Casper HollemansSenior software engineer · 14 yearsIndependent senior software engineer · 14 years
This should be working. So why is AI not working for your team?
I help your team set AI up so you can trust what comes out of it. It starts with an audit: where is it going wrong now, and what do you tackle first?
€750
for the full audit
For a team of 4–6 developers.
A report of bottlenecks and fixes, within 10 working days · about 2 hours of your team · find nothing worthwhile and you pay nothing
No obligation. No sales pitch.
Introductory price
€500
for the first two audits
In exchange for your honest feedback and a short reference.
Teams I have built with
From now to after
You are not the only one. Google's DORA research, covering almost 5,000 professionals, shows that AI acts as an amplifier: it magnifies what already works, and just as readily what does not. Teams with a sound way of working do get faster; for the rest that speed arrives with more mistakes and rework. The difference is not in the tool but in how you work with it.
“I'm not sure AI is making us any faster.”
Bigger and more complex work, and faster too
“We're spending more time fixing than building.”
Reliable AI code
“Everyone's just winging it with AI.”
A team that improves itself
“I'm done with AI.”
Back in control
What the audit gives you
After the audit you have a clear picture of the bottlenecks in your current AI workflow, and for each one a proposed fix the whole team can pick up straight away.
The list is ordered from high impact to low. So you know exactly where to start, and you can get going straight away on what pays off most.
You receive a written report with all the findings. It is yours to keep.
The audit is the first step. If you want to go further, I implement the fixes together with your team and teach you the principles that keep you improving on your own.
- HighNo shared definition of 'done'Proposed fix: Put the review criteria where the AI reads them
- HighArchitecture rules live in the wiki, not in the toolingProposed fix: Move the rules into the repo, beside the code they describe
- MediumEveryone repeats the same context in every chatProposed fix: One shared context layer per project
- LowPrompts that work are lost when the editor closesProposed fix: Keep them in the repo, the same as scripts
Invented findings — this is the shape of the list you get after the audit.
How it works
- 1
Kick-off
We go through the pain points your team has collected (1 hour).
- Read access to your repositories — read-only, I change nothing. I am happy to sign an NDA, and working inside your own environment is fine too.
- Pain points from your developers, collected before the kick-off: where does AI go wrong for them most?
- 2
Analysis
I dig into your codebase and AI setup.
- 3
Going through the findings
We go through the report together and decide what to tackle first (±1 hour).
Usually starts within two weeks. Findings within 10 working days of the kick-off. About 2 hours of your team's time in total.
Price
€750for the full audit
For a team of 4–6 developers.
For teams already using AI who cannot rely on what it produces — including teams who are well along.
Not for: Teams not yet working with AI · Solo developers · Anyone looking for a tool recommendation
If I find nothing worth acting on, you do not pay.
Introductory price
€500
for the first two audits
In exchange for your honest feedback and a short reference.
If you go on to implementation, the audit fee comes off the price. Larger teams or setups: price agreed after the intro call.
About me
I have been building software for more than fourteen years. As tech lead at route.nl I built Routiq, an international platform, from the ground up, and after that I was founding cloud engineer at mosa.cloud. So I know how teams work, and why a tool only pays off once the whole team can handle it.
For a long time my conclusion was that AI is not for enterprise. On complex systems it broke down fast, or I could not trust what came out of it. Then I had to check all of it carefully, which often cost me more effort than building the thing myself. I set out to find why. It was rarely the tool: the AI did not know how the system fitted together, and what 'good' meant was written down nowhere. Out of that I am developing Anchored Principles: principles that make sure you know what you want before you set AI to work, and that what comes back stays as close as possible to what you expected.
I like to get stuck into a hard problem and come back with a simple solution.
What you can count on
"Casper was voor mij de ideale tech lead: inhoudelijk sterk, verantwoordelijk en iemand waarvan ik wist dat als hij zich ergens aan committeerde, ik erop kon vertrouwen dat het ook afkwam."
Frequently asked questions
- Can we not just do this ourselves?
- Yes, in principle. But it tends to be the thing that never gets done: important, but never urgent. And it only really works once the whole team shares the same principles, not when one developer has a setup of their own. I have already done that groundwork, so your team does not have to.
- Do we really need an AI workflow of our own?
- Not one fixed workflow for everyone. What matters is a shared set of principles, with your architecture anchored in them. Within those principles the workflow is open: every team, and every developer, can fill it in their own way. That way AI works to your standard, and everyone keeps the freedom to work the way they want.
- Can AI not do this itself?
- I use AI heavily, including in the audit itself. But AI does not know your architecture and your conventions on its own. Someone has to decide what "good" is and write it down. That is where the work is.
- AI keeps getting better — can we not just wait?
- Better models get cleverer, but they do not pick up your architecture and conventions by themselves. And because AI amplifies what is already there, the gap between teams with a sound way of working and teams without only widens. Waiting means falling behind. Principles also work with any model: what you set up now grows with every new version.
- What do we actually get out of it?
- A prioritised list of bottlenecks and fixes, from high impact to low. With that you know exactly where you are losing time and what to tackle first. I do not promise hard numbers up front; those differ per team. What I do promise: after the audit you know where it goes wrong and how to fix it.
Find out where AI is going wrong for your team
Book an intro call — no obligations. Then we can look together at whether this fits what your team is running into.
€750 — and if I find nothing worth acting on, you do not pay.