How it works.
One agent rather than a fleet: it knows your business, has a personality your company recognises, and stays our responsibility for as long as you have it. It arrives the way a good hire does: slowly, then permanently.
-
01 Audit
A written scope, signed by both sides
- Interviews with the people who will use it, and a map of every system it touches
- You decide what it is called, where it lives, and what it is like to work with
- A signed scope, and a fixed price to build it
-
02 Integration
An agent built inside the systems you already run
- Built into your own systems, on your own data
- Tested against real work, never against a demo
- One quarter, because you open up your processes and your people at your own pace
-
03 Retainer
Responsibility that stays with us
- Moved onto new models as they arrive, and repaired when that breaks something
- Re-scoped once a quarter, as your work changes
- When you call, you reach the team that maintains your agent
From 2 000 EUR a month, no end date, for as long as you have the agent
Model usage is billed to you at cost and sits outside the retainer. Moving the agent onto a stronger model is inside it, not a separate fee.
How it starts
One team, one agent, one quarter. Focused rather than small, so the agent proves itself on real work before you widen it. One team, then the department, then the company: the same agent, and neither the integration nor the retainer changes as it widens.
-
Month 0
Scope signed
-
Month 3
Agent live
-
Month 6
Team and model review
-
Month 9
Team review
-
Month 12
Team and model review
What stays constant
None of these change when the technology underneath them does, which is the whole reason the relationship lasts.
-
The way you talk to it
You talk to it where your team already works. Nobody relearns anything when the model underneath is replaced.
-
Your data
It works inside your environment, on your systems. Nothing is copied to a new platform, and nothing is pooled with anyone else’s.
-
Its memory
Every task and decision stays with the agent and compounds. The part that is worth more each year, and the part a new tool throws away.
-
Your price
A flat monthly retainer. Model usage passes through at cost, so what you pay us does not move when the market reprices.
What we move underneath it
The intelligence is a component, not the product. We replace it deliberately, on a schedule, so you take the upside of a fast market without absorbing its instability.
-
The model
Each task runs on the model suited to it. When a stronger one arrives we move that task across, and your team notices nothing except better work.
-
The vendor
No one owns your agent. Moving it from one vendor to another is our work and not your risk.
-
The cadence
We review the field every six months and the working relationship every three, with your teams in the room. Every change is agreed before it happens.
Never stuck on last year’s AI. Never rebuilding around this year’s.
Your data stays yours
The usual trade is capability against confidentiality: either you keep the good models away from the real work, or you let your operating data walk out the door. The agent is built on three separable levels, and you decide how far each one goes.
-
It is built in your repository
The documents the agent is grown from, and in time its code, live in your own version control system. What makes the agent yours sits where your source already sits.
-
It runs on your infrastructure
The servers the agent runs on can be yours. Nothing has to be hosted by us for it to work.
-
The model can be yours too
At the far end of the dial the model itself runs on hardware you own, on open weights, and nothing leaves the building. Capability and confidentiality trade against each other at that setting, so it is a choice you make rather than a default we impose.