
AI trends for SMEs: seven developments that matter
In short: The most important AI trends for German SMEs are less about technical novelties than about shifts in everyday work: AI has reached the majority of companies with 20 or more employees, AI agents are coming into focus, data protection and data sovereignty matter more, internal knowledge becomes usable through AI, and regulation and training are becoming mandatory topics. The difference is made by businesses that adapt their processes and track costs and benefits per task.
There are plenty of AI trend lists. For managing directors and department heads in small and medium-sized companies, a different question counts: which of these affect my business directly, and what should I do in the coming months? This article puts seven developments into context that show up clearly in current surveys, and derives a first step for each.
Two figures show how much the picture varies by company size. According to Bitkom, 57% of German companies with 20 or more employees use AI in 2026, a majority for the first time; a year earlier it was 36%. For the German Mittelstand as a whole, including many micro businesses, KfW Research arrives at a much lower figure: in the period 2022 to 2024, 20% of SMEs used at least one AI technology. The figures do not contradict each other; they measure different groups and periods. Together they show that larger SMEs are in the middle of adoption, while smaller ones are often still at the start.
Trend 1: AI becomes everyday, especially for text and language
The most widespread applications are those that can be used without a major project. According to KfW Research, 14% of German SMEs use technologies that generate natural language and 10% use text recognition, whereas more demanding methods such as data analytics are used by only 3%. According to Bitkom, AI is used most often in customer contact, for example when handling enquiries (72%), and in marketing and communication (54%).
What this means for you: Starting with text, email and enquiries is the obvious route. The question is no longer whether employees use chat tools, but whether they do so in a regulated, secure way that actually affects the process.
Trend 2: AI agents come into focus
The next step after the chat assistant is AI agents that complete several work steps on their own: read an enquiry, look up data, create a draft, submit it for approval. Among companies that use AI or are planning or discussing it, agents are no longer a topic for only 24%, according to Bitkom. The difference between the two approaches is explained in Agentic AI vs. chatbots.
What this means for you: Start with an agent for a frequent, clearly defined task and decide from the outset what the agent may do itself and what a person approves. Examples with a labelled ROI calculation are in AI agent use cases.
Trend 3: data protection and sovereign AI become selection criteria
As usage grows, so does concern about data. Among companies using AI, 66% name data protection requirements as an obstacle, according to Bitkom. Accordingly, more businesses ask where their data is processed, whether a model can run in the EU or on their own infrastructure, and how trade secrets stay protected.
What this means for you: For each application, clarify which data it processes and where. Options from EU cloud to on-premise are compared in Sovereign AI and on-premise LLMs. This overview is not legal advice.
Trend 4: internal knowledge becomes usable through AI
Manuals, work instructions, quotes and project files contain a lot of knowledge that hardly anyone can find. AI assistants that access these documents and cite their sources are gaining importance: according to Bitkom, 22% of companies use AI for internal knowledge management in 2026, up from 11% the year before. The technical basis is usually retrieval-augmented generation, explained in the RAG guide.
What this means for you: Before a knowledge assistant can answer well, the sources must be current, findable and have access rights. That is often the real work.
Trend 5: regulation and AI literacy become mandatory topics
The EU AI Act has arrived in companies' day-to-day work. According to Bitkom, 37% of companies expect to be affected by the AI Act as users. Which obligations apply when, and what has changed through the Digital Omnibus, is summarised in EU AI Act: what companies need to do now. Article 4 of the regulation addresses staff training; more on this in AI literacy obligation under Article 4. Many businesses are already acting: according to Bitkom, 70% of all companies now train their employees in using AI.
What this means for you: Record which AI applications run in your business, who is responsible for them and how employees are trained. A simple overview is the first step towards AI governance.
Trend 6: from tool to adapted process
Many businesses find that an AI tool on its own changes little if approvals, handovers and roles stay the same. KfW Research states that broader AI use requires not only investment in IT infrastructure and skills, but also changes to work organisation. At the same time, Bitkom shows what is possible: 48% of companies using AI were able to speed up their internal processes significantly.
What this means for you: Appoint a responsible person for your first AI process, define approvals and measure beforehand. Why this matters so much is described in The AI paradox.
Trend 7: budgets rise, costs arise elsewhere than expected
According to Bitkom, 45% of all companies are spending more on AI in 2026 than the year before. The major cost items are rarely licences: 51% of AI users name infrastructure such as cloud and compute, 50% data preparation and 41% integration into existing systems. Among companies without AI, 63% cite costs that are hard to calculate as a reason for holding back.
What this means for you: Plan data preparation, integration and ongoing operation from the start and calculate the benefit per task.
Overview: trend and first step
| Trend | First step for your business |
|---|---|
| AI in everyday work | Record the use of chat tools and set rules |
| AI agents | Choose a frequent, clearly defined task as a pilot |
| Data protection and sovereign AI | Clarify per application which data is processed where |
| Internal knowledge | Tidy up the most important documents and clarify access rights |
| Regulation and literacy | Create an overview of AI applications, plan training |
| Process instead of tool | Appoint a process owner, measure baseline values |
| Costs | Include data, integration and operation in the calculation |
What the trends mean for smaller businesses
For businesses with 10 to 50 employees, the biggest hurdle is rarely the technology but time and know-how. Among companies not using AI, 85% name a lack of technical know-how for implementation as the reason, according to Bitkom. KfW Research sees identifying the first practical applications as a central hurdle. One approach is to introduce AI not as a collection of tools but along the org chart: which role has which tasks, and which of them can an AI specialist take over or prepare?
This is how mysoftwarelab GmbH from Worms, Germany, works as an external AI department for businesses with 5 or more employees: it builds and runs a company cockpit in which AI specialists take on roles while your team steers and approves. The starting point is an org chart analysis of the company.
Frequently asked questions about AI trends for SMEs
How many companies in Germany use AI?
According to Bitkom, 57% of companies with 20 or more employees use AI in 2026, a majority for the first time. For the Mittelstand as a whole, including micro businesses, KfW Research arrives at 20% for the period 2022 to 2024. The figures measure different groups and periods.
What is the most important AI trend for small and medium-sized businesses?
The move from individual chat tools to AI agents that take over entire work steps and submit results for approval. This only becomes effective if the process around the agent is adapted.
Does my company need to comply with the EU AI Act?
Anyone using AI in their business should check whether and how they are affected. Our article on the EU AI Act summarises the current status of obligations and deadlines. If in doubt, seek legal advice.
How much does getting started with AI cost?
It depends less on licences than on data preparation, integration and operation. Calculate the benefit per task and start with a pilot whose baseline values you measure beforehand.
Do SMEs need sovereign AI?
Not for every application. But for trade secrets, personnel data or customer data, it is worth considering providers that process data in the EU or running models on your own infrastructure.
Conclusion: the trends are clear, the advantage comes from execution
AI has reached the majority of larger SMEs; agents, data protection, knowledge management and regulation shape the next steps. Businesses that now choose a concrete task, adapt the process, regulate approvals and measure the benefit build an advantage that cannot be bought. If you want to translate the trends into a plan for your business, you will find support through our AI consulting or directly by booking an appointment.
Sources
Who builds your company cockpit
Benefit from over 20 years of enterprise experience
Andreas Indorf
Managing Director, mysoftwarelab GmbH
Qualification: Founder of mysoftwarelab GmbH; with his team, builds company cockpits with AI agents for SMEs
Expertise: Over 20 years of developing and implementing IT systems for DAX companies and international corporations. Specialized in AI automation for mid-sized businesses since 2021.
Hands-on Experience: mysoftwarelab is a model company: around 80% of its recurring tasks are done by AI agents from its own cockpit. This hands-on experience flows directly into our client consulting.
Focus: Company cockpits in which AI agents take over recurring tasks – for companies with 5+ employees, built and run by us.
More articles on AI transformation
AI Costs for SMEs: Planning a Realistic Budget
What AI costs mid-sized firms: cost blocks according to Bitkom, indicative values per project type, an example calculation for an AI agent and five steps.
How to Implement AI in a Business: 7-Step Guide
How to implement AI in a business with the org chart method: 7 steps from task analysis to your first AI agent, plus timeline, roles and obligations.
Specialized AI agents: job profiles, not all-purpose AI
Specialized AI agents by role: expertise, rules, checklists, limits and approval. With three tiers, a pool of 51 job profiles and a checklist.
Which of your tasks belong in the cockpit?
In a free callback we clarify whether a company cockpit with AI agents fits your business – for companies with 5+ employees.
- An individual company cockpit instead of single AI tools
- AI agents take over recurring tasks, your team approves
- Built and run by us – individual offer
From the lab
The most important AI news of the week – as a comic.
Every Wednesday, ByteUndGlitch explains one real AI story in eight panels. Satirical, accessible, free.
Read the latest episode