We have seen it play out across organisations in Europe. The tools are fine. The models are impressive. The pilots look promising. And then nothing moves, not the revenue line, not the operating model, not the P&L.

The bottleneck is not the AI. It is the organisation running it.

We have a name for what gets in the way. We call it Organisational Incompetence. And if you are a leader responsible for making AI actually work, this is the article you need to read.

Before we get to the diagnosis, let us ask you something.

How many of these does your organisation already have?

• Claude, Cursor or Copilot rolled out company-wide
• A scalable AI platform and infrastructure
• Engineers running multiple coding agents
• Sales, Marketing and Customer Service building their own agents

If you said yes to most of those: good. Now the harder question: has it really moved the needle? Have you seen return on investment? Do you actually feel faster, better, stronger as an organisation?

That hesitation before you answer is the gap we want to talk about.

Think about an ice rink. A few people have figured out how to skate — they are flying. Individual contributors, specific teams, certain departments. Fast, productive, visibly impressive.

But the organisation around them? Still walking. Still in shoes.

What happens when you put elite skaters in a rink full of people learning to stand up with skates? The fast ones get frustrated.

The slow ones feel left behind. Tempers fray. And in the heat of that friction, someone will eventually say it out loud: that the people who have not kept up are falling behind. But that is the wrong conclusion.

The problem is not the people. It is that the organisation put everyone on a new surface without asking whether they had the right equipment, the right support, or even enough time to learn.

Nobody is actually going anywhere together.

That is what AI did to most organisations. The surface changed.

Not everyone got skates.

And even the ones who did cannot move freely when everything around them is still designed for a different surface.

We saw this in the data too. In software engineering, AI has dramatically reduced the time spent on coding: from roughly 65% of development activity to around 25%. That sounds like a revolution. But the work did not disappear.

It moved. To code review. To QA. To rework.

The bottleneck did not go away. It just changed location.

If nothing else in your organisation changes, your processes, your decision-making, your operating model, a faster coding team just creates a faster backlog somewhere else.

If the rest of the delivery system stays the same, AI helps teams produce code faster than the organisation can review, integrate and absorb it.

AI doesn’t fix dysfunction. It scales it.

This is the uncomfortable truth we keep seeing with clients across Europe.

AI amplifies. Point it at a broken process and it runs the mistakes faster. Wrong decisions scale instantly. Silos get automated in place. Bad habits become industrialised.

AI masks. You deploy the tools, the dashboards go green, leadership feels like transformation is happening. But there is a gap, sometimes a vast one, between activity and actual value. We have seen organisations where the key performance indicator for AI adoption was the number of tokens burned. Of course people found ways to burn tokens. That is not transformation. That is performance — for the board deck, not for the business.

AI splits. A small group figures it out and races ahead. Everyone else falls behind. Instead of bringing people along, organisations start creating an AI elite and quietly writing off the rest. That is not a strategy. That is abdication.

AI is an amplifier. It compounds whatever it is pointed at. If your organisation is well-led, well-structured and genuinely learning, AI will accelerate all of that. If it is not, AI will accelerate that too.

The thing nobody says out loud

We use this term deliberately. Not to insult anyone. To name something that everyone sees and almost nobody says out loud.

Organisational Incompetence (OI) (n.):
A persistent, systemic inability to turn intent into coordinated action and reliable outcomes.

This is not about unintelligent people. It shows up as unclear ownership, conflicting priorities, slow or avoided decisions, weak feedback loops and leadership that does not create clarity, resolve trade-offs or own outcomes.

The root cause is the gap between where an organisation believes it is and what it can actually deliver.

Everyone sees parts of it. Few name it. Fewer accept it. Fewer still stay with the work long enough to fix the fundamentals.

And here is what makes it particularly dangerous right now: AI does not remove this gap. It exposes it. Faster. At scale. In front of customers, internally, boards and competitors.

What our consultants told us. None of it is surprising.

We asked our AI consultants across Europe, people embedded in real AI assignments right now, what they see as the actual blockers. Not what the vendor decks say. What actually stalls progress.

The top organisational challenges they named were:

1. Leadership and strategic misalignment
2. Skills and capacity gaps
3. Functional silos and poor knowledge sharing
4. Slow decision-making and bureaucracy
5. Cultural resistance and low AI literacy

Look familiar? These are not new problems. They existed before AI. AI just made them more expensive and more visible.

Then we asked a different question. What single fundamental needs to be in place before you scale AI?

The answer was nearly unanimous: C-suite sponsorship and budget. Before the tools. Before the platform. Before the strategy document. Leadership commitment.

Read those two answers back to back.

The number one challenge and the number one prerequisite are the same thing. That is not a coincidence. It is a pattern.

Hey leader, are you the blocker?

Even the so-called technical blockers, data quality, legacy systems, fragmented infrastructure, trace back to organisational decisions. Who owns the data? Who funds the governance? Who resolves the conflict between departments? Those are leadership questions, not engineering questions.

The one thing the maturity models keep forgetting

We started with an AI maturity model to help organisations assess where they actually were.

The first version covered five dimensions: strategic alignment, technology foundation, governance, delivery flow, and people and culture.

It looked comprehensive.

It wasn’t.

Leadership and sponsorship were missing.

At first, we treated leadership as something around the model — important context, but not a dimension in its own right. Our work with clients and our conversations with AI consultants changed that view.

Leadership was not just another factor. It was the condition that determined whether the other dimensions could move.

So we added a sixth dimension: leadership and sponsorship.

Without leadership and sponsorship, an organisation cannot expect to get far on its AI journey.

Every dimension matters. But one determines how far you can get on all the others.

Leadership.

Here is what the progression looks like in practice:

• L0: No sponsor. Leadership disengaged or sceptical.
• L1: Lone experiment. No funding, no belief. Leaders are often misaligned, with competing agendas.
• L2: Sponsorship exists but is inconsistent. Risk appetite varies by leader. There is no unified direction.
• L3: A real C-suite sponsor. Protects the risk appetite. Resolves conflict between business units. Ties incentives to adoption.
• L4: Sponsorship is distributed and aligned across the organisation. AI literacy is a leadership succession criterion. The organisation bets on AI.
  

Which of these descriptions would your leadership team recognise, honestly?

Most organisations we encounter are somewhere between L1 and L2. They have tools, experiments and isolated champions, but not yet the consistent leadership commitment needed to scale them.

And they wonder why adoption stalls.

3 months in Cologne. 3 phases that changed everything.

Florian spent several months working with a mid-size software company in Cologne. They build charging systems. When he arrived, they were at roughly L2: some willingness to experiment, especially in engineering, but no clear mandate and significant variation between leaders.

They had already bought 2 tools: Claude Code for the engineers, and Langdock for the broader organisation. The question was: what now?

Phase 1: stepping on the ice.

The assumption was that people would figure it out alongside their regular work. Some did. AI champions slowly surfaced: curious people who had been using tools in their own time and were willing to lead informal onboarding. Workshops started. Everyone got a basic Claude introduction.

And then reality set in. Some people got genuinely fast. Others were afraid of losing their jobs. And some were frustrated: engineers producing more code than their colleagues could review, because review had not been reimagined alongside coding. The ice rink problem, live.

Phase 2: getting skates and learning to use them.

Leadership made a decision. A real one. They accepted that there would be a delivery dip. They mandated AI exploration tickets: time explicitly protected for developing AI literacy during working hours. They communicated a company-wide picture and kept working on it even when it felt uncertain.

Critically, leaders had to walk the talk. Not just say “use AI.” Actually use it themselves. The head of engineering became one of the highest AI users in the organisation. Leaders who had been far from operational work found that AI brought them back into it. That change in leadership behaviour was unexpected and powerful.

Phase 3: sharpening the skates.

With the foundation in place, the organisation made structural changes. They got rid of Scrum for some teams. Removed sprints and refinement cycles where AI could handle the scaffolding instead of humans. They built what they called CompanyOS: a central, machine-readable knowledge hub containing everything about the products, teams and processes. A shared organisational brain that agents could actually use.

The journey from phase 1 to phase 3 took months of deliberate sequencing. The operating model change only worked because the people and leadership foundations came first.

Organisational Incompetence + Artificial Intelligence means your existing dysfunctions: the unclear ownership, the misaligned leadership, the silos, the avoided decisions, are now running faster, at scale, with a budget behind them. Mistakes accelerate. Gaps widen. The workforce splits in 2. You spend more and move less.

Organisational Intelligence + Artificial Intelligence means your ability to learn, align, decide and act together is now amplified by the best tools ever built for the purpose. You move faster because the organisation underneath is actually ready to move.

Same technology. Same ambition. Same market pressure.

The preparation is what changes everything.

Honestly: which one are you actually on right now?

What separates the organisations that make it

The organisations that get real return from AI are not the ones with the best tools or the biggest budgets. They are distinguishable by a small number of things that cannot be bought off a shelf.

They know where they actually are. Not where the strategy deck says they should be. Not where they were planning to be by now. Where they are. Most organisations assess themselves against aspiration, not reality. The gap between those 2 points is the real starting point, and it will not close by pretending it does not exist.

Do the assessment. Name the Organisational Incompetence honestly. Then do the work of turning it into Organisational Intelligence. If you mask the gap instead of closing it, AI will not transform your organisation. It will just run your current reality faster. And faster dysfunction is not transformation. It is just more expensive dysfunction.

They account for the full cost. Not just licences and compute. The harder costs are organisational: the delivery dip while people learn, the change management that nobody scoped, the reskilling investment, the weeks where a team slows down because they are rebuilding how they work, not just what they use. And then there is the human cost. People who have spent years building deep expertise are watching AI agents replicate it in seconds. That is not just a workflow problem. It is an identity problem. The organisations that make it do not dismiss that. They answer it.

They treat people as the point, not the footnote. The Cologne case shows this clearly: the tools were bought in month 1. The real transformation happened when leaders started walking the talk, when AI champions were found and empowered, when the emotional journey of the workforce was taken seriously alongside the technical one. The organisations that create an AI elite and quietly leave everyone else behind have not transformed anything. They have built a faster version of their existing inequality.

Their leaders go first. Not in the strategy document. In practice. The head of engineering in Cologne became one of the highest AI users in the building. That single behavioural shift changed what was permissible for everyone else. Leadership that mandates AI adoption without visibly participating in it creates compliance, not transformation.

These are not complicated conclusions. They are just the ones most organisations are not yet willing to act on.

The question nobody asks before buying the next tool.

Most AI strategies describe a destination. Almost none describe the organisational condition required to reach it.

That is the gap. And it is expensive.

You cannot deploy AI into a structure that cannot make decisions. You cannot scale something your organisation is not ready to receive. You cannot automate your way out of misalignment, or buy your way out of unclear ownership. The tools will arrive. They will be impressive. And then they will sit in the exact shape of the organisation that deployed them, for better or worse.