What 550 hours of retyping a year really looks like
At a home furnishings retailer we worked out what retyping by hand costs. Four hours in a quiet week, twenty-six in a busy one. Here is how to work it out for your own company.
ArticlesWhat we run into along the way
No trend pieces. What we see on the shop floor at small and mid-sized companies, what works, and where we have watched ourselves miscalculate.
We do not build ERP. Why we advised a client against it
A client asked us to replace an order system that had been running for twenty years. Our advice: buy a standard package, and let us only do the layer on top.
Soon your customer will ask ChatGPT. Are you in the answer?
GEO is what SEO was ten years ago. The difference: an AI shows no ten blue links but one answer. What that means, and how to measure whether you appear in it.
Eleven out of eleven emails had no project number
Automation rarely fails on the technology. It fails on an assumption nobody checked. An example from a project at an events agency.
Two buyers, same car, five thousand euros apart
At a dealer group, two experienced buyers valued the same car five thousand euros apart. That is not a staffing problem. That is a process problem.
What 'we host in the EU' really means, and what it does not
Almost every AI supplier says your data stays in Europe. We said it too, and it was not entirely true. Four questions to ask before you believe it.
Six to seven minutes to find a wiring diagram. Per fault.
In technical companies the knowledge sits in the heads of the engineers, and they are ageing. At one service company, finding a single wiring diagram costs six to seven minutes on the phone.
AI tell or just bad writing? Eight things that give away that a text is not yours
Our own site scored 20 percent on anti-slop, the lowest of all the sub-scores. These are the eight patterns we started watching for, and why it costs you search traffic too.
We gave our own site a C. And this is what we are doing.
We scanned our own website with our own tool. Result: 55 percent, a C. Here are all eight sub-scores, what was wrong, and what we have done about it.
A digital colleague is not software. It is a colleague you have to train.
Building is the easy part. Getting people to use it is the work. What we learn from the tools that grind to a halt after three months.
Your new employee: 24/7, no salary, the odd big mistake
Just Eat Takeaway does not call its AI agents tools or models, but interns. That is the best explanation there is: quick and eager to learn, but with guidance, clear instructions and someone who checks the work.
The art of prompt engineering is dead, long live agentic workflows
Forget the prompt cheat sheets. The focus is moving from how you ask for something to how you set up a process, with digital colleagues that check each other's work and fix mistakes before you see the result.
Claude Design: your new visual assistant on the shop floor
With Claude Design, Claude now builds websites, presentations and dashboards directly in the chat. No expensive designer for a first draft, and what you make you export on to Canva to finish.
AI is the new natural gas. But who lays the pipes at your place?
When the Netherlands found natural gas, the big players fought over the source, but the heating engineer made every house ready one connection at a time. AI is no different: the money is in the connection, not in the source.
My visit to Uber EMEA: the power of AI clean-up
At Uber EMEA in Amsterdam, Daniel Melado showed how AI agents cleared out 500,000 lines of dead code. The lesson for SMEs: AI is not only there to build on top, but above all to clear out.
Why we scan the US and Chinese AI market every day
The US makes the hypes, China the execution. We scan both markets daily so that you do not get experimental software, but proven tools you can already use next week.
AI through a different lens: opportunities for SMEs vs. the political reality
At the book talk for The AI Matrix in Pakhuis de Zwijger, the focus was on the power of Big Tech and the risks to society. A solid counterweight, but abstract for the business owner who just wants his CRM fixed.
AI in SMEs: who carries the blame when it goes wrong?
Your team is experimenting with AI everywhere, but nobody knows who is responsible when it goes wrong. You do not get a grip on it with thick reports, but with five practical documents you can draw up in an hour each.
The 2026 shift: why you have to stop 'chatting' and start using 'agents'
Chatting with AI stayed manual work: you type, the bot answers. An agent gets a goal and carries out the tasks on its own inside your systems, and that is exactly why this is the breakthrough for SMEs.
Rather have a conversation than an article?
One conversation is enough. No obligations, no report. Just a straight answer on what AI is worth to you.
Book a free intake call