capinorium · Article · AI
The real barrier to AI in mid-sized companies: knowledge, not creativity
What holds mid-sized companies back on artificial intelligence is above all a lack of knowledge: in the key German surveys, 72 to 85 per cent of companies without AI cite this reason, while only 11 to 30 per cent cite a lack of ideas for applications. How high the figures turn out, however, depends heavily on whom a study asks.
The starting point seems clear. According to the Bitkom AI study of 14 September 2026, 57 per cent of German companies with 20 or more employees now use AI – a majority for the first time. Yet 87 per cent of these users say they exploit its potential only a little or not at all (Bitkom presentation, slide 9). Other studies measure usage rates from 20 per cent. For decision-makers the question is therefore: which barrier is real, and which figure holds up?
Knowledge comes first, ideas for applications trail far behind
Across all studies that survey non-users or companies actively considering AI, a stable ranking emerges. The ranges combine values from different surveys and should therefore be read only as orders of magnitude.
| Rank | Barrier | Range | Sources |
|---|---|---|---|
| 1 | Lack of knowledge / expertise | 72–85 % | Destatis, Bitkom |
| 2 | Law and data protection | 49–66 % | Destatis, Bitkom, BNetzA |
| 3 | Time and staff | 51–66 % | Bitkom, BNetzA |
| 4 | Costs cannot be calculated | 63 % | Bitkom |
| 5 | Integration into existing systems | 45 % | Destatis |
| 6 | Data (availability, quality) | 40–44 % | Destatis, Bitkom |
| 7 | Employee acceptance | 33 % | Bitkom |
| 8 | Lack of use cases | 11–30 % | Bitkom (11 % among users), Destatis |
The difference between cost level and predictability is revealing: according to the Federal Statistical Office (Destatis), only 32 per cent cite "costs too high", whereas 63 per cent of non-users in the Bitkom study cite costs that are hard to calculate. The problem is less the price than the lack of a basis for calculation. Self-assessment confirms the knowledge finding: 66 per cent of all respondents rate their employees' AI skills as low (Bitkom, slide 13).
Why the studies measure so differently
Three factors explain most of the discrepancies. First, the base of the question: Destatis only asks companies that have considered AI but not introduced it, and so arrives at 72 per cent for lack of knowledge. Statistik Austria uses the same European survey module but calculates on the basis of all non-users and ends up at 14.6 per cent, because 77.3 per cent have never considered AI. Even there, lack of knowledge remains the most common reason.
Second, the size threshold: some surveys start with micro-enterprises, Bitkom only at 20 employees. Third, definition and timing. The German development bank KfW measures 20 per cent usage for the period 2022 to 2024, Bitkom 57 per cent in summer 2026. Nor do the Bitkom figures of 41 and 57 per cent, often quoted against each other, contradict one another: the 41 per cent comes from the digitalisation survey of 11 March 2026, the 57 per cent from the AI study around six months later.
It is not size that holds companies back, but their knowledge base
Small companies use AI less often, but size hardly explains this. A KfW regression analysis of around 6,800 mid-sized companies shows that once graduates, digitalisation and innovation activity are taken into account, the differences by size are not statistically significant. According to KfW Fokus 554, the decisive factors are the degree of digitalisation, digital know-how and innovativeness. The official data fit this picture: at Destatis, lack of knowledge lies between 67 and 73 per cent in all size classes.
Once companies start, the barriers shift
Companies already using AI report a different profile. In the Bitkom study, lack of know-how falls from 85 per cent among non-users to 25 per cent among users, and lack of use cases from 30 to 11 per cent. Data protection, by contrast, rises from 49 to 66 per cent (slide 12). The comparison is a snapshot of different groups. It does not prove that companies acquired their knowledge through use; it is equally possible that knowledgeable firms were more likely to start. At the threshold from testing to operation, the Federal Network Agency (BNetzA) reveals a further barrier: 58 per cent of those who abandoned AI projects cite failure to meet their objectives.
The legitimate objection: are use cases not missing after all?
There are dissenting voices. KfW describes identifying the first practicable applications as "a central barrier". However, this is the authors' interpretation, derived from a moderate correlation with the use of external information sources, not a direct survey question. In a Swiss study on AI in marketing, 44 per cent cite "no clear benefit/need", but at the same time 57 per cent cite at least one knowledge-related barrier (Handelsverband.swiss).
Both findings can be read as a translation problem. Ideas are not scarce; what is scarce is the knowledge of which idea is both useful and feasible in one's own business. Attitudes support this: according to Bitkom, 91 per cent consider AI the most important technology of the future.
What this means for decision-makers
Anyone using study figures for their own decisions should first check who was surveyed, from what company size and when. More importantly, if knowledge is the bottleneck, collecting yet more ideas will not help much. What is needed is the ability to test an application against data, law, costs and process before budget is committed. The decisive questions are then: which assumption should a pilot test? Which mistake would be expensive? And what must be proven before the next budget step? How AI potential can be methodically transferred to processes and products is the subject of the next article in the capinorium.
If you would rather not take this step alone, a structured potential analysis or feasibility assessment with capinori can serve as a starting point.
Frequently asked questions about barriers to AI adoption in mid-sized companies
What is the biggest barrier to AI in mid-sized companies?
The biggest barrier is a lack of knowledge about AI. According to the German Federal Statistical Office (Destatis), 72 per cent of companies that have considered but not introduced AI cite this reason. In the Bitkom AI study 2026, the figure is 85 per cent of companies not using AI.
Is a lack of ideas or use cases a main reason against AI?
No, a lack of use cases ranks far down in the studies. Bitkom measures 30 per cent among non-users and 11 per cent among users (Bitkom 2026), Destatis 18 per cent for "not useful". What is evidently harder is judging which application is useful and feasible in one's own business.
Why do the figures in AI studies differ so much?
The studies ask different groups, start at different company sizes and define AI use and time periods differently. Destatis, for example, only asks companies that have considered AI; Statistik Austria includes all non-users. That is why "lack of knowledge" is 14.6 per cent there rather than 72 per cent.
Do the Bitkom figures of 41 and 57 per cent AI usage contradict each other?
No, the two figures come from two different Bitkom surveys from 2026. The 41 per cent is from the digitalisation survey of 11 March 2026, the 57 per cent from the AI study of 14 September 2026. Both cover companies with 20 or more employees and are around six months apart.
Do small companies use AI less because they are small?
Hardly. Small companies do use AI less often, but according to a KfW regression analysis the differences by size are not statistically significant once graduates, digitalisation and innovation activity are taken into account. What matters, therefore, is a company's knowledge base and innovation routines.
What barriers do companies already using AI face?
Companies using AI mainly cite data protection (66 per cent), legal uncertainty (56 per cent) and costs that are hard to calculate (54 per cent), according to the Bitkom AI study 2026. Only 25 per cent cite a lack of know-how. The comparison is a snapshot and does not prove a learning effect through use.
Sources (9)
- Bitkom: AI study 2026, press release "Erstmals nutzt Mehrheit der Unternehmen KI", 14 September 2026 (German). Link
- Bitkom: presentation "Künstliche Intelligenz in der Wirtschaft 2026" (Dr Ralf Wintergerst), 14 September 2026 (German). Link
- Bitkom: digitalisation survey, press release of 11 March 2026 (German). Link
- Destatis: reasons against the use of AI by employee size class, ICT survey 2025, as of 24 November 2025 (German). Link
- KfW Research: Fokus Volkswirtschaft No. 533, 11 February 2026 (German). Link
- KfW Research: Fokus Volkswirtschaft No. 554, 29 July 2026 (German). Link
- Bundesnetzagentur: KI in Unternehmen – Einsatz, Ressourcen und Herausforderungen, July 2025 (German). Link
- Statistik Austria: IKT-Einsatz in Unternehmen 2025, Vienna 2026 (German). Link
- Handelsverband.swiss: report on the KMU Digital Pulse 2026 (localsearch/HSLU), 29 September 2026 (German). Link
Rudy Winklhofer is Owner and Director R&D of capinori GmbH.
More articles by Rudy Winklhofer
