According to Bitkom and the Institute for Employment Research, the share of German businesses using generative AI quintupled between 2023 and 2025 – from 5 to 25 percent. That sounds like a breakthrough. But at the same time, a third of companies report higher AI costs than expected. For small and mid-sized businesses, where every euro is counted twice, this is no academic problem – it's reality.
At Softstorm, we support businesses from Bad Tölz-Wolfratshausen and the wider Oberland region in rolling out AI solutions. And we see the same mistakes time and again. Here are the five most common ones – so you don't have to learn them the hard way.
Mistake 1: Starting with the tool instead of the problem
The most common mistake of all: a business owner reads about ChatGPT, buys a licence – and three months later wonders why nothing has changed. AI isn't a plug-and-play solution you can just slot in somewhere. It solves specific, repeatable problems. Which one of those do you have?
Before you even look at a single tool, you should have a clear answer to this question: Which process costs me or my team the most time every day – and is it structured enough to be automated?
Mistake 2: Starting too big
Many businesses want to do everything at once with their first AI project: customer service, accounting, marketing, quote generation. The result is a half-finished system nobody really uses – and a team that's frustrated.
In 8 out of 10 of our projects, lack of focus was the first bottleneck. A single, clearly defined use case – for example, automatically pre-qualifying incoming enquiries – delivers more than five half-baked ideas running in parallel.
Mistake 3: Not bringing the team along
New technology rarely fails because of the technology. It fails because people don't understand why they should change anything. If you introduce AI without involving your team early on, you'll meet resistance – sometimes openly, sometimes quietly.
Plan a brief introduction for your staff from the start. Not a training session with slides, but a real conversation: What's changing for you? What stays the same? What will get easier?
Mistake 4: Choosing the wrong tools – or too many at once
Our honest view: most AI consultancies sell complexity that no SME actually needs. A small carpentry business doesn't need a self-hosted agent infrastructure with five microservices – it needs a solution that's up and running in two weeks and saves three hours a week.
For automation tasks – such as automatically routing customer enquiries, populating CRM fields, or sending reminders – we frequently rely on n8n. The reason: n8n can be self-hosted, is in line with GDPR requirements without any extra effort, and offers a visual interface that even less technically experienced staff can follow. Zapier is an alternative – but with Zapier, costs quickly run into three figures per month once volume grows. n8n stays cost-efficient even as usage scales.
When more intelligent, multi-step processes are involved – such as an agent that evaluates enquiries, asks follow-up questions, and sends a summary to your inbox – we use Flowise. Flowise is built on LangChain but has a visual interface that requires no coding knowledge. That makes it one of the most accessible entry-level tools for AI agents in an SME context, without sacrificing flexibility. You can find out more on our AI Agents page.
Mistake 5: Not defining clear success criteria
"We want to use AI" is not a goal. "We want to reduce the processing time for customer enquiries from 4 hours to under 30 minutes" is. Without measurable criteria, you won't know after three months whether the project was a success – or whether you simply burned through money.
Define before you start: What should change, by how much, and within what timeframe? And who is responsible for tracking it?
A carpenter from Holzkirchen used an automated quote process – built with n8n and a simple language model in the background – to cut his average quote preparation time from 45 minutes to under 10 minutes. Not because he launched a massive AI project, but because he tackled exactly one process: structuring and pre-filling quotes based on customer conversations.
The result: more time for the actual work, fewer errors in the documentation – and not a single team member who rejected the system, because it was built in a way they could understand. You can find similar approaches for the trades sector in our project on AI quote generation.
AI automation is not a cure-all. Especially when customer enquiries touch on legally sensitive areas – such as liability questions, warranty claims, or sensitive personal data – every automated solution hits its limits. An AI agent can prepare, structure, and route. A human must make the decision.
Anyone who ignores this during the build and hands too much responsibility to the system risks mistakes that become expensive. So always plan a clear handover point: at what stage does a human take over?
Introducing AI doesn't have to be complicated. But it does have to be thought through. Anyone who starts with a concrete problem, involves their team, chooses the right tools, and makes success measurable has significantly better odds than someone who simply buys a tool and hopes for the best.
If you want to know which process in your business would be the best candidate for automation first, talk to us. No pitch, no consultant jargon – just an honest conversation about what actually makes sense for you. Our Automation Service gives you a first overview of what's possible.
We analyze your processes, identify automation potential and show you which digital systems really make sense.
