AI consultant for small and midsize companies in Germany

Why AI stays out of many smaller companies

The main reason is missing know-how. In a Bitkom survey of 603 companies with 20 or more employees, published on September 14, 2026, 57 percent used AI. Of the companies without AI, 85 percent said they lacked the technical know-how to introduce it (Bitkom).

A smaller company rarely has a person whose job is to follow AI. The managing director reads about agents and chatbots, and staff try tools on their own. Nobody has the time to test which task would repay the effort. That test is the part I take over.

What an AI consultant does

An AI consultant finds the tasks in a company that AI can take over and plans the first project with a way to measure its result. The candidates come from the people who do the work, because they know which tasks repeat: the offer text rewritten for each customer, or the delivery note typed into a second system.

Each candidate then goes through the scorecard below, and the task with the highest total becomes the pilot. Some candidates are better served by a plain script or a better form than by AI, and the proposal says so.

The AI use case scorecard

I score each candidate task on seven questions, with 0, 1 or 2 points each. You can go through it before our call with the tasks you have in mind.

Question	0 points	2 points
How often does the task come up?	A few times a year	Each week or more often
How does the input arrive?	On paper or by phone	As email, PDF or web form
Is the rule written down?	It lives in one person’s head	A new colleague could follow it
How fast can someone check the result?	Only the customer notices a mistake	A person or a script checks it in minutes
May the data go to an AI service?	Your data protection officer says no	Approved for the plan you use
What does a mistake cost?	It reaches a customer at once	It can be corrected internally
Who owns the task?	Nobody in particular	One person who wants the change

A task with 0 points on the data question waits, or it runs on an open-weight model on hardware you control. The last row counts as much as the others: a pilot without an owner has nobody who carries it into daily work.

What my advice is based on

I work with AI agents in client work, and my proposals come from what held up there. On a client’s platform, coding agents and people work in one repository. The agents have limits at the database and the branches, and a hook checks each tool call before it runs. Details: AI agents under the same rules as humans.

For the same client I set up agents with roles on a control plane: a chief hands out the work, and a script that the agent cannot change measures each result. The first job showed why: the work was right and the agent’s report was wrong. Details: Virtual organizations of autonomous agents.

Who I am

I am Michael Wutzke, an AI engineer and consultant in Frankfurt with more than twenty years in IT and media. I was Head of Decentralized Finance and Node Operations, then CIO, at the Frankfurt-based company Blocksize Capital, and I held leadership roles as IT Team Lead and Managing Director before. Today I help a client rebuild its platform with Claude and Codex as coding agents, and I teach at Claude Hacker House. I am also interested in open-source AI models that a company runs on its own servers. Details: Career stages.