The Sandwich Trap: Why Germany's Classic Mid-Market Falls Behind on AI
Published on 9/3/2026 · André Hellmann
AI in the mid-market is growing fast. 54.5 percent of companies in Germany use AI in their business processes, up from 40.9 percent a year earlier (Source: ifo Institute, 2026). One figure in the same survey breaks the expected pattern: mid-sized companies do not just trail large ones. They also trail small ones.
Discuss your next step in a free diagnosis call. Book a slot →
Contents
- Behind the corporations — and behind the small firms
- What the ifo figures show
- The sandwich trap: too big to improvise, too small for an AI team
- What hidden champions can do differently
- Getting started without risk
- Conclusion: the middle needs a different route
- Frequently asked questions about AI in the mid-market
- Sources
Behind the corporations — and behind the small firms
That large corporations lead on AI surprises nobody. They have budgets, data teams and departments built for exactly this.
The second half of the finding is the surprising one. Small companies use AI more often than mid-sized ones: 51.2 percent against 47.2 percent (Source: ifo Institute, 2026). The middle sits at the bottom.
This is not a measurement error or a rounding effect. It is a structural pattern — and it hits precisely the companies Germany exports as hidden champions.
What the ifo figures show
The ifo Institute tracks AI adoption among German companies through its regular business surveys. As of June 2026: 54.5 percent use AI in their business processes, up from 40.9 percent the year before (Source: ifo Institute, 2026).
Broken down by company size, the picture turns unusual.
Large companies reach 67.2 percent, small ones 51.2 percent, mid-sized ones 47.2 percent (Source: ifo Institute, 2026). The order is not linear — and that is the finding.
A second value from the same survey explains part of it. Only 18.7 percent of AI-using companies develop their own AI systems. Nearly three quarters use paid external solutions, 48.4 percent also use free applications (Source: ifo Institute, 2026). AI almost always comes from outside. The only question is how well it gets built in.
A free Self-Check shows where a company stands in a few minutes.
The sandwich trap: too big to improvise, too small for an AI team
Why does the middle in particular fall behind? The explanation lies in operating size, not in technology.
Small companies can simply start. Twenty people, short paths, no works agreement. Whoever has a good idea tries it on Monday. AI enters through individuals and spreads informally. That is messy, but fast.
Large companies can organize it. They have an IT department, a data protection team, often a dedicated AI unit. They build structures because they can staff them.
The middle can do neither. It is too big for the informal route — at 600 employees, “just try it out” stops working the moment customer data is involved. And it is too small for the structural route. A dedicated AI department would not be fully utilized, and the head of IT already holds two other full-time roles.
The middle does not fail on technology. It fails because nobody owns AI without dropping something else.
There is a further effect that numbers rarely show. The classic mid-market has working processes. That raises the barrier: a company with a well-drilled quotation process built over fifteen years will not rebuild it for a tool whose value nobody has quantified yet.
The underlying pattern is the Implementation Gap in its German form: not missing tools, but missing redesign.
What hidden champions can do differently
The good news sits in the same survey. If AI comes from outside anyway — only 18.7 percent build their own — then in-house development is not the bottleneck. The bottleneck is integration into the operation.
That is a capability, not a department. Four things make the difference:
- Pick a workflow, not a tool. Not “we are introducing AI” but “we are cutting turnaround time on tender reviews”. A workflow has a start, an end and a volume.
- Assign ownership without creating a role. One person from the business unit, not from IT, with a clearly capped time budget. Running AI requires knowledge of the workflow, not of the model.
- Build for production, not for a pilot. A pilot has to convince, an operation has to run. That difference is decided in construction, not in evaluation.
- Buy operating capability instead of building it. This is exactly the gap between too big and too small — and it closes with a partner, without founding a department.
The last point is the actual way out of the sandwich trap. A mid-sized machine builder does not need an in-house AI unit. It needs someone to run AI as a function — much as it does not write its own payroll software.
How that operation is structured is described in our engagement steps.
Getting started without risk
The most common mistake in the middle is an oversized first step. A year-long program with a steering committee generates effort before it generates insight.
The better route starts smaller: one workflow, six weeks, one number. If the number holds, the second workflow follows. If it does not, the stake was contained. Why we always start this way is set out in Getting Started Without Risk.
And since measurability decides everything here: how to prove the value cleanly is covered in the article on measuring AI ROI.
Which workflow fits a specific case is something we map out in the free diagnosis call.
Conclusion: the middle needs a different route
Germany’s classic mid-market trails both corporations and small businesses on AI (Source: ifo Institute, 2026). Not for lack of willingness, but because of a size for which neither the informal nor the structural route works.
The way out is unspectacular. Not more tools, not more strategy, but operating capability: a named workflow, an accountable person, one metric, a partner for day-to-day operation.
The alternative is predictable. AI adoption keeps growing across every size class. A company that is the only one not rebuilding does not lose overnight — it loses slowly, through quotation speed and response times that others undercut.
Frequently asked questions about AI in the mid-market
How many mid-sized companies use AI?
Mid-sized companies in Germany reach 47.2 percent, small ones 51.2 percent and large ones 67.2 percent. Across all sizes, 54.5 percent use AI in their business processes (Source: ifo Institute, 2026).
Why does the mid-market trail small companies?
Because both routes fit it badly. It is too big for the informal route and too small for a dedicated AI department. Small firms simply experiment, large ones organize. The middle is the wrong size for either.
Does a mid-sized company need its own AI team?
Usually not. Only 18.7 percent of AI-using companies develop their own systems (Source: ifo Institute, 2026). The bottleneck is not development but integration into existing workflows — which a partner can solve.
Where should a mid-sized company start?
With a recurring workflow that touches money and has a measurable volume: quotations, tender reviews, complaint handling. Six weeks is enough for a first defensible number.
What does getting started cost?
That depends on the workflow, not on company size. Which entry point makes sense in a specific case is something we clarify in the free diagnosis call.