Looking at AI adoption in the DACH mid-market structurally, one dynamic stands out that is rarely named: companies don't learn AI – individual teams do.
At first glance, that doesn't sound like a problem. On the contrary: it looks like what we usually wish for. Bottom-up adoption. Initiative. Pragmatic employees who find and deploy tools themselves, without anyone having to launch a change programme.
This is exactly the most underestimated management challenge in the mid-market right now.
What is actually happening
If you look closely at a typical mid-market B2B organisation with 80 to 300 employees today, you usually see the following:
The marketing team is already working substantially AI-augmented. Content briefings, research, first drafts, competitive analyses, reporting – much of it runs through ChatGPT, Claude or specialised tools. The speed at which output is produced is significantly higher than twelve months ago.
In sales, individual employees experiment with automation: account research, personalised outreach, call preparation. Not systematically, but individually. Everyone has their own stack.
In operations and finance, quiet workflow automations emerge – usually wherever a single person has the curiosity and the courage to dig in.
And leadership sees the outputs. They see things getting finished faster. They see the language becoming more precise. They see that some employees suddenly seem more productive. But they have no complete picture of what is actually happening.
This observation is not an isolated case: studies – including the McKinsey Global Survey on AI 2024 – show that generative AI usage varies strongly by function, with particularly high adoption in marketing and product development, while other areas adapt far more slowly.
The problem isn't speed
The real problem doesn't arise because AI-capable teams are too fast. It arises because the speed differences within the same organisation grow larger than existing management structures can absorb.
Concretely, that means:
Marketing produces volume and quality that sales, in its current form, cannot process. Content gets created in weeks that used to take quarters – but the sales pipeline and the underlying sales logic haven't scaled accordingly.
Individual employees make operational decisions with AI support that used to be made two hierarchy levels higher – not out of arrogance, but because they can move faster. The leaders above them notice it but don't know whether they should welcome it or not.
Strategic decisions still take the time strategic decisions take. But the operational layer underneath has accelerated. This creates a growing gap between the speed at which output is produced below and the speed at which decisions are made above.
Why this isn't a leadership weakness
It is tempting to read this as a failure of leadership. "The leadership isn't AI-savvy enough." "Top management is lagging behind." Such statements come up in many consulting circles right now.
That is wrong. And it is unhelpful.
CEOs structurally have less time for deep tool exploration than someone who works with a marketing workflow all day. That isn't a bug, that's a feature of their role. Anyone who spends 40 hours a week with an AI system learns it faster than someone who, between customer meetings, board sessions and operational decisions, dips in for a few minutes three times a week. That is mathematics, not character.
What does happen, though: the gap between those who use AI daily and those who have to assess and contextualise its outputs becomes a structural risk.
The three tension fields
Three recurring tension fields emerge from this dynamic in practice.
First, a quality assessment problem. When a marketing team produces ten times as much output as a year ago, leadership can no longer review that output in the same detail as before. What usually happens is one of two things: either leadership gives up on detailed review and trusts blindly, or it tries to keep reviewing and becomes the bottleneck. Both are suboptimal.
Second, a governance vacuum. When individual employees introduce tools on their own and feed data into external systems, legal, technical and reputational risks emerge that nobody tracks systematically. Most GDPR violations that will occur in German companies over the next two years won't come from bad intent – they will come from someone trying to be pragmatic without knowing where the boundaries are.
At the same time, studies – including reports from BCG and Deloitte on AI adoption 2024/2025 – observe that employees frequently drive the introduction of generative AI independently, while formal governance structures lag behind.
Third, a strategy-operations split. Leadership still makes strategic decisions on a quarterly or annual cadence. But the operational layer now reacts in weeks. This leads to strategic direction lagging behind operational reality – or, conversely, to operational realities undermining strategy without anyone consciously deciding so.
What companies actually need
The right answer to this dynamic isn't "introduce more AI". Nor is it "introduce less AI" or "hit the brakes until we've caught up". Neither works.
What works is a deliberate framework that does three things at once:
Create visibility. An honest, current inventory: who in the company is already using which AI tools, for which tasks, with which data? In most mid-market companies this inventory doesn't exist – and just the act of creating it changes the dynamic.
Prioritise instead of forbid. Out of the many individual AI initiatives, two or three are selected for strategic support. The rest either continue under clearly defined rules or are deliberately paused. Leadership makes a decision about the portfolio – it doesn't try to control everything.
Establish a learning cadence. The speed at which tools and possibilities change makes one-off decisions worthless. What helps is a rhythm – stopping briefly every six to eight weeks, looking, adjusting. Not as bureaucracy, but as practice.
What this means for you concretely
If, as a CEO, you read this and feel that some of it describes your own organisation – that is normal. I see this dynamic in nearly every mid-market company that doesn't actively ignore AI.
The most important insight is usually not analytical but emotional: it isn't your fault that this is happening. It is a structural consequence of a key technology seeping into organisations faster than organisations can adapt their structures. Past technology waves looked similar – the introduction of the internet in the late 1990s, cloud in the 2010s. What's different this time is the speed.
What you can do as leadership isn't "get better at AI". It is: create a framework in which speed differences between teams are used productively, instead of becoming destructive.
Structural observation from market analysis and 18 years of B2B marketing experience in the DACH mid-market. More in the CEO Guide to AI.