Your team already uses AI. Everyone on their own.
We configure and bring order to the tools your team already uses — Claude, ChatGPT, Copilot — so they stop being scattered accounts and become a shared way of working.
If your team works like this, it's your case.
Small teams whose output is documents and decisions rather than transactions: usually fewer than thirty people, or one department or committee inside a larger company. A lot of sensitive information and few people to organise it.
Everyone uses AI on their own.
Their own account, their own depth of use. What one person learns never reaches the rest.
Context gets pasted in by hand every time.
And the output is then moved into Excel, PowerPoint or Word. Same again next month.
The same analysis is rebuilt from scratch.
Every cycle reassembles what the last one already assembled, with almost nothing reusable.
There is no central administration.
Nobody has a clear picture of which accounts exist or who can reach what.
Some information can't be seen by everyone.
Client material, confidential third-party documents, matters under analysis. Today each person draws that line themselves.
From how your team works today to how it could
Neither column is about new tools. They are the same ones, put in order.
- Individual accountsCentral administration
- Uniform or unclear accessProfiles, permissions and perimeters by source
- Context supplied by hand each timeConnected sources and shared context
- Ad-hoc conversational useDocumented, versioned procedures
- Uneven outputCommon formats and standards
- Knowledge held by individualsKnowledge held by the organisation
The order is not negotiable.
Access, then context, then procedures. Each phase rests on the one before it, and nothing gets connected or automated before the perimeter is defined.
Phase 01
Diagnosis and architecture
We look at which tools and accounts exist, what data they touch and how they are really used. We design the access boundaries, list the processes that repeat and measure the starting point. You leave with a document that is yours.
Phase 02
Perimeter
Central administration, profiles and permissions, and clear boundaries by source and by folder. Only then are sources connected, and always inside those limits.
Phase 03
Shared context
We organise the knowledge, prepare the context each process needs and fix common output formats, so nobody pastes the same thing in by hand again.
Phase 04
Procedures pilot
We turn two or three recurring processes into repeatable, versioned procedures, each with a validation step before real use.
Phase 05
Measurement and handover
We measure against the starting point and leave you the administration runbook, the standard for writing new procedures and clear roles: who creates them, who validates them and who administers access.
The point is that you can carry on without us. Administration, procedure authoring and governance roles are handed over in the last phase, with a runbook and an authoring standard. Staying involved is optional, not structural.
The perimeter is work, not a configuration step.
Teams like this hold several categories of information with different visibility rules: your company's, your clients', and confidential third-party material about matters that may never go ahead. The perimeter design defines profiles, permissions and boundaries by source and by folder. And it determines how far each procedure and each connected tool can reach.
- Profiles and permissions by source, not per individual
- Boundaries by folder and by where the information came from
- Defined before anything is connected or automated
The recurring work that ends in a document
What repeats every month or quarter and today gets rebuilt from scratch.
Periodic reviews
Monthly business and strategic plan tracking, in the same format, without reassembling it each time.
Analysis and benchmarks
Recurring studies and comparisons where the structure repeats and only the data changes.
Reporting and follow-up
Financial and activity reporting drawn from sources that arrive in different formats and at different intervals.
Committees and meetings
Preparing and following up recurring meetings: pre-reads, decisions taken and what is still open.
Before touching anything, this
Sometimes something underneath needs automating
This door builds nothing. But mapping the processes almost always surfaces something that does need automating — preparing the information before it comes in, for instance. That is scoped separately.
See bespoke automation →Security, one of our foundational values
EU-hosted
Database and processing in EU regions. No transfers outside the EEA without your consent.
Minimum access
Only the team assigned to your project touches your data, and only to keep the agent running.
Deletion on request
You ask, we delete — within 30 days, including operational copies.
No reselling, no training
Your data is never sold or used to train public AI models.
Thirty minutes to know where to start
A Google Meet video call. We look at your case, tell you which door is yours and what we would do first. No commitment and no hard sell.
- We review your processes and bottlenecks
- We tell you which door is yours and where to start
- You talk to the team, not a bot
Available times