There's a moment that happens in a lot of Belgian companies right now. The board, or a big customer, or a competitor's press release puts pressure on the leadership to "do something with AI." And the instinct — the trained, sensible-sounding instinct — is to commission an AI strategy.
I want to argue you should resist that instinct. Not because strategy is bad, but because for a 30-to-200-person company, "we need an AI strategy" is usually the wrong size of answer to the actual question. What you need is smaller, cheaper, and far more useful: a decision.
Why "strategy" is the expensive trap
An AI strategy sounds responsible. It's also where a lot of money goes to die.
Here's the pattern. A company feels behind on AI, so it pays a consultancy €40k–€80k for a strategy. Twelve weeks later it gets a thick deck: a maturity model, a vision statement, five "horizons," a transformation roadmap, some slides with hexagons. Everyone nods. The deck goes in a drawer. Nothing ships. Six months later the company is exactly where it started, minus the fee, and still under the same pressure.
The deck failed not because it was wrong but because it answered a question nobody actually had. The leadership didn't lie awake wondering about their "AI maturity horizon." They lay awake wondering: is there anything here worth doing, and if so, which one thing first?
That's a decision, not a strategy. And decisions are cheap to make well and ruinous to avoid.
Strategy vs. decision, plainly
| A strategy | A decision | |---|---| | Asks "where could AI take us in 3 years?" | Asks "what's the one thing worth doing first?" | | Produces a vision and a roadmap | Produces a yes/no and an order | | Costs €40k–€80k and twelve weeks | Costs a fraction and three weeks | | Often ends in a drawer | Ends in something you actually build or deliberately don't | | Useful for a 5,000-person enterprise | Right-sized for a Belgian KMO |
A strategy is the correct tool when you're a large enterprise coordinating fifty teams. For most of the companies I work with, it's a category error — buying a cathedral when you needed a door.
What a decision actually requires
The good news: deciding well is mostly about asking four honest questions, in order, and refusing to skip ahead. You can run this yourself on a whiteboard. (It's also, not coincidentally, the spine of how I run the AI Opportunity Assessment — but the questions are yours to use either way.)
Question 1 — Where does the work actually hurt?
Not "where could we use AI." Where does your business bleed time or money today? The repetitive thing a smart person does for three hours every week. The back-office bottleneck. The thing customers wait too long for.
Start from the pain, never from the technology. AI applied to a problem you don't actually have is the single most common way companies waste money on it. If you can't name the pain in one sentence, you're not ready to spend on the cure.
Question 2 — Is AI genuinely the right tool, or just the fashionable one?
For each painful thing, ask honestly: would AI actually fix this, or would a better form, a fixed process, or a boring bit of automation fix it faster and cheaper? Sometimes the answer is "you don't need AI here, you need to delete this step." A good adviser will tell you that and lose the sale. Most won't.
This question alone saves more money than any other. The most valuable thing I sometimes tell a client is "don't build that."
Question 3 — What does it actually cost — including the parts no one mentions?
The licence or API cost is the small, visible number. The real cost is integration into your existing systems, the change management to get your team to use it, the ongoing run cost, and — increasingly — the EU AI Act compliance if the use turns out to be high-risk. (Recruitment and people-management AI almost always is; I wrote about how to tell.)
A decision made on the licence price alone is not a decision, it's a guess. Cost the whole thing or don't proceed.
Question 4 — What's the honest order?
Once you've got a shortlist of things that hurt, that AI genuinely helps, and whose full cost is known — rank them. Highest value, lowest risk, least disruption goes first. Not the most impressive one. Not the one the vendor is keenest to sell. The one that ships, works, earns trust, and pays for the next.
One good first project that lands beats five ambitious ones that stall. Sequence is most of the decision.
What you get at the end
Run those four questions properly and you don't have a vision. You have something better: a short, ranked list of AI projects worth doing — each with an honest cost, a risk classification, and a clear first move. Often the list is shorter than the leadership expected, and one or two "obvious" ideas have been honestly crossed off. That subtraction is worth as much as the additions.
That's a written roadmap of decisions, not a strategy deck. You can act on it next week, hand it to a developer, or use it to say no to the next vendor with a straight face.
Where this becomes the Assessment
You can do all of this yourself, and some companies should. Where WDC comes in is when leadership doesn't have the internal AI depth to answer Question 2 and 3 honestly — to tell, technically, whether AI really fixes the thing and what it'll truly cost to integrate and run.
That's exactly what the AI Opportunity Assessment is: a fixed-price, three-week engagement that runs these four questions across your business with senior engineering judgement behind them, and hands you the ranked, costed, AI-Act-classified roadmap in writing. €4,500, fixed — and if you decide to build one of the projects with WDC within 60 days, the whole €4,500 is credited against it. The decision pays for itself either way: worst case, you've spent a known, small amount to find out which expensive things not to do.
The bottom line
The pressure you're under is real, but it's pushing you toward the wrong-sized answer. You don't need a transformation. You need to know whether there's anything here worth doing and what to do first. That's a decision — and a good decision is cheap, fast, and the thing that actually keeps you out of the drawer.
I write a short, practical newsletter for Belgian businesses — real engineering, the EU AI Act, and AI integration, in plain language and without the hype. Subscribe below if that's useful.
And if you'd rather make this decision with senior engineering judgement in the room, that's what the AI Opportunity Assessment is for — start with a conversation.