Services · developerz.eu
AI integration
Practical use of AI in business processes — from automating recurring tasks through assistants on your own data to documentation that accounts for data protection and the EU AI Act from the outset.
- Process automation with AI
- Assistants on company data (RAG)
- Integration with Claude / OpenAI API
- Documentation under GDPR and the EU AI Act
AI where it saves time
I start with the process, not with the technology. First I measure what a process costs today — then I implement a pilot and compare. If the saving does not show up on paper, I say so. A project that does not pay for itself helps neither you nor me.
Typical areas of use
- Pre-qualifying enquiries — classifying incoming messages, summarising them and answering with drafts.
- Analysing documents — summarising contracts, reports and paperwork and making them searchable.
- Assistants on company data (RAG) — answers from your own documents, with sources cited rather than invented.
- Automation in the background — recurring steps within existing processes.
I do not train my own models. I integrate existing ones, cleanly and traceably — which is what a company actually needs.
Data protection first
What may go to a model and what may not is defined and documented before implementation. Processing through APIs without training on your data, a data processing agreement under Art. 28 GDPR as standard, and on request a restriction to EU regions. As an EU provider I need no standard contractual clauses for this.
The EU AI Act: what I deliver and what I do not
Regulation (EU) 2024/1689 has applied since August 2024 and brings its obligations into force in stages; the stage due on 2 August 2026 has now been reached. For a purchase in the DACH region this is no longer a question about the future but a point that comes up in the review. So I prepare it rather than waiting to be asked.
- Classification per use case, in writing. The regulation does not regulate “AI” wholesale, but the specific use. Most operational integrations — summarising texts, classifying enquiries, producing drafts — do not fall into the high-risk category. It looks different as soon as systems reach into areas such as recruitment, credit scoring or access to essential services. Which classification applies to your case is examined and recorded with reasons — not assumed.
- Transparency where people are affected. Where users interact directly with an AI system, or output content is machine-generated, labelling obligations apply. I implement these technically: recognisable notices in the interface, labelling of generated content, a route to a human.
- Traceability in operation. Logging of calls, defined permissions, and the documentation of which categories of data reach which model. Without that foundation nothing can be evidenced after the fact.
What I do not deliver: legal advice, certifications, declarations of conformity or an assurance of compliance. I provide the technical documentation on which your legal department or your external adviser bases its decision. Anyone promising you guarantees at this point is selling you something they cannot uphold.
When it makes sense
- Your team spends a lot of time reading, writing and sorting
- Enquiries or documents arrive in large numbers
- You want to adopt AI soberly, not as an end in itself
- You have to evidence the use internally — to data protection and compliance
How we work together
- 01
Process analysis
I look for the processes in which AI saves measurable time — and name the ones in which it does not.
- 02
Classification and data concept
Which categories of data go to a model, which do not, and how the use case is to be classified under the EU AI Act. In writing, before the pilot.
- 03
Pilot
One process, real data, a verifiable result — within a few weeks.
- 04
Integration
Connection to your systems, permissions, logging.
- 05
Training
The team learns where the use makes sense and where it does not.
- 06
Operation
Support and adjustment, or handover to your team.
What you get
- Pilot with a measurable result
- Integration with your systems
- Documentation of the data processing (GDPR)
- Documented classification of the use case under the EU AI Act
- Training for the team
- Source code owned by you
FAQ
Where does AI actually deliver something in a company?
Everywhere people repeatedly read, write or sort: answering enquiries, summarising documents, preparing paperwork, categorising data. I start with the process, not with the technology.
Does our company data end up training the models?
No. I work through APIs in a mode in which your data is not used for training, and I process sensitive documents in a segregated environment. Which data may go where is part of the concept — not a question first asked after go-live.
What about the GDPR?
The processing is documented, a data processing agreement under Art. 28 GDPR is standard, and I record in writing which categories of data go to a model at all. On request I restrict processing to EU regions. As an EU provider I need no standard contractual clauses for this.
And the EU AI Act?
The obligations of Regulation (EU) 2024/1689 apply in stages; the stage due on 2 August 2026 is now in force, so the topic is no longer announced but applicable. In practice that means the use case is classified and the classification documented. Most operational integrations — summarising, classifying, drafting — do not fall into the high-risk category. Where people interact directly with an AI system or content is AI-generated, transparency obligations apply, which I implement technically. I produce the documents on which your legal department bases its decision — this is expressly not legal advice, and I hold no certifications.
How will we know whether it was worth it?
I measure how much time the process costs before the rollout and compare afterwards. If the saving does not show up in the numbers, I say so and advise against it.
Planning a new project?
Get in touch — in a first call we go through your goals, the scope and whether I am the right person for it, with no obligation.