Past autocomplete. Building with models instead of typing into them.
Most developers use AI as smarter autocomplete and leave it there. The interesting part starts where it enters your own system: a model behind an endpoint, with your data, your rules and an error path that holds. That is what this session is about.
- Duration
- Half a day, 3.5 to 4 hours
- Format
- In-company, in your own environment
- Group size
- 4 to 12 participants
- Level
- You code daily and know HTTP APIs
- Language
- Dutch or English
- Location
- At your office, or online
Who it is for
This session is for
- Development teams wanting to build AI into a product or internal tool
- Developers who want to go further than a chat window
- Teams that built something and are hitting cost, latency or unpredictability
- Tech leads who want to know what is realistic before promising it
And not for
- Non-technical teams — start with Getting started with AI
- Anyone mainly wanting no-code integrations — that is AI workflows and agents
- Teams looking for an off-the-shelf product rather than building something
Is your situation listed above? Say so in the request and we will point you to the session that does fit.
What it delivers
After this session you can
- Call a model from your own code, with structured output you can validate
- Pick the right model per task, weighing cost, speed and quality against each other
- Build an assistant with your own tools, and know when an agent does more harm than good
- Bring in proprietary knowledge with embeddings, and recognise when retrieval beats stuffing context
- Build an error path for answers that are wrong, too slow or too expensive
- Test and measure something that does not give the same answer every time
Programme
What the session looks like.
- 30 min
The model as a dependency
What really happens on a call, what tokens and context windows mean for your architecture, and why cost and latency are design choices here rather than an afterthought.
- 45 min
Structured output and validation
A model returning JSON is not yet a contract. We build a call with a schema, validate the output and make the path for when it goes wrong.
- 45 min
Tools, assistants and agents
Letting a model choose among your functions. Where that is strong and where it fails, plus the honest story about agents: when you want one and when a dumb workflow is better.
- 40 min
Getting your own knowledge in
Embeddings, search and retrieval. When you send context along and when you fetch it, and what that does to cost and quality.
- 35 min
Keeping it in production
Evaluating, logging, versioning prompts and dealing with a vendor replacing its model. The parts everyone skips and pays for later.
- 25 min
Your own use case
Together we sketch the setup for something you actually want to build, including where it will probably get stuck.
An indication, not a timetable. After the intake we move blocks around, or drop one, if your situation calls for it.
What you take home
Tangible, and yours.
- Working example code from the session, in your own language and environment
- A sketched-out setup for the use case we worked on
- A checklist of what is needed before something like this goes to production
- The thirty-day plan
Preparation
What you arrange beforehand.
- A laptop with a working development environment and the rights to install
- An API key per participant or per team, with a small budget on it
- A real use case from your own backlog
- Access to a test environment or a dataset you are allowed to practise on
This is the most common reason a session only half succeeds: accounts that do not work or missing permissions. We send this list with the confirmation too, but sort it out earlier if you can.
The trainers
We run every session together.
We are not a training institute but an agency that does this work. What we show runs at clients, not only in the classroom.
Maarten Clement
Founder, AI workflows and automation
Builds the workflows, GPTs and integrations that survive a session. From first prototype to something that runs every day.
Michael van Wijhe
Founder, performance marketing and data
Works daily on campaigns, dashboards and scoring models for clients in the Netherlands, Belgium, Germany and Central Europe.
Investment
What determines the investment.
We do not put a price on the site, because no two sessions are the same. What does determine the price we can happily explain — so you know what you are looking at before you ask.
Number of participants
Up to fifteen works well. Above that we split the group, and that changes the setup.
Duration
Half a day or a full day. A full day is not two half days: it adds practice, not theory.
Degree of tailoring
Working with your own accounts, data and campaigns means preparation on our side.
Location
At your office, at ours, or online.
You get a programme proposal with a fixed price within one working day, no strings attached.
Would you rather see a price first and sign up on your own? The masterclass performance marketing and AI is also in the open programme at NIMA.
Practical
Other things worth knowing.
In-company, no open enrolment
We come to you and the session is about your work. The masterclass via NIMA is the one exception.
Always two of us
One in front of the group, one walking around. In hands-on work that is the difference between keeping up and dropping out.
Dutch or English
Mixed teams too: we run the session in English when that works better, materials included.
Rescheduling is fine
Free of charge up to two weeks beforehand. After that we find a solution together.
Frequently asked
What people ask us beforehand.
What does it cost?
That depends on the number of participants, the duration, how much tailoring is involved and where we run it. Request the programme: within one working day you have a proposal with a fixed price, no strings attached.
Can a single person sign up?
Not for our own sessions: those are in-company and built around your team's work. Anyone wanting to attend alone can join the masterclass performance marketing and AI, which is also in the open programme at NIMA.
How quickly can you do this?
A session can usually be scheduled within three to four weeks. If you are in a hurry, say so — sometimes it can be sooner.
What happens after I submit the form?
We read your request and call or email within one working day for a short twenty-minute intake. The programme proposal with the price follows after that. You are not committed to anything.
Do participants get a certificate?
From us you get a confirmation of attendance, not an examined qualification. If you want an examined certificate, the NIMA route is the right one.
Does this help with the AI literacy obligation under the AI Act?
Yes. The AI literacy obligation has applied since February 2025. The AI Omnibus (July 2026) made it lighter: organisations must support the AI knowledge of their staff rather than guarantee a specific level. Our sessions cover safe use, privacy and the limits of a model as standard, and you get a participant list to evidence who has been trained. We are not an examining body; for a formal certificate we point you to NIMA.
Which language and which models do you cover?
We work in the language you use; the examples are ready in TypeScript and Python. We cover models comparatively rather than by brand — what you learn still holds when you run something else next quarter.
We cannot send data outside. Is this still useful?
Yes, and it gets more interesting. We cover what is possible locally and inside your own tenant, the trade-offs that involves and what it costs in quality. Mention it in the request and we will shape the session around it.
Request
Request programme and price.
You get a proposal with a fixed price within one working day. No strings attached, and without having to sit through a demo first.
- We reply within one working day.
- In the call we sharpen the bottleneck together. That costs you an hour.
- If it fits, we build a working prototype within five working days. Free of charge.
Other sessions
Not quite the right fit?
AI for managers
Half a day, 3 hours
Management session: introducing AI
Half a day, 3 hours
Masterclass performance marketing & AI
A full day
