Four shapes
1. Architecture review - 1 to 2 weeks
For teams with something already running that is slow, expensive, unreliable, or about to meet a security review. You get a written teardown: failure taxonomy from your own traces, eval and guardrail gaps, a cost and latency model, and a prioritised fix list with effort estimates. Fixed fee. No obligation to continue.
2. Discovery Sprint - 2 to 4 weeks
For teams who know the outcome they want and not the path. You get a working thin slice, an architecture your engineers can build against, an eval harness, a security pre-check, a cost and latency model, and a roadmap your board can read. Fixed fee, credited against a build.
3. Embedded build - 4 to 10 weeks
Shipping alongside your team, in your repo, in your review process. Behind flags, gated by evals, with traces and runbooks. Monthly block. The deliverable is a system in production and a team that can extend it.
4. Fractional - recurring
A standing block each month for teams who have engineers and need senior AI judgement: architecture reviews, hiring loops, vendor decisions, eval strategy, and a second opinion before the expensive commitment. Monthly, cancellable.
{{/* Concrete day rates and monthly block prices intentionally omitted until Alexander sets them. Add them as a table here when decided - the FAQ above already explains why a number needs a scope. */}}
What drives the number
- Integration count. Every internal system is auth, rate limits, pagination, error semantics and a staging environment that does not exist. This is the largest variable by a wide margin.
- Data condition. Clean structured data is a week. Scanned PDFs with tables and inconsistent templates are a month.
- Compliance surface. On-premise, data residency, audit logging and a formal security review add real weeks. Worth doing, not free.
- Team readiness. An engaged engineer on your side makes the work faster and the handover real. No counterpart means everything routes through me, which is slower and more expensive.
- Definition of done. “A working demo” and “an SLA your support team can live with” are different projects.
Contract versus a full-time hire
Hiring a strong agentic AI engineer full-time is the right long-run move if you will have this problem forever. It is also a two to five month search, a competitive comp band, and a bet made before you know what the system needs to be.
| Contract / fractional | Full-time hire | |
|---|---|---|
| Time to first commit | Days | Months |
| Risk if the use case dies | Ends with the engagement | A salaried role without a mission |
| Best for | Getting to production, de-risking a bet, unblocking a team | Owning a system that is already proven and permanent |
| Knowledge retention | Depends entirely on handover discipline | High by default |
| Cost shape | Higher per week, bounded in total | Lower per week, unbounded in total |
The sensible order for most teams: contract to get the first system into production and prove the use case, hire full-time to own it, and keep a fractional block for the next bet. Handover is written into the engagement for exactly this reason - if your new hire cannot take it over, the engagement failed.
Next step
A 20-minute call. Scope, constraints, budget, and an honest answer - including “you do not need this,” which is a real outcome and costs you nothing.