The build got cheap. The judgment did not.

Earlier this year I spent a day at the Vlerick Business School M&A Conference. Two tracks that seemed to have nothing to do with each other, buy and build on one side, AI on the other, and by the end of the day they were telling the same story. I am not an M&A expert, I was there as someone who advises founders on pricing and growth and gets asked, more and more often, “and what about acquisitions?” So this is what I took home, written down before it fades.

The buy is not where the value is

Mieke Verstraeten of Group Induver, on stage with Sjarel De Bondt, told the story of a Flemish insurance broker that grew faster than most. In 2024 they had 120 people, today 520. From 24 million in revenue to 90 million, with HG Capital as PE partner and 23 acquisitions in two years. And still, the buying is not where they make the difference.

They started with full integration from day one. Too fast, it turned out, so they flipped it. Now there are four things that are not negotiable and happen immediately: compliance, cybersecurity, commercial reporting and financial reporting. The rest is a menu. Keep your own brand? Fine. Run your own HR? Also fine. The North Star is clear, one integrated company eventually, only the pace and the order are flexible. Most buy and build stories I hear are either integrate everything on day one or leave everything standalone. This push and pull model sits in between, and it works, as long as you are clear about where you are going.

A detail that stuck. Induver used to communicate an acquisition as a merger as it sounds softer. It created trust problems with new employees and customers, so they changed the wording: an acquisition of shares and a merger of visions. Call a dog a dog. Honesty at the start prevents friction later, and that pattern came back all day.

The first generation does not work for you

Gert Bervoets of H.Essers, the family business in transport and logistics that went from 80 million to 1.1 billion in twenty years without an outside shareholder, half organic and half through acquisitions, was the most direct about it. A real entrepreneur, first generation, the person whose name is on the door, is not going to work in your structure. Not with your reporting lines, not with your governance, not with your processes. Second or third generation is a different story, they know the difference between owner and operator. Plan a transition of two years at most. It sounds harsh, but look back at the acquisitions that struggled and the cause sits exactly there.

The invisible cost

Roel Caers of Guardsquare gave the most practical talk of the day. Four M&A processes, two as seller, two as buyer. His most recent deal was a carve-out of 8.5 million, and external help cost about a million. At a deal of 15 or 20 million that cost would have been roughly the same, which is a proportion worth knowing before you start.

His real lesson was different. In their first deal, Battery Ventures in 2018, they did everything themselves. It worked, but the business suffered for three to six months, strategic work stayed on the shelf, and the 2019 numbers felt it. In their later deals they used an external banker, external lawyers, an external PMI consultant. More expensive, yes. But the company kept running, and that is the bill that counts in the end. It is the same lesson Omar Mohout gave me from the M&A side of the table a few weeks ago: the real cost of running your own deal is not the fee you save, it is the year-one forecast you miss while doing it. You only sell once.

One more of his rules I want to keep next to Omar’s point about investors who negotiate every day: mirror the structure of the other side. If they bring a banker, an external lawyer and a project manager, you bring the same. Without that you have no umbrella to shelter behind, you show your cards too early, and the relationship with the other party suffers for it.

Alignment you cannot fake

The strongest point of the day came from Induver again. Partners who sell their company have to reinvest part of the proceeds in the holding, not stay a local shareholder but own a piece of the whole. The result: 15 percent of all employees are shareholders. When a new acquisition is announced, people are enthusiastic instead of afraid of the workload. You cannot fake that.

A reasonability check

Then the other track. Scott Moeller, ex Morgan Stanley and now professor at Bayes Business School, came with what he called a reasonability check, a slide of ten reasons why AI may be overrated. The first line on it: “AI lies, not just ‘hallucinates’.” It gives an answer that looks reasonable, and when it has none, it invents one that looks reasonable, references to papers that do not exist included. Then a copycat that cannot imagine, no killer app yet, capabilities oversold. And in the same talk: JP Morgan invested 2 billion dollars in AI, half their staff uses it daily, and due diligence without it is barely thinkable anymore.

I will be honest, it was not the session that stuck. It felt like watching from the academic stand while the pitch had already moved. But his last reason on that slide turned out to be the one that mattered, and I come back to it at the end.

The chatbot is dead

Paul Geertsema, associate professor at Vlerick, drew the line most people have not drawn yet. There are chatbots and there are agents. A chatbot talks: you ask, you get text back, advice, an email, a summary. Cheap talk, he called it. An agent does: it runs on your machine, has access to your files, executes tasks. Install software. Build a model. Write a report.

He described what had just happened elsewhere at that moment: sixteen AI agents together built a C compiler, normally a project for thirty people over a year or two and millions in cost, for about 20,000 dollars in a few weeks. With intensive human steering, and that is the nuance. Then he showed an agent building a full M&A merger model in seventeen minutes. Twelve tabs, DCF, price-to-book multiples, comparable deals. His warning came in the same breath: he would not trust it, and he would not advise us to either.

Not because it was bad. Because you have to understand the model before you use it. The bottleneck has moved, from building to understanding.

The productivity numbers Moeller quoted, an efficiency gain in M&A somewhere between 9 and 20 percent, fit that picture: not nothing, not the revolution you hear in podcasts either. The real gain is not speed, it is capacity, the same people can do more, but only if they understand what the AI is doing. A contractor Geertsema spoke to summed it up: AI makes him ten times more productive, three times he keeps for himself, three times he gives to the company, everyone happy. But that is someone who has been working with agents for months.

That is the crux. The people getting value from agents now are the ones who started six months ago. It is like piano, Geertsema said, you cannot just start. His advice: find someone in your organisation who feels like it, give that person one afternoon a week, no course, no consultant, just time to play. If your competitor starts in two years, they are two years behind, and catching up is hard.

People, processes, systems

Pieter Casneuf closed the conference with the most uncomfortable session of the day, in Dutch, without slides you could forward. One of his trades is preparing family businesses for an exit, doing the transformation and then leaving. His order never changes: people first, then processes, then systems. Most companies start with systems when there is a problem, new ERP, new CRM, and that is the last step. Get the talent right, then the processes, and the stronger the team, the fewer processes you need. His travel policy is one sentence, spend company money as if it were your own. His holiday policy is unlimited, for thirty years, never a problem.

Then the AI part, and the room went quiet. Three weeks earlier, he told us, his board had approved a plan to cut half of the development team and double the output. Not a forecast. A founder who called him that weekend, back in student mode, had spent the whole weekend programming in Cursor and built a data lake an external team had taken a year over. Alone.

His broader claim was harder still: we are on the eve of endless scalability at no cost, cognitive and manual, white collar and blue collar. IBM crashed on the stock market over COBOL, the cash cow nobody under retirement age still knows, now refactorable by AI. His prognosis for due diligence: legacy IT projects with a lead time of two years or more become a risk factor, not an asset. Buyers will ask why you do not ship every two weeks.

The tools get cheaper. The judgment does not.

Put the two tracks next to each other and there is one thread. AI makes the building cheap. Models, analyses, code, documents, the cost of producing them is heading to zero. And on the buy and build side the same thing was true all along: the buy is the easy, expensive, replicable part, the build is where value is created or lost.

What gets more valuable is the judgment about what to do with it. The question you ask, the context you bring, the nuance you catch. That is where the last line of Moeller’s slide lands, and it was the best thing he said all day: “AI is overrated because we tend to underrate humans: people really are versatile, talented and multifaceted.” Geertsema and Casneuf confirmed it from practice, and with far more conviction.

The most honest observation of the day: the apprenticeship is flipping. Juniors used to learn the trade by doing the boring work. Now the AI does the boring work, and seniors learn from juniors how to use it.

Three things I take with me

1. The difference between a buy and build that creates value and one that only creates PowerPoint value sits in the integration, in the build, not the buy.
2. Honesty costs nothing and saves everything, in communication, in expectations, in what you call a deal.
3. And the invisible cost, of M&A and of AI, is not the invoice from your advisors. It is your attention. Every hour on the deal, or on the tool, is an hour not spent on your customers.

Vlerick Business School M&A Conference, 2026. My personal notes for future reference.

---

Preparing your company for a deal and wondering what pricing does to the multiple? Let’s talk.