The Best Advice I Ever Gave Didn't Make Me Any Money

The best advice I ever gave a client cost me a project. They wanted to build a custom AI system; I told them not to. As an AI consultant in the Netherlands, I am supposed to sell builds — and I talked them out of one. We did not sign anything that month. Two years later, that same conversation has sent me more work than any proposal I have ever written. So no, the advice did not make me money. Not directly. It made me something more useful: trusted.

I want to tell you about a few of those moments, because they shaped how Crux Digits works, and because I think the willingness to say "do not build this" is the single most underrated quality in an AI partner.

The day I told a client to use a €30 tool instead

A mid-sized firm came to us wanting an AI document classifier. They had budgeted for a real project — the kind that pays well. We sat down, mapped what they actually needed, and it became obvious within twenty minutes: an off-the-shelf tool with a sensible rule set would do ninety percent of it for almost nothing. The remaining ten percent was not worth a custom build yet.

I could have built it anyway. Plenty of consultancies would have. Instead I said: "Spend thirty euros a month on this, try it for a quarter, and call me when you hit its limits." They were surprised. A little suspicious, even — people are not used to a vendor turning down work.

They called back fourteen months later. They had outgrown the tool exactly as predicted, they knew precisely what they needed, and they did not get a second quote. That is the thing nobody tells you about honest advice: it does not lose the deal, it delays it — and it removes every competitor from the room when it lands.

Why "do not build this" builds more trust than any pitch

Trust is the real currency in AI consulting, and it is earned in the moments where your interests and the client's diverge. Anyone can recommend the expensive option that happens to pay their invoice. The signal that you can be trusted is when you recommend against your own short-term interest and the client can see you doing it.

There is a reason "slideware" is a dirty word among the technical buyers we work with. A polished deck arguing for a big build tells the client what the consultancy wants to sell. A blunt "your data is not ready, fix that first" tells them what they need to hear. The first wins a contract. The second wins a relationship — and relationships send referrals for years.

This is also just good engineering. The best first project is rarely the most ambitious one; it is the smallest thing that proves value and builds the foundation the ambitious thing will need. Saying so out loud, even when a bigger scope would pay more, is how you end up with systems that actually ship instead of proofs of concept that stall. We wrote about scoping that honestly in our guide on AI implementation.

The advice that costs you a project and earns you a market

Here is the pattern I have watched repeat. You give someone the honest answer. You lose the immediate sale. And then three things happen that no marketing budget can buy.

• They remember you as the one who was straight with them. When the real, build-worthy problem arrives — and it always does — you are the first call, not one of three quotes.
• They tell other people. "Talk to him, he will tell you if you actually need it" is the best referral a consultant can get, and you only earn it by occasionally talking someone out of spending.

• You stop wasting your own time. Projects that should not exist are miserable to deliver. Killing them early protects your team's energy for the work that matters.

None of that shows up on the invoice for the conversation itself. All of it shows up on next year's.

The proof of concept I refused to extend

Another one stuck with me. We had built a proof of concept for a logistics company — a forecasting model that worked, technically. The numbers were promising. The client was excited and wanted to jump straight to a full production rollout, with the budget to match. On paper, that is the dream: a happy client asking to spend more.

I said no. Not "no, never" — "no, not yet." The proof of concept had worked in a clean test environment, but their live data pipeline was held together with manual exports and a spreadsheet someone updated by hand on Fridays. Push a production model onto that foundation and it would drift, break, and quietly lose their trust within a quarter. The honest sequence was to fix the data engineering first, unglamorous as that is, then scale the model onto something solid.

It pushed the big invoice out by months. It also meant that when we did scale it, the thing actually held — and they have since expanded it twice. A production AI system is not a project you finish; it is something you operate, and operating it on shaky data is how good models get blamed for bad plumbing. Saying that early, even when it delays your own payday, is the difference between a system that lasts and a demo that quietly dies.

What this means for choosing an AI consultant

If you are a business in the Netherlands weighing up an AI consultant, here is the test I would apply, knowing how the work actually goes. Watch what they say no to. A partner worth keeping will, at some point in the first few conversations, tell you not to do something you asked for. They will say the data needs work before the model does. They will point you at a cheaper tool when a cheaper tool is the right answer. They will scope smaller than your ambition because they want the first thing to succeed.

The ones who say yes to everything are the ones to worry about. Endless enthusiasm usually means they are selling certainty they have not earned — and you will pay for that certainty when the build meets reality. If you want to see how we think about this trade-off in numbers, our transparent pricing is public for exactly that reason: an audit comes first, deliberately, so the build only happens when it should. And our case studies are the short version of what gets built when the honest answer is "yes, this is worth it."

This is not a knock on ambition. We build big, hard systems for clients all the time. The point is that the ambition has to be earned by a clear problem, ready data, and an owner who will run the thing — and a good consultant tells you when those pieces are not in place yet, even if it means a smaller invoice today.

The honest close

The best advice I ever gave did not make me any money, and I would give it again tomorrow. It cost me a project and bought me a reputation, which turns out to be the better trade by a wide margin. If you take one thing from this, take this: when you are choosing who to trust with your AI work, the most valuable thing a consultant can say is "you do not need this yet." Find the partner who will say it.

Book a free consultation and I will tell you honestly whether your idea is worth building — or whether you would be better off spending the money elsewhere. Either way, you will leave the call knowing.

Frequently asked questions