Eight weeks for AI to become part of the work, not just the talk.
We start from your company's context, train your teams, review all your processes, select three to five and put them into production one by one, each with its process sheet. At the end, your team knows what the AI does, why, and how to develop it further.
The right time, and the wrong one.
For you if
- several teams lose time on the same tasks;
- your staff already use ChatGPT or Copilot, without shared rules;
- you want documented processes, not a collection of tools;
- someone in-house can take ownership of the subject, even part time.
Not yet if
- you want to test it alone first: the owner-only format;
- you are not sure whether AI has a place in your business: the free diagnostic;
- you expect an AI that never makes mistakes: we measure its errors, without promising there will be none;
- you are a municipality, a public social welfare centre (CPAS/OCMW) or a non-profit (ASBL): the public sector approach.
Eight weeks, then a month of adjustments.
Your context, in writing
Two workshops with the business owner: business, clients, rules, and what must never be automated. The context file is written, reviewed and versioned.
The team is trained
One day on site, eight people at most per group: strengths and limits of AI, data never to expose. The usage charter is written with your teams. Each participant receives a named certificate documenting the measure taken under Article 4 of the AI Act.
Mapping and costing
All your processes reviewed with the people who do them: who is involved, frequency, duration, bottlenecks. Three to five are selected and costed in hours per week, errors and lead times. Before handing a process to AI, we check that it should exist in that form: sometimes simplifying it is enough, without AI. You choose the order.
One process at a time
Each process gets its process sheet: objective, data, steps handed over to AI, human check, version. The team concerned tests it for a week on its real files, we make corrections, your business lead signs off the validation, then it goes into production. Technical automation comes last, if the gain and your tools allow it.
A month of adjustments
Process sheets corrected on real cases, prompts refined, incidents recorded in a log. At the end of the month, your team runs its processes on its own, or moves on to monthly follow-up.
For each process
Assessment
We look at the data involved, the rules of your profession and the right hosting. The people who will sign on your side are named at this stage.
Approved process sheet
What the AI does, the accepted error threshold and the stop criterion are written down and signed on your side before anything is built.
Verified configuration
We configure the model, the instructions and the connections, and check that they match the sheet.
Test on your files
Results are measured against the threshold. Your team signs, or the process is corrected, or it is stopped.
Go-live
The team is trained and the date of the first review is set.
Follow-up
At each review, the process is kept, corrected or stopped.
The number of signatories follows the size of your team. There is always at least one signature on your side: we never validate on our own what we have built.
Starting alone, in four to six weeks.
For the business owner whose team is not ready, or who wants to try the method before extending it. The company's context is mostly in their head, and the most expensive processes often go through them.
Everything stays with you.
- Your company's context file
- The map of your processes, each one costed
- Three to five process sheets in production, versioned, with the prompts and the model used
- The internal AI usage charter
- The register of authorised AI tools
- One validation file per process: cases tested, error rate, your business lead's signature
- The incident procedure
- The named training certificates, Article 4 of the AI Act
- The dashboard of hours before and after, per process
Continue, or not.
Monthly follow-up
Process sheets kept up to date, incidents handled, a new process launched when the team is ready. No minimum term.
Training only
Week 2, sold separately, as a documented measure under Article 4 of the AI Act: on-site training, certificates and usage charter.
Public sector
The same method, respecting professional secrecy and with a suitable public procurement procedure.
Five commitments, put in writing.
At least one signature from your team at every step: we never validate on our own what we have built.
An error threshold and a stop criterion written before the test. Each process is kept, corrected or stopped on its results.
No confidential data sent to an AI without your written consent. Towards a public AI, the data leaves anonymised.
Every deliverable belongs to you, in an open format: process sheets, validation files, usage charter, register of tools.
A single named contact, written into your contract.
After the diagnostic, you receive a written briefing and, if you wish, a quote within three working days.
Three entry prices, shown.
AI visibility measurement
We run your site through the questions your clients ask the assistants, and we tell you who is cited in your place. The measurement commits you to nothing; the work that corrects it starts at 2 900 €. See AI visibility.
Process audit
We review your processes and tell you which ones AI can take on, which ones it must not, and what each would require. You keep the document, whether or not you entrust us with what follows.
First process in service
From the process sheet to the test on your own files, through to going into production and its measurement. The price depends on the process selected and the size of the team to be trained.
The thirty-minute diagnostic remains free. Beyond the first process, the quote is drawn up after the audit: there is no price list by company size.
What people ask us.
How much time do your teams need to set aside?
One day of training per participant. During mapping, 45 minutes to an hour per person involved. One hour per week for the team testing the current process. The rest happens without interrupting work.
And if a process does not work with AI?
We note it in the process sheet and move on to the next one. The costing in weeks 3 and 4 is there precisely to rule out processes where AI would waste time. Sometimes a process improves without AI, by changing one step: that is a result too.
Does the AI need to be connected to your software?
Not at first. The AI prepares, a person validates and enters the data. Full automation comes last, if your tools have an API and the gain justifies it. Otherwise, we tell you at the scoping stage.
How much does it cost?
Process audit: 2 900 €. First process in service: from 7 500 €. Beyond that, on quotation after the free diagnostic. The price depends on the number of processes selected and the size of the team to be trained. The quote details the eight weeks and the month of adjustments. There is no price list by company size.
Who does the work?
Benjamin de Bruijne, from diagnostic to go-live. A single named contact, written into your contract.
Is it a subscription?
Not for the engagement. Follow-up is a monthly flat fee on quotation, with one month's notice and no twelve-month commitment. Its price depends on the number of process sheets maintained and the pace of new processes. If your AI runs on a dedicated server in Belgium, the quote also includes hosting.
Does the team need technical skills?
No. We work on your premises: we install the AI, test it on your files and stay until it runs. Your team brings its knowledge of the business and validates the results.
How is this different from an automation agency?
An agency installs a tool. We first set the acceptable error level in writing, test on your real files, and your team validates. If the process does not deliver its results, we stop it. There is always at least one signature on your side: we never validate on our own what we have built.
Our processes are not written down anywhere. Is that a problem?
No, it is the most common case. We describe them with the people who do them, in a process sheet that you keep.
What does the AI Act change for us?
Article 4 of the European AI regulation, as amended by Regulation 2026/1744, asks companies that use AI to take measures to develop their staff's AI literacy. Obligations for high-risk systems apply from 2 December 2027, and from 2 August 2028 for regulated products. For most professions, the most pressing question lies elsewhere: professional secrecy, data protection and liability when something goes wrong. AI governanceCompliance and data.
The quote starts with thirty minutes.
Free diagnostic by video call, on one of your real processes. The quote follows, week by week. You leave with the process best suited to AI, how we would test it, the hosting that fits your rules, and what is better not to hand over to AI. All of it is given to you in writing, in a briefing you can share internally.
You speak with Benjamin de Bruijne, from diagnostic to go-live.