An AI map and tools implemented in your operations, from the first prompt to an automated working system. For managers, owners, and the teams they run.
Two formats. A half-day that gets a team using Claude, ChatGPT, Perplexity or Copilot properly, with the decision kept where it belongs. And a build program that leaves your team with a solution that runs on Monday, owned by someone inside your organisation.
Most organisations already pay for the tools. What they do not have is a team that uses them the same way twice. One person gets remarkable results and cannot explain how. Another pastes a client file into a public chat window. A third stops using the tool after a bad answer and goes back to doing everything by hand. Six months in, the licence is renewed and nothing in the operation has changed.
A map: which tasks in the team stay human, which get assisted, which get automated, and in what order. A method: instructions written down so the tool produces the same quality for everyone, not only for the person who wrote the prompt. An owner: someone inside the organisation who keeps the solution running after the trainer has left the room. This line delivers the three, in that order, on your team's actual work.
Better, at work. The tool does the repetitive part. Your people keep the judgment.
The half-day is the floor. The build program is where the operation changes. Both start with your team's real tasks, never with a demo.
Three and a half hours, online, for up to twelve people. Each participant arrives with two real tasks from their week and leaves with both of them done with the tool, plus a written way to repeat it. What is covered: how these tools actually produce an answer and why they fail in predictable ways; what never goes into a prompt, in your organisation, with your data; how to keep the decision on the human side and use the tool for the work that benefits; how to check an answer before acting on it. One tool is chosen with you before the session so the team practises on what it will use on Monday. On quote
Five working sessions with one team, online, over four to six weeks, on one process the team already runs by hand. Session one produces the map and the usage policy and captures the baseline: how many hours the process takes today. Sessions two to four build the solution piece by piece with the team at the keyboard: reusable skills for the repeated steps, memory and instruction files that hold your standards and vocabulary, and a versioned repository so every change can be reviewed and rolled back. Session five hands the solution to the internal owner your organisation named at the start, and closes on the measurement plan. Three weeks later, one number: hours per week reclaimed on the process the build covers. On quote
The depth described here, skills, memory files, instruction files and a versioned repository, is built on the platform your organisation chooses; the map, the usage policy and the measurement apply whatever the tool. The scoping call settles which platform fits your licences and your data rules before anything is quoted.
Larger teams, several processes, or a second build on the same team: quoted on the same basis, session by session.
One page. It is the first thing produced and the document a manager can show upward before a single tool is touched.
The team lists what it actually does in a week. Each task lands in one of three columns: stays human, because judgment or accountability is the point; assisted, because the tool drafts and a person decides; automated, because the task is repetitive, rule-based and checkable. Most teams discover that the automated column is shorter than they feared and the assisted column longer than they expected.
Personal information, client data under contract, anything the organisation would not put in an email to a stranger. The map names it explicitly, per task, so the rule is not left to individual judgment on a busy afternoon. This is the raw material of the usage policy.
Which automated task gets built first: the one with the most hours behind it and the clearest way to check the output. The order protects the program from the classic failure, where the most interesting task is built first and the most valuable one never is.
For each assisted task, the map names the person who signs off on the tool's output. A solution without a named decision point is a solution that quietly becomes the decision-maker. The map prevents that in writing.
A written usage policy: what goes into a tool, what never does, and who keeps the decision, drafted with Loi 25 obligations in view and handed to your privacy officer for review.
A versioned repository for every skill and instruction file, so anything the solution does can be reviewed and rolled back.
A baseline captured with you in the first session, and one number measured at three weeks: hours per week reclaimed on the process the build covers.
An internal owner, named by you before the build starts and trained during it, so the solution runs without outside help.
Your team already has the licences, or is about to, and nobody has been shown how to use them the same way twice. You can name one process the team runs by hand every week. You will name an internal owner before the build starts, because a solution nobody owns stops working the week the trainer leaves.
An inspiring talk about the future of AI is what you are after; nothing here is about the future. A tool bought for the team but not yet chosen is not a blocker, the scoping call settles it. A build with no baseline and no owner is not delivered, because it cannot be measured and will not last.
One. Fill in the short form below: your organisation, the format you are looking at, and the size of the team. Two. A ten-minute call decides whether the service fits; nothing is delivered in it. Three. A written proposal within two working days, with the format, the price, the platform, and the financing lines your organisation can use.
Individual AI setup sessions have their own door, their own format and their own pricing. Nothing on this page applies to them.
Go to Personal →Your organisation, the format you are looking at, and the size of the team. Reply within two working days.
A team that knows which tasks to hand to the tool, and which to keep, works differently within a month. Put that team next to others and the operation changes.