AI advice from someone who builds the systems, and tells you what broke.
You do not need to be technical to make a good decision about AI. You need someone who sells no software, does no staffing, and has no stake in the outcome. I build the things myself first. Then I tell you what worked and what did not.
// I will help you understand the road ahead, but I do not build it for you.
Four situations where I am useful.
Seven AI pilots are running and none of them is finished
Every department started something. Nobody can say which of the seven makes the company money. The problem is almost never the model. It is that nobody wrote down the decision.
A vendor promises 40% efficiency and you cannot check it
You do not need to be an engineer to test that claim. You need someone who has built the equivalent system and who tells you which question to ask before the signature goes on the contract.
AI bolted onto a process that already rattles
An agent on an unclear process automates the unclarity, only faster and four hundred cases at a time. Process first, then the model. In that order, always.
Nobody on the board knows what it really costs
The licence is the smallest line item. In my own agentic sessions I measured that 85 to 98 percent of token spend was overhead. That number appears in no vendor deck.
Four shapes. All of them end in something written.
No timesheets without an outcome. Every engagement ends with something you can forward to your board, your accountant or your vendor.
AI reality check
A day at your table. We put your plans, pilots and vendor proposals next to what I have actually built and measured. You leave with a short memo you can circulate internally: what continues, what stops, and why.
Deliverable: Decision memo, 2 to 4 pages
Process before technology
We take a process that hurts and cut it open: where the real work sits, where the waiting sits, and which part is a candidate for automation at all. Half of it usually turns out not to need AI.
Deliverable: Process map plus a prioritised shortlist
Fractional CTO or CIO
You have no CTO on staff and you do not want one. A fixed number of days per quarter where I read along on architecture, vendor contracts and build versus buy, and where I say no when no is the answer.
Deliverable: Fixed day per month or quarter
Vendor and build due diligence
An independent second opinion on a vendor, an acquisition target or an internal build proposal, before the budget commits. I resell nobody, so I have no stake in the outcome.
Deliverable: Written opinion plus a question list
What I do not do: implementation contracts, staffing, or reselling third-party software. A scoped proof of concept to settle a decision is fair game; a team on-site for six months is not. I would rather state that now than after the first invoice.
I run my own company on a system of plain files. You are welcome to borrow it.
Six projects, deadlines, decisions and a time log, all of it in folders and text files on a laptop. No SaaS subscription, no vendor holding your knowledge hostage. I once moved the whole thing to a different machine and a different AI subscription in an afternoon, and lost nothing. That is the entire point.
What the light version is
- One overview screen with the projects, the deadlines and the next action per project
- Decisions with a date and a reason, so nobody has to reconstruct the why three months later
- Everything in plain files. No database, no vendor holding your knowledge hostage
- Works in dark and light mode, on a laptop, with nothing to install on a server
How you get it
It comes with the engagement. It is not a product with a price tag. After a reality check or a process round I set the structure up with your own projects in it, and explain how you keep it going yourself. After that you do not need me for it, and that is deliberate.
// Fair warning: this is not a Jira replacement and not meant for a team of fifty. It is meant for a management team that wants to know where the projects stand without buying another tool.
I have no vendor to defend.
Wilco de TreeOwner, ZeroNonsense.devDirector of Architecture & Software Engineering, Valona IntelligenceI have sat on your side of the table. That is the whole offer.
Sole proprietor. No partner programme, no commission. If the answer is that you should do nothing with AI, it costs me nothing to say so.
Thirty years in technology, currently Director of Architecture & Software Engineering at Valona Intelligence, where I own the platform architecture of two merged companies. I have sat on your side of the table.
I publish what went wrong too. An expertise agent scored perfectly on its own test, and that score went in the bin after I audited the test itself.
And I still build. A pipeline of 4.8 million verified rows from eight national registers plus the EU-wide E-PRTR, with lineage per row. Advice from someone who has not touched code in ten years is not advice, it is history.
Start with half an hour. No deck, no intake form.
You tell me what you are stuck on, I tell you whether I can be useful. If I cannot, I will say that too, and it will have cost you half an hour.