Six steps. Each one ends with a deliverable.

From a 20-minute conversation to a system your team runs, measured on your own files throughout.

The six steps

One step at a time.

First conversation

About 20 minutes. Which tasks cost your team the most time, how many people do them, what has been tried. Nothing is sold at this step.

Deliverable A clear next step, or an honest no
Data rules first

Confidentiality, where data is processed, retention and deletion, agreed in writing before any access. A dedicated mailbox, read-only first.

Deliverable Signed data rules
Diagnostic on your data

Recurring tasks mapped, measured in hours per month and ranked, with a prototype built on a sample of your files.

Deliverable Ranked hours table, prototype, roadmap, IT security file
One task, end to end

The highest volume among the simplest tasks, run on real files with a green, orange or red status. Your team is trained while it is built.

Deliverable The task running on real files, team trained
Acceptance on a measurable criterion

One criterion, agreed before work starts, checked on a sample of your real files.

Deliverable Acceptance report on a sample
Go-live, run and improve

Hosting, monitoring and maintenance. The next task costs less, because your company context is already written down.

Deliverable Monthly review

Principles

AI prepares, your teams decide.

The system flags and drafts; people check, decide and sign.
A person approves every outgoing action.

Nothing is sent automatically.
Data rules come first.

Agreed in writing before any access.
Measured on your files.

Not on industry averages.
One task first.

Prove it end to end, then extend.
Your data stays yours.

Exportable at any time, deleted when you leave.
Said plainly.

What is promised and what is not, running costs included.
No fake urgency.

No countdowns, no invented scarcity.

Deciding what comes first

Illustration, fictitious data

Ranked by hours per month

Tell us where you want AI to make a difference first. No commitment at this stage.