AI won't replace your team. Or your management.

This week, three CEOs and two PE partners asked me the same question: how do I scale with AI?

I gave the same answer every time. And every time, I disappointed them.

Because it’s the wrong question.

AI cuts the cost of execution. Not the cost of confusion.

For the past eighteen months, it’s been the same chorus in every board and exec committee I sit in: we need to AI-ify, we need agents, we need Claude Code, Cursor, n8n. As if the next release of some tool was going to fix ten years of sloppy steering.

It won’t.

AI cuts the cost of execution. It doesn’t cut the cost of strategic confusion. And it’s almost never execution that kills scale-ups. It’s what sits above it.

Any competent human — any properly deployed AI agent — can deliver inside a clean operating system: clear direction, honest KPIs, feedback loops short enough to catch drift before it becomes a fracture. If the frame holds, execution follows. Always.

Where things break is rarely on the ground. It’s upstream. When the business model doesn’t hold the plan you sold at the raise. When KPIs push one way and strategy the other. When monitoring catches the drift three quarters too late. There, no AI stack saves the company. AI just industrialises the problem — faster, at greater scale, with prettier dashboards.

A multiplier doesn’t turn chaos into excellence. It accelerates it.

Sortlist: what I understood six months too late

Sortlist went from 4 to 100 people in a few years, 9 countries, €10M ARR, three M&A deals. On paper, a story of marketplaces, SEO, matching algorithms. In reality, the story of an operating system that started cracking the moment we hit the gas.

We tried every tool. Different CRMs, different BI, different sales methodologies. Everything worked roughly fine as long as the global system held. Everything broke the moment it started slipping — regardless of the tools, regardless of the people.

The hardest moment of my career wasn’t the layoff plan after the Series B. That’s heavy, but it’s an operational decision: you make it, you own it, you move on. The hardest moment was realising, six months too late, where the root cause actually was.

We had visible people problems. The post-raise lure of equity had attracted mercenaries: brilliant CVs, in for the ride and the equity, not for the project. Middle management was young, ill-equipped to scale an organisation that had doubled in twelve months. All of that was true.

But none of it was the root cause. The real subject was pricing. We were on a Pizza Buffet model — a flat subscription, customers consuming at radically different intensities, and an economic equation that no longer tracked the growth curve we’d sold. The shift from a Pizza Buffet to a Paid to Value model should have happened before the Series B, not after. I told the story of how we found the problem in this article, and the forced repricing that ended up saving the company here.

Once that equation was broken, the entire system became uncontrollable. Sales KPIs pushed reps to sign faster and broader — exactly the wrong direction. Monitoring caught the drift too late, because the pivot metric — the value actually delivered to each customer — wasn’t measured. Inside that frame, any team would have derailed. Junior or senior. Brilliant or average. Motivated or mercenary. Not because the people were bad. Because the playing field no longer made sense.

That’s the lesson I hadn’t yet digested at the time, and that I now repeat to every scale-up CEO who’ll listen: you don’t scale a company. You scale what it already is. If the equation is healthy, you scale healthy. If it’s broken, you scale the mistake. The more fuel you add — cash, talent, AI — the harder the drift becomes to recover from.

A multiplier corrects nothing

The 2026 picture is clear. AI agents now run alongside humans on entire processes. Output per head is shifting structurally. I have clients where a senior dev runs four Claude Code agents in parallel and ships ten times what he did two years ago. That’s not marketing. That’s the Git log.

But here’s the trap: a multiplier multiplies what’s already there. Not zero. Not noise. Not a broken business equation.

On an aligned company — clear purpose, clean KPIs, shared North Star, a team that knows why it gets out of bed — AI delivers x10. Sometimes x100. Massive and probably durable. On a company without purpose, without alignment, without a system, AI also delivers x10 — but on the drift. On the burn. On the confusion. You get exactly the same decline, simply faster, with a dashboard that flatters the board until the runway runs out.

AI doesn’t decide why your company should exist. Anthropic didn’t outsource its safety mission to an LLM; Dario and Daniela Amodei laid it down, fought for the legal structure that protects it — Public Benefit Corporation with a Long-Term Benefit Trust — and they fire the behaviours that betray it. Compare that to OpenAI, whose mission has been rewritten six times in nine years, with the word safely finally dropped during the for-profit repositioning. You can prefer one or the other — that’s beside the point. The point is elsewhere: a purpose reveals itself in the legal structure, in the roadmap, in who you keep and who you let go. Not on the homepage.

AI also won’t define the values that make you turn down a toxic client the month you need cash. It won’t feel the urgency to protect a team when an investor pushes for a brutal cut. It won’t carry the existential responsibility of a project you’ve defended for five years. It executes. Inside a frame a human has set.

And above all, AI doesn’t create the mindset — that intangible texture that makes a team go the extra mile instead of ticking boxes. Mindset doesn’t come out of a prompt. It comes from a management team that transmitted a direction, recruited the right people, let go of the wrong ones at the right time, and held the line when it got hard.

The system that makes a company scalable, in three layers

If I had to compress in a single diagram what separates a company that scales from a company that scales its mistakes, it comes down to three layers — and the order matters.

One — Purpose. Not the slogan. The honest answer to: if we vanished tomorrow, who would sincerely complain, and why? The canonical example is still Apple — challenge the status quo. Not invented in 2002 to dress up “Think Different”. Already there in the product choices, in the refusal to compromise on design, in the way Jobs hired. The campaign just made visible what was already in the DNA. If the answer to the question is “our investors would lose money”, you don’t have a purpose. You have a cap table.

Two — Translation into a heading. Purpose translates into one number that everyone, from the CEO to the intern, reads the same way. That’s the North Star — a concept formalised by Sean Ellis, the man who coined “growth hacking”. The most pedagogical illustration is still Airbnb: their North Star is nights booked. Not GMV. Not user count. A booked night is a host who put up a quality listing, a traveller who trusted a stranger, an exchange that closed. The whole marketplace lives in that one number. Cruel test: if ten people in your company give nine different North Stars, you don’t have a strategy problem. You have a leadership problem.

Three — Execution discipline. Day to day, OKRs remain the most solid framework for aligning an organisation vertically. Invented by Andy Grove at Intel, brought to Google by John Doerr in 1999 when the company had 40 employees, still in place twenty-five years later. The framework’s name doesn’t matter. What matters is the discipline it demands — the discipline most exec committees don’t have: the ability to say no to 80% of the quarter’s good ideas, and the courage to grade Key Results honestly at the end (red zone at 70%, not at 30%).

When the three layers hold, here’s what happens: a junior salesperson in Madrid knows why this week’s deal matters for the Q3 product roadmap. A backend engineer in Brussels knows which business KR her sprint serves. Everyone rows in the same vector, because everyone sees the same heading.

It’s not a kickoff with a tray of croissants. It’s a system.

What PE funds are missing

To every PE partner reading this: your due diligence is biased. And the cost of that bias is rising.

You spend 80% of your DD time on market, unit economics, cap table, competition. You spend 20% on the team — and that 20% often boils down to checking exec committee CVs and a dinner with the CEO. In nearly every deal I’ve seen go wrong — including some of mine — the root cause was structural: a business model that didn’t hold, fragile unit economics, market assumptions that were too optimistic. But what turned a detectable problem into a disaster was the team in place: its capacity — or incapacity — to see the issue in time, to name it, to fix it before it became unrecoverable.

That’s where the delta sits between a company that pivots and a company that collapses. And AI is going to widen that delta, not close it.

Three tests we increasingly fold into the fractional advisory work I do with Belgian and French funds. None take more than a day. All are worth more than a month of financial modelling.

One — Audit the system, not the CVs. Ask to see last quarter’s OKRs. If they’re sleeping in a Notion no one remembers, you know. Ask a middle manager what the North Star is. If their answer diverges from the CEO’s, you know. Ask when the last underperformer was managed out. If the answer is over six months ago, you know.

Two — Hiring capacity above the current bar. A scale-up that can’t hire above itself is a scale-up that’s plateauing. Ask for the last five senior hires. How many stayed eighteen months? How many were at the level? Who personally ran the recruitment — a search firm, or the CEO? A healthy company has a CEO spending 25 to 30% of their time on hiring. Not on sourcing — on evaluation, closing, and onboarding.

Three — Transmission test. Sit down for an hour with three middle managers, without the CEO. Ask them what the company is trying to do in five years, and how their function contributes. Three coherent, energetic, angle-different answers: you have an asset. Three corporate answers that could apply to any company in your portfolio: you’re buying a shell with a nice deck.

Add to that a few weak signals that never lie. How the CEO talks about people who’ve left — with dignity or with bitterness. When the board prepares its meetings — the night before, or on the morning. At what level of the org bad news reaches the CEO — directly, or filtered through three layers. Management reads in those details, not on the org chart slide.

What I actually believe

The next decade won’t be won by the companies with the best AI stack. It will be won by the ones that built an operating system clear enough for humans and agents to deliver inside it without derailing — and that hired the humans capable of evolving that system as the market shifts. The rest is technology.

AI won’t replace your purpose. It won’t replace the conviction that makes you turn down a bad client. It won’t replace the manager who delivers bad news to her team with dignity. It won’t replace the strategic call made at 11pm after a day that ran too long. It will execute. Fast, hard, at scale. But the frame — direction, values, alignment, discipline — stays with you.

AI is a multiplier. The lever is the same as twenty years ago: a purpose that holds up, a heading everyone can read, real execution discipline, and humans who chose to fight for it.

Hire better. Manage better. Align better.

The rest, AI will handle.

Thibaut Vanderhofstadt is co-founder of Sortlist — the European leader in B2B agency-client matchmaking, present in over 140 countries. After twelve years as CEO and two M&A operations, he now advises SaaS founders, B2B scale-ups, and private equity funds on management, alignment, and unit economics through MetSaaS. Book a diagnostic →