How the work runs.
Three stages, each earning the right to the next. Every stage is priced off what the stage before it found, so you are never committing to work that hasn't been justified yet.
Diagnose
2–3 weeks · fixed fee
We map where AI can move your numbers, size it in dollars, and read your real conditions: data, systems, processes, and people. You end with a ranked roadmap, a scoped first build, and an honest go or no-go.
Build
6–12 weeks · fixed or value-based
We build the highest-impact use case and put it into production on live data, used by real people, instrumented against a metric you already track. Senior builders, constrained scope, working software.
Run
Ongoing · monthly
Anything deployed decays. Accuracy drifts, systems change, models move monthly. We monitor, tune, and extend what we built, and report performance against the metric it was built to move.
We agree the number before we build anything.
One operating metric, chosen with you and instrumented in the first days of the work, measured at the end off the same instrument.
It sounds obvious and it is the step most often skipped. Without an agreed starting number, the result at the end is unreadable: everyone has an opinion about whether the thing worked and nobody can settle it. This is also the clearest pattern in the research. Organizations with full visibility into their AI operating costs report established ROI at five times the rate of those without.
It is why we will tell you when the evidence says to pause on a build. A diagnostic that recommends nothing is still worth its fee if it stops you spending six figures on the wrong workflow.
How we read readiness.
Our diagnostic scores twelve conditions across four areas. Together they show where an organization can move and where it is likely to get stuck.
Business direction
- Strategic clarity
- Executive mandate
- Value-pool clarity
Organization
- Business ownership
- Change capacity
- Workforce readiness
Technology and controls
- Data readiness
- Technical foundation
- Risk and governance
Execution system
- Opportunity discipline
- Integrated delivery
- Measurement
Each condition is scored on observed evidence rather than self-assessment, and tied to a value pool in dollars. The scoring rubric and interview instrument are ours, developed across engagements and refined each time.
Most AI programs stall on a condition nobody assessed. A company with excellent data and no change capacity fails differently from one with a clear mandate and unusable data, and the right first project is different in each case.
What we actually do in the room.
The same five moves, at every scale. AI expands the solution space. It does not replace the discipline.
- Understand friction. Stakeholder and workflow mapping. Where does the work actually slow down, and who feels it?
- Identify opportunity spaces. Value pools and workflow priorities, sized against the P&L rather than against what is technically interesting.
- Assemble the team. Business and technical people in the same sessions, because a workflow nobody owns does not survive the handover.
- Define solutions and estimate ROI. Impact, feasibility, effort, and risk, scored side by side so the trade-offs are visible.
- Build the prioritized roadmap. Sequence, owners, milestones, and the resources each step needs.
What makes this different.
We bring your team along for the build
Your people are in the working sessions, not interviewed for them. They help shape what gets built, so they are bought in by the time it ships and able to run and change it afterwards. That is the difference between a tool your company owns and a tool your consultant owns.
Fixed fee, scoped from what we find
Never hourly. Each stage is priced off the stage before it rather than off assumptions made before anyone looked at your business.
Practitioners, not advisors
Senior operators who have built and shipped. The people in the room are the people doing the work, and there is no second team waiting behind the pitch.
Model-agnostic and conflict-free
The right tool for the problem, whether that is commercial, open source, or nothing at all. We don't resell a stack, so there is no lock-in and no conflicted advice.
We hand it over and leave
The test we hold ourselves to is whether it is still running after we go. Ongoing support is available and it is a choice you make, not a dependency we design in.
A pace the organization can absorb
Adoption usually decides whether the work shows up in the numbers. Change that outruns the people doing it tends to get dropped, so we set the rate against what the team can carry.
The first step is small on purpose.
Two to three weeks, one company, one function, a price agreed before we begin.