Six weeks. One agentic flow. It pays for itself.
We have run this blueprint inside financial services, construction, retail and pharma businesses. Three weeks working out where the time and the money actually go, three weeks building. You finish with at least one agentic flow in production that recovers the cost of the whole engagement in six to twelve months.
The same blueprint,
pointed at your operation.
The shape holds every time: find the work that costs the most and changes the least, build one flow that does it, then measure what came back. What changes is which process you point it at. That is what the first three weeks are for.
Where it has run
Three weeks looking.
Three weeks building.
The looking half is the one people want to skip. It is also the half that decides whether the building half pays for itself.
Find the flow
worth building
We sit with the people doing the work and cost it honestly. Most of what looks automatable turns out to be cheap already; the expensive work is usually somewhere nobody has measured.
- Where the hours go, and what an hour costs
- A shortlist, ranked by payback. The most interesting candidate is rarely the one that pays
- One flow chosen, with its definition of done written
- The security and data questions answered before anything is built
Build it and
put it in production
A pod, our small team of senior people directing agents, builds against the definition of done from phase one. Same model as every other engagement: agents execute, people approve each step, and nothing reaches production without a human releasing it.
- The flow, built and iterated in lower environments
- Verified against the acceptance criteria written in phase one
- Infosec sign-off, then a human release
- The measurement wired in from the first day it runs
The same pod model as everything else we do · see how a pod works
At least one flow,
running, with a number on it.
The flow we pick has to clear one bar before we build it: the hours it gives back have to cover what the six weeks cost, inside a year. That constraint is what keeps the first project honest.
The engagement pays for itself before the next budget cycle.
Consulting and implementation together, recovered from the hours the flow gives back. We model it in phase one and we publish the working, so you can argue with the number before you spend anything on phase two.
The first one
is the expensive one.
Most of the cost of a first agentic flow has very little to do with the flow. It is the access, the approvals, the security review, and the long argument about whether any of this is safe to do at all. All of that is paid once. Once it is settled and there is a working example inside the building, the second and third arrive in a fraction of the time.
Access and approvals, already granted
The systems are connected, the permissions are scoped, and the governance paperwork exists. The next flow inherits all of it.
A security posture that has already been reviewed
Guardrails, the authorization process and the review checkpoints were argued through once. The second flow is a much shorter conversation.
A measurement framework everyone has agreed to
The first flow establishes how you count hours given back and what a good result looks like. That is what lets you rank the next ten by payback, on numbers everyone has already agreed to.
A working example everyone has seen
The most useful output of the first six weeks is often internal: people stop debating whether this works and start bringing you candidates.
One flow is the deliverable · the capability to keep going is what you keep
Four things
it moves.
Which of these leads depends on the process you point it at. In practice the first flow usually moves more than one.
Process
Work that waited on a person to start it now starts itself, and the handoffs between steps stop being where things go to sit.
Operational cost
The hours come back first in the team that owns the process, and the saving is measured in the same terms the business already budgets in.
Client satisfaction
Most of what irritates a customer is waiting. Flows that used to run overnight or over a week tend to finish while they are still paying attention.
More with the same team
Capacity arrives without hiring, which matters most to the teams that were never going to get the headcount anyway.