Common Questions About AI in the Mittelstand

Probably yes, but not in every area. Most mid-sized companies find at least two or three use cases where AI creates significant value. Common areas are knowledge assistance, document processing, prediction models and configuration support. Which use cases are worth it for your company is what we clarify in Getting started with AI.

Yes. In fact, that's the norm. Most AI projects in the Mittelstand are delivered with external support. What matters is that you have someone internally who drives the project on the business side, and that the external consultancy involves your people in a way that lets them run the system on their own later on.

It depends on the kind of project. The orientation at the start, collecting, assessing and prioritising use cases, takes a few weeks. A first productive system in a department usually needs a few months, including integration and test operation. A platform for the whole organisation is a programme over several months. Anyone promising you a productive AI system in two weeks is exaggerating.

There is no flat rate, because a single automation is a different project from a platform for the whole organisation. What drives the effort is explained in What does an AI project cost?. After a short intro call we give you a ballpark with the assumptions behind it; the proposal then comes with a fixed price.

In three steps: in the intro call we classify your project and give a first ballpark. For a concrete project we look at process, data and systems in one or two sessions. From that comes a proposal with clear scope, deliverables and a fixed price.

That depends on which AI platform you use. With most European providers (or with the European subsidiaries of large US providers, such as Azure OpenAI in the EU), your data stays in Europe in a GDPR-compliant way. For sensitive data or special requirements we build private AI platforms with RAG that run on-premises or in your own cloud. We discuss this explicitly in every project.

As a rule, no. Modern AI applications can be integrated into existing systems via APIs, whether ERP, CRM or industry-specific software. Major rebuilds are only necessary if your data is so scattered that there's no usable data foundation. We clarify this in a data audit.

For most Mittelstand use cases, the requirements are manageable. What matters is the obligation around your employees' AI literacy (in force since February 2025). High-risk applications (e.g. in hiring) carry more extensive obligations. The deadlines for these have been pushed back with the "Digital Omnibus": to 2 December 2027 for most high-risk systems and to 2 August 2028 for AI embedded in products. More on this in our pillar article on the EU AI Act.

That's why we deploy AI where a human checks or approves the results. For every use case we define the limits of use, make decisions traceable and build in checkpoints, part of sound AI governance.

The bigger risk is often not the AI itself but shadow AI: when AI is banned, many people use private tools anyway, and company data flows out uncontrolled. The better path is a vetted, GDPR-compliant platform for everyone, plus clear policies for the genuinely critical risks (personal and regulated data). That keeps control where it matters, without slowing innovation.

We build so you can switch: open standards, interchangeable models, and your data and your logic stay with you. Where it fits, we use self-hostable components. The solution belongs to you.

The biggest hurdle isn't the technology, it's acceptance. We involve the eventual users from the start, build solutions for real problems and train hands-on. An AI that solves a genuine annoyance catches on by itself.

By building the pilot for production from the start: clean code, real integration, defined success criteria. After the pilot we move the system into operation (with monitoring and ongoing development by us if you want) or hand it over cleanly to your team.

We'll tell you honestly when a standard product is enough for your use case and no custom development is needed. Even when there's nothing in it for us. We make a living from clients recommending us, and that only works if we don't try to squeeze the maximum out of every project.

Of course. Many clients, after getting started with us, take the results and implement them on their own. That's explicitly a desired outcome. We won't necessarily sell you the next project. If the collaboration went well and you want to continue, that's naturally welcome.

Yes. Before every project, and on request before the intro call as well, we sign an NDA (non-disclosure agreement). We also publish our case studies only in anonymized form, unless we have explicit approval.

Three things: First, from the very first meeting you talk to the people who will later develop the concept or build the application. Second, we combine consulting and development in one team, without handovers between separate departments. Third, we know regulated industries: documentation duties, approval processes and evidence are part of the concept with us, not an afterthought.

Usually within a week, often sooner. You can book the appointment directly online.

Then we say goodbye and wish you all the best. The initial call is free and non-binding. We believe you're the best judge of whether the chemistry and the approach are a fit.

Yes. Through the Nao Training Hub we offer an AI-literacy programme under Article 4 of the EU AI Act, as a self-paced course or live training, with your own rules and tools, a final quiz and a certificate per participant.