
A 30-minute scoping call is enough to tell whether the architect is the first seat, and whether a squad follows.