According to Bitkom (2025), 36% of German SMEs are already using AI tools – but only 5% are achieving measurable business results with them. That's not AI scepticism; it's an implementation problem. And that's exactly what we see every day in projects across Bad Tölz, Wolfratshausen and Rosenheim: the current AI trends aren't the problem. The problem is that they're introduced without a clear goal.
Here are the five developments that are genuinely relevant for SMEs in Upper Bavaria right now – with figures, tools and what they actually cost in practice.
The biggest leap of the past twelve months: AI systems that don't just answer a question, but independently carry out multi-step workflows. An agent today can review quotes, contact suppliers, evaluate responses and produce a decision brief – without a single employee touching each step.
A concrete example from the region: A metal construction company from Penzberg used an AI agent for quote generation and cut the turnaround time from enquiry to finished quote from an average of 3.5 days to under 18 hours – a time saving of around 40%. Since then, the sales manager manually coordinates two quote processes per week instead of the previous five.
For orchestrating agents like these, we at Softstorm prefer CrewAI. The reason is straightforward: CrewAI lets you have multiple specialised AI agents (e.g. one for data research, one for content drafting, one for quality checking) work together in a defined sequence. Compared to more generic frameworks like LangChain, the entry point for practical business processes is significantly faster – without having to dive deep into prompt engineering. Find out more about how we build systems like this in our AI agent projects.
According to Salesforce and the German Mittelstands-Bund (2026), 51.2% of mid-sized companies are already using or testing AI solutions – an increase of 54% compared to 2024. The main reason for this rise: the tools have become simpler and cheaper.
n8n is one of the tools we use in almost every automation project. It connects existing software (ERP, CRM, email, Google Sheets, accounting) without requiring deep programming knowledge and can be self-hosted – which is a decisive factor for many businesses in Upper Bavaria, as it means data doesn't end up in American cloud services. Compared to Zapier or Make, n8n is more flexible for complex workflows and significantly cheaper in the long run, since there are no transaction-based costs.
Softstorm observation: In 7 out of 10 of our automation projects, the first bottleneck wasn't missing technology – it was unclear or undocumented internal processes. If you don't know how your own workflow actually functions, you can't automate it.
If you want to find out which processes in your business should be automated first, take a look at our automation service.
Two trends that look separate at first glance but are often used together in practice:
Voice AI (e.g. via Vapi) enables AI-powered telephony – systems that answer incoming calls, qualify them and route them accordingly. For service providers in Holzkirchen or Miesbach who miss a lot of enquiries outside office hours, this is a real lever. Compared to simpler solutions, Vapi stands out for its low latency and the ability to integrate custom language models – making responses more natural and adaptable to specific industries.
RAG systems (Retrieval-Augmented Generation) solve a different problem: employee knowledge locked away in PDFs, internal wikis or old emails is made searchable by AI. Instead of repeating the same onboarding effort with every new hire, a RAG system built on Qdrant (a local vector database) can search internal documents in seconds and deliver structured answers. Like n8n, Qdrant can be self-hosted – a clear advantage for privacy-conscious businesses in Bavaria over cloud-based alternatives such as Pinecone.
A real drawback you should know about: RAG systems are only as good as the quality of the documents you feed them. Outdated, contradictory or unstructured internal materials lead to incorrect answers – and you may not even notice. Before introducing a system like this, you need an honest audit of your knowledge base. That's an effort that's often underestimated.
Our clear position: Treat AI as a software project and you'll fail. Treat it as a process change and you'll succeed.
Most providers sell ready-made AI solutions tailored to generic use cases. For a trades business in Geretsried that wants to map its specific quoting process with individual material prices and customer categories, that doesn't help much. What does help: a small, concrete pilot project with a clear, measurable goal – for example, "produce quotes in 50% of the time" – followed by gradual expansion.
This is backed up by KfW Research (2026): the share of mid-sized companies using AI has grown fivefold in six years. But growth in usage doesn't automatically mean growth in results. The 5% that achieve measurable results according to Bitkom don't differ from the rest because of better tools – but because of clearer goals.
If you want to know what a structured AI rollout could look like for your business – without consultant jargon and without project costs spiralling out of control – our AI consulting service is the right first step.
The next step: Give us a call or drop us a quick message about what's taking up the most time in your business right now. In a 30-minute conversation, we'll show you whether and where AI could already make a measurable difference for you today – no obligation, no presentation.
We analyze your processes, identify automation potential and show you which digital systems really make sense.
