Skip to content
Services

AI, automation and system integration for companiesFrom process problem to a running solution.

I build custom solutions for companies when existing flows create too much manual work, systems do not play together cleanly, or AI can improve a concrete process.

This is not about using as many tools as possible. I first analyse the flow and then decide whether AI, classic automation, Microsoft 365, a system integration or custom software is the right setup.

The setup should not only work at go-live. It should stay traceable, maintainable and adaptable.

AI

AI assistants & knowledge access

Problem

Information sits scattered in documents, emails, SharePoint or other systems. People spend time searching, summarising and asking around.

Solution

An AI assistant reaches relevant company data in a controlled way and puts knowledge into day-to-day work.

Depending on the use case that can run via existing cloud platforms, a knowledge base of your own or a local model. What counts are data sources, permissions, knowledge quality and the concrete task.

My own local speech assistant shows, for example, how speech processing, language models and knowledge access can also run without the cloud.

More in the blog: Local AI speech assistant without cloud

FX

Process automation

Problem

People copy data between systems, check the same information again and again or handle recurring cases by hand.

Solution

Recurring flows are automated. Systems are connected and information is passed to the next station automatically.

Not every process has to be solved with AI. If rules and classic automation already work reliably, that is often the better variant.

AI is added where unstructured information, language or ambiguous cases have to be processed.

Typical use cases are document processing, email classification, incoming payments, data transfer and internal approval flows.

∫

System integration

Problem

ERP, CRM, Microsoft 365, email, files and other apps sit side by side. People become the interface.

Solution

Existing systems are connected via APIs, interfaces and workflows.

The goal is not necessarily a new platform. Often an existing process improves when the right systems talk to each other and manual handoffs drop out.

I work with existing APIs and platforms as well as with custom software when a standard setup is not enough.

For delivery I combine existing platforms, workflow automation, APIs, AI models and custom software, depending on the requirement.

See the technical basis
◇

AI consulting & technical design

Problem

There are many models, vendors and automation tools, but no clear decision yet on what makes sense for your own process.

Solution

I translate the concrete process into a technical architecture and decide with you which components are actually needed.

That includes, among other things:

  • AI or classic automation?
  • Local model or cloud?
  • Which data has to be processed?
  • Which systems need to be connected?
  • Which interfaces and workflows are needed?
  • How does the setup stay maintainable and extensible?
  • Where do unnecessary lock-ins to individual vendors arise?

Not every task needs AI. And not every AI setup needs a large language model.

Architecture

Tech is part of the solution, not the solution itself

A working AI application is not just a language model. Data sources, permissions, interfaces, workflows, logging and maintenance decide just as much whether a setup works day to day.

That is why I look at architecture and operations from the start.

A system must not only work today. It must also be updateable, monitored, extended and, if needed, taken over by another developer.

That view matters especially when a prototype becomes a production application.

More in the blog: Why maintenance routines decide project success

AI

Local, cloud or hybrid

The question is not whether local AI or cloud AI is better in general. The question is which architecture fits the process and the data risk.

Local AI

If data should not leave the company, a local model on your own hardware or an internal server can make sense.

Cloud AI

If stronger models, scale or complex tasks matter more, an enterprise API can be the better setup.

Hybrid

Local AI Cloud AI

Sensitive processing can stay local, while less critical or especially compute-heavy tasks run via a cloud API.

The architecture is chosen by the process, not by the current AI hype.

Delivery

From idea to a productive process

  1. Understand the process I first clarify how the flow works today, where manual work arises and which systems and data are involved.
  2. Design the setup Then automation, AI, interfaces and architecture are fixed. Data risk, maintainability and operations are part of that.
  3. Build and test The setup is built and tested against concrete cases. Error cases and exceptions belong in that as much as the normal flow.
  4. Operate and extend After delivery the setup stays traceable and maintainable. Updates, adjustments and extensions can be planned instead of improvised.

Not sure yet which tech fits?

Briefly describe the process that currently creates unnecessary work.

I assess whether AI, classic automation, integration or another technical setup makes sense.

Discuss the process