Diagnostic
3 weeks
For AI spend you are about to make expensive: seat renewals, an agent rollout, a build-vs-buy call. You leave with a recommendation, the alternative we rejected, and a first slice scoped to a team you already have.
Approach
A short path from a vague technical worry to a decision someone owns.
We do not arrive with a framework to install. We arrive with a sequence: understand the constraint, choose, prove it in production, and leave your team able to continue without us.
How an engagement moves
01
Frame
We sit with the constraint. Interviews are short; artifacts — cloud bills, seat licenses, margin reports, support queues — are long. We leave with the problem written in one paragraph both a CTO and a CFO will sign.
02
Choose
Options, including the one where you spend less: renegotiate seats, cap token sprawl, kill the pilot that cannot pay back. The output is a decision record, not a menu.
03
Prove
A thin slice in the real operation: one team, one workflow, metered end to end. We would rather be wrong on a small bill than confident on a model.
04
Hand over
Your team runs the rest. We stay through the first measured readout — cost per outcome, margin defended — then leave the operating rhythm behind: owners, budgets, and the meeting where the number gets read aloud.
Ways to engage
Pick the length that matches the decision.
Build partnership
A quarter
We embed with your engineers and finance through one outcome — an agent fleet under budget, a cost-governed rollout, a margin defended. We do not staff a standing army. We staff the outcome.
Advisory
Retainer
A partner in the room for the decisions that should not wait for a project: pricing, vendor negotiation, a bill that revealed a missing owner, an incident the agents caused.
Principles
Decisions, not decks
Every engagement ends with a choice someone can execute: a price, a budget, a scope item killed. If a finding cannot become an owner, a date, or a deleted cost, it does not ship.
Founder on the work
The person who scopes the work does the work. No handoffs to a bench of generalists, no layer whose job is summarizing the people who understand the system.
Numbers before narratives
A decade in audit and international tax before a decade shipping systems. We agree the measure — cost per outcome, margin impact, payback period — before we design anything.
Fit the stack you have
The brave architecture that ignores your vendors, your contracts, and your finance team is not brave. We design for the company that will still be here next year.
If the problem is still fuzzy, start with a diagnostic.
Three weeks. A written recommendation. No standing team required until the slice is worth building.