63 Percent Are Already In: Where Does Your Company Really Stand?

63 percent. This number has been showing up for the past few weeks in practically every board meeting, every investor conversation, and every LinkedIn feed that has anything to do with digitalization. It comes from the study "Unlocking AI Potential in Germany 2026" by Strand Partners on behalf of AWS, which surveyed 1,000 executives and 1,000 members of the public in Germany – part of a Europe-wide survey. The result sounds unambiguous: 63 percent of German companies already use artificial intelligence, a marked jump from 53 percent the previous year and clearly above the EU average of 54 percent. 83 percent of AI users report measurable productivity gains, and 91 percent expect AI-driven growth over the coming year.

If you're reading these numbers for the first time, they can trigger two reactions. Either a relieved "Good, so we're not the last ones" – or a quiet unease, because you know exactly that your company has so far done little more than have a few employees use ChatGPT for emails. Both reactions are understandable. Both are also, as it turns out, not quite accurate based on the headline alone. Because the truly interesting question isn't "Are we using AI or not?" but: where exactly does your company really stand – compared to businesses your size, not to the average across corporations and micro-businesses?

63 percent: a figure that covers every company size – and can mislead precisely because of it

The first important point, lost in most headlines: the 63 percent figure from the AWS/Strand Partners study is weighted by company size, industry and region, and thus represents the German economy as a whole – from the publicly listed corporation to the ten-person trade business. It's representative, but it isn't Mittelstand-specific. And that's exactly where a gap opens up that other, independently conducted studies make clearly visible.

The ifo Institute published its own survey in May 2026 and arrives at an overall figure of 54.5 percent AI adoption in Germany – already noticeably below the AWS study's 63 percent. What matters more, though, is the breakdown by company size: large enterprises sit at 67.2 percent, while the classic Mittelstand sits at only 47.2 percent. Bitkom puts active AI use at 41 percent for 2026, with another 48 percent still only planning it. One survey focused specifically on generative AI in the Mittelstand cites just 38.7 percent adoption – and, more remarkably, only 4 percent company-wide embedding. This last study is based on only 134 companies and should therefore be treated with caution statistically, but the pattern matches the larger surveys: the smaller and more mid-sized the company, the lower the actual AI penetration.

What does that mean for you? If your company belongs to the classic Mittelstand and you're wondering why you don't see yourselves in the 63 percent – the answer is simple: you shouldn't be comparing yourself to that figure at all. The more honest benchmark sits closer to 41 to 47 percent. That changes the starting position considerably, but it also takes some of the artificial pressure off that flashy headlines tend to create.

The second, more important look: usage isn't the same as depth

Even more revealing than the size question is a second finding from the same AWS study that gets lost in most reporting: of the 63 percent of companies that use AI at all, only 15 percent deploy it transformatively – meaning with advanced systems or agentic AI that independently takes over tasks. 57 percent operate at a basic usage level: public chatbots, off-the-shelf standard solutions, isolated applications without deeper integration. Another 28 percent sit in between, at a mid-level integration stage.

What's notable isn't just the low share of transformative users, but the direction: the previous year, that share stood at 21 percent – so it has fallen, not risen. While more companies overall are coming into contact with AI, the share of those going truly deep is shrinking at the same time. A plausible explanation: the broad mass of latecomers is only just getting started with simple tools, diluting the share of advanced users without those advanced users themselves having become fewer.

The real divide doesn't run between companies that use AI and those that don't. It runs between the 15 percent who have genuinely built AI into their processes, and the 85 percent stuck somewhere between experimenting and surface-level usage.

For your own self-assessment, this is the decisive question. "We use AI" can mean: a few colleagues type requests into a public chatbot. But it can also mean: AI agents independently handle customer inquiries, draft quotes, maintain your CRM, and run around the clock without anyone needing to intervene manually. Between these two poles lies a difference that shows up in real business numbers – not just in whether a license is active.

Agentic AI: the real dividing line between pioneers and the rest

Within the 15 percent of transformative users, there's an even sharper line: agentic AI, meaning systems that independently complete multi-step tasks rather than just responding to individual prompts. According to the AWS study, only 22 percent of companies have even heard of it – and only 4 percent have fully implemented agentic AI. That's a surprisingly small number given how much media attention the topic gets.

Things get interesting when you look at the results among companies that have already taken this step: 92 percent of them report productivity gains, compared with 83 percent across all AI users on average. 48 percent record higher revenues, compared with 34 percent on average. 96 percent expect stronger growth over the coming year, compared with 91 percent overall. The gap might look moderate at first glance, but it shows a clear pattern: those who go beyond the chatbot and genuinely embed AI into processes achieve above-average results – consistently, across multiple metrics.

This trend is measurable in the Mittelstand too, even if at a lower level: the Salesforce AI Index Mittelstand 2026 shows that AI agent adoption in the Mittelstand rose from 8.7 percent in 2024 to 16.6 percent in 2026 – nearly doubling. In parallel, another Mittelstand survey reports that 51.2 percent of mid-sized companies now use or at least test AI solutions, compared with 33.1 percent in 2024. So the picture isn't "the Mittelstand is hopelessly behind," but rather: the Mittelstand is catching up noticeably, though starting from a lower baseline and with a smaller group of pioneers.

For concrete use cases with documented ROI potential, industry surveys mainly cite three areas: quote generation, customer service and chatbots, and sales forecasting. In suitable use cases, provider studies describe a reduction in manual workload of 40 to 60 percent. Important for context: these ROI and cost figures – typical entry costs for specialized Mittelstand setups are, according to the same sources, around 3,000 to 10,000 euros one-time, plus ongoing support starting at a few hundred euros a month – come from vendor and marketing sources and haven't been independently verified. They're useful as a rough guide, not as a reliable market average.

The counter-trend: why over a million people are deliberately opting out of AI

An honest self-assessment also requires a look at the countermovement – because "more AI everywhere" is by no means the only story of 2026. In August 2026, the open-source project LibreOffice released version 26.8 and hit a remarkable record: 1,031,162 downloads in the first week from the official website alone – excluding Linux package sources like Flathub, Snap or PPA, which separately mean significantly more usage still. It was the highest first-week figure in the project's history, and for the first time there were two consecutive weeks with over a million downloads each.

What's notable is the reasoning LibreOffice itself gives for it. In a blog post dated September 3, 2026, titled "Yes, no AI is now a feature," the project explicitly promotes the fact that it deliberately forgoes built-in generative AI – as a contrasting offer to Microsoft, which has firmly integrated Copilot into Microsoft 365, and to Google, which embeds Gemini into Docs, Sheets, Slides and Drive. The causality of "download record because of opting out of AI" is an interpretation from industry observers and the project itself, not a causally proven study – but the sheer number of downloads is real, verifiable, and speaks for itself.

What does this mean for your own AI strategy? Not that AI is fundamentally dispensable – the AWS study's productivity numbers speak clearly against that. But that "AI everywhere" isn't an end in itself. Classic word processing in the back office is an area with demonstrably low AI necessity for many users. Sales, customer service and recurring back-office processes, by contrast, are exactly the areas with documented ROI potential. A mature AI strategy distinguishes between the two, rather than forcing AI indiscriminately into every tool that already works fine.

The real hurdle isn't the technology

If you're wondering why your company hasn't progressed further despite existing interest, the AWS study offers a revealing answer – and it has little to do with a lack of tools. 49 percent of surveyed executives cite a lack of AI or digital skills as the biggest hurdle for AI adoption. 39 percent of companies still don't have a dedicated AI budget at all. On average, 21 percent of the IT budget is currently earmarked for AI, with 86 percent of respondents expecting this share to keep rising.

These figures are notable for two reasons. First: the technology itself is clearly no longer the limiting factor. Powerful models, mature agent frameworks and specialized industry solutions have long been available. The actual bottleneck lies in competence and budgeting – questions that can be resolved far faster with the right preparation and the right partner than a technological breakthrough one would have to wait for. Second: whoever still has no AI budget today will, in a year or two, be competing not just with rivals who use some AI, but with those who already have agentic systems running productively – with the correspondingly higher productivity and revenue figures described above.

In practice, that means: the most sensible first step is rarely "buy yet another tool." A more sensible approach is an honest inventory – which processes currently tie up the most manual time, where does the team already have experience, and where is knowledge simply missing to even recognize existing opportunities. This is exactly where paths diverge in practice: companies that do this inventory and then invest specifically in competence and a first, clearly scoped use case move from basic to mid-level or even transformative usage within a few months. Companies that instead try out tools at random tend to remain stuck among the 57 percent of surface-level users – regardless of how many licenses end up in use.

Where does your company really stand? An honest self-assessment

Let's summarize the numbers once more the way they actually fit together, rather than treating them as a single headline. 63 percent of all German companies of every size use AI – but if your company belongs to the classic Mittelstand, the more honest benchmark is closer to 41 to 47 percent. Of all users overall, only 15 percent deploy AI transformatively, a share that has actually declined recently. Agentic AI, which delivers above-average results, has been fully implemented by just 4 percent. And the biggest hurdle on the way there is, in nearly half of cases, not the technology but a lack of competence and budget.

Applying these figures to your own company typically yields three possible positions. First: you belong to the roughly 40 to 55 percent of mid-sized companies that don't yet use AI at all, or only sporadically. The good news then is that you're in large company; the bad news is that the window for an easy, relaxed entry is closing steadily while the pioneer group extends its lead. Second: you belong to the 57 percent that already use AI at a basic level, say via a public chatbot – then the biggest lever usually isn't new tools, but integrating existing processes and making the leap from a single application to genuine automation. Third, the smallest group: you already deploy AI transformatively or even agentically – then you should read this group's above-average results as confirmation to keep consistently pursuing and expanding the path you've taken.

Whichever of these three groups you belong to: the evidence clearly shows that an honest assessment is worth more than a catchy percentage. 63 percent sounds like a race you've already lost. The more nuanced numbers, by contrast, paint a more realistic picture – one where there's still plenty of room for a well-thought-out entry tailored to your company, as long as you don't keep pushing it off indefinitely.