From the wish
Dear Sir or Madam,
We hereby inform you of the additional costs that have arisen. Additional costs are costs that were not included in the original quote. The costs amount to [amount].
We trust that this has informed you sufficiently.
Workshops of one, two or three days, at your location, on your own work. After that they can do it without us.
Just write a quick email about those extra costs.
AI is a fast apprentice who has read everything and has never set foot in your company. You give it a clear brief and you check its work.
The difference isn’t in the AI. It’s in what you ask. Give it a wish and you often get the email on the left. Give it a brief and you get something like the one on the right.
From the wish
Dear Sir or Madam,
We hereby inform you of the additional costs that have arisen. Additional costs are costs that were not included in the original quote. The costs amount to [amount].
We trust that this has informed you sufficiently.
From the brief
Rush delivery on Tuesday: your approval, please
Hi Mark,
You asked for delivery on Tuesday. That wasn’t in the quote, so we’d like your approval first. The rush delivery costs €185 excl. VAT. You’ll then have it on Tuesday before 12:00.
Is that OK? If so, please let me know by email by Friday.
Assumed: the contract says nothing about rush orders.
Not sure: whether Friday is early enough for the driver.
You read it through. The “Not sure” line tells you where to look.
Demo data: we wrote both emails to show the difference.
More than half of the time they work themselves, on their own work.
For work that keeps coming back, made in the AI tool you already have.
A traffic light for what may go in, and six working agreements, for all staff.
It matters more than asking well. Of everything AI makes, your team asks five questions:
Facts and figures held up against the source.
Names, dates, amounts, rules: copied or invented?
What would you add that AI couldn’t know?
Would you say it like this to this customer?
If not, it doesn’t leave the building.
All prices exclude VAT, for up to 12 participants.
Now
A free talk. We see whether it fits and which format you need. If it does, you reserve the dates.
2 to 3 weeks ahead
A half day with the client and two key people, or an hour for the first day. Which work comes back, which tools are there, who handles privacy?
After that
We build your anonymised examples into the exercises. You test the logins.
The days themselves
On site, 09:00 to 16:30, consecutive or spread out. Short demos, then hands-on with their own work.
Within a week
The agreements for everyone, the certificates digitally, and a report: which assistants exist, who the champions are, what’s still open.
After 30 and 90 days
Everyone shows what their assistant does now. Then we choose the next step: more people, a second assistant or a real automation.
More than half of the time they work on their own work. The full programme
The programme was first written for a construction company. In the intake we replace the examples with ones from your sector.
All four are reasonable, and they come up in every group. A sceptic who finds a mistake is a gift: that’s when we show how to steer.
Thought: AI isn’t for our kind of work.
Revised: A lot of time leaks away between the shop floor and the office: retyping, searching, chasing.
Thought: AI makes mistakes, so it’s useless.
Revised: An apprentice makes mistakes too. That’s why you check.
Thought: I first need to understand how it works.
Revised: You learn it by doing it: fifteen minutes a day.
Thought: AI will take our jobs.
Revised: Some work will go, that’s the honest answer. The question is who chooses which. Here, you do.

Laurens van Riel builds AI workflows at Van Riel Online Services in Rotterdam that run every day: texts in a company’s tone of voice, purchase invoices held against the price list, a phone line that answers in Dutch. He gives the workshops himself.
This isn’t demo data. These workflows run at our company, at fixed times, and a person keeps the last word.
An introductory talk is free. Together we look at which work keeps coming back, which AI tools you have and which format fits.