AI that runs in production — not just in the demo.

It depends on the use case — a single n8n workflow is quick to build, a RAG chatbot over large document sets is more involved. We work audit-first: in a use-case audit we check what's genuinely worth it, and you get a fixed-price offer before we build.

Retrieval-augmented generation combines a language model like Claude with your own documents. The chatbot searches your manuals, tickets and knowledge bases and answers based on them — with source citations and far fewer hallucinations than a freely answering bot.

Yes, that's the core of our approach. We run n8n self-hosted in the EU, draw clear data boundaries and put the necessary data-processing agreements in place. On request we use local or European models instead of a connection to US providers.

No. Over the commercial API tiers your content is not used to train the models. We configure the integration so your data is only processed for the specific request.

We run the automation layer (n8n) self-hosted in the EU. For the model connection we choose EU data residency or European/local models depending on the requirement. So you keep control over where data flows.

Typical ones are knowledge chats on your own docs, mail triage and ticket routing, lead scoring right in the form, and extracting structured data from invoices, PDFs and emails. In the audit we prioritise the use cases with the best effort-to-value ratio.

A classic chatbot answers individual questions. An AI agent can complete multi-step tasks, call tools and APIs and make decisions along clear guardrails — for example check a request, look up data and trigger an action.

Because they start as an impressive demo but lack data boundaries, quality measurement and integration into existing systems. That's exactly where our audit comes in: we prioritise operable use cases over hype and build for production from day one.

A clearly scoped workflow is often in production within a few days; a RAG setup or an agent with several integrations takes correspondingly longer. We start with a small, measurable proof of concept on real data.

As soon as personal data is processed, yes. We help with the necessary data-processing agreements with the providers used and set up the processing in a GDPR-compliant way.

That's exactly why every project is built to be handed over: a standard stack instead of exotic tools, clean and tested code, documented architecture and deployment — and the repository is yours. Any competent developer can carry on without me. You're never chained to one person.

No. The project is cut into milestones with a fixed scope and a fixed price. You pay per milestone, get a working result after each one, and decide whether to continue. No open-ended hourly tab, no surprise on the invoice.

Often faster. No coordination overhead, no ticket ping-pong, no friction between junior and senior — you talk directly to the person who builds. If scope grows, I bring in vetted partners selectively, always transparently agreed.

Then you take the code, the docs and all access with you — it's yours anyway. I deliberately build in no dependency that locks you to me. Good work doesn't need a lock-in clause.

We clarify that before building, not after. We start with one concrete use case whose value can be measured — for example a knowledge chat on your documents. If it isn't worth it, I'll say so plainly instead of selling a project that fizzles out.