Lots of access, little use
Copilot or a company GPT has been rolled out, but only a handful of enthusiasts use it. There are no concrete use cases in daily work, no support, and too few leaders going first.
AI ownership for mid-sized companies and scale-ups—as a project, interim, or permanent
The AI licenses are paid for. Hardly anyone uses them.
I bring AI into your processes so your people actually use it. With leadership I decide where AI pays off, with your teams I rebuild the first process, and after about 90 days a first result is measurable.
Works council and data protection officers are involved from day one.
At Bibel TV: An AI writing assistant for letters, emails, and social media requests. Complex letters in minutes instead of hours. Introduced without a mandate—now used by initially skeptical colleagues too.
- NinoxHead of Product Design, for a time also product management; member of the Senior Leadership Team
- DatameerDesign lead for an analytics platform used by leading banks and insurers
- HypoportTurned internal services into standardized offerings with the executive team
- Bibel TVAI introduced in viewer services and product development
Situation
Where AI gets stuck in mid-sized companies.
The bottleneck is rarely the model or the tool. What is usually missing is a clear priority, a better process, and someone who brings leadership, business teams, and IT together.
Many ideas, no priority
Every department tries something different. Nobody decides which process changes first or how its value is measured.
A core process is stuck
Domain knowledge, exceptions, and handoffs cannot be mapped in off-the-shelf software. The team needs a better process—and a tool that fits it.
Recognize yourself? Then you are not missing technology—you are missing someone who takes ownership of it.
Why now
AI has arrived. The benefit has not.
of German companies with 20+ employees use AI. For the first time, that is more than half.
Bitkom, September 2026 · in my reportof those AI users say they use its potential fully. Six in ten say they do not use it at all.
Bitkom, September 2026 · in my reportArticle 4 of the EU AI Act requires companies that use AI to ensure sufficient AI literacy among their staff.
Regulation (EU) 2024/1689Assess yourself
Where do you stand? Two ways to find out.
Ten questions, five stages, your next step.
Usage, processes, ownership, enablement and measurement: you see where you stand and where I would start. No sign-up, nothing is stored.
Start the check“Bought is not adopted”
Why more than half of companies use AI and hardly any get much out of it. From current studies and 187 podcast episodes with practitioners.
Read the summaryExamples
Three examples from practice.
Bibel TV · Viewer services
An AI assistant the team uses by choice.
The assistant drafts replies to letters, emails, and messages in the house voice; a person checks and sends every one. Introduced without a mandate: a trial week, then one-to-one sessions. Skeptics came around after it helped with their hardest case.
Minutes, not hoursfor complex letters covering several topics. Temporary staff write in the house voice from day one.
Full exampleBibel TV · Product development
From depending on an agency to building in-house.
Almost every change used to depend on an external agency, with no in-house developers. I led by example, coached product managers and owners one to one, and introduced Claude Code, Spec Kit, and clear rules with them. Today much is built in-house.
No agencyLeadership gave me the rebuild of the app, and I built it myself, with AI.
Full exampleNinox · AI in the product
Plain language becomes an editable application.
With the CEO and the technical co-founder, I developed an AI onboarding: users describe their work, AI generates the data model, and everything stays editable in the Ninox editor.
Edit directlybecame the interaction model of Ninox’s public AI beta
Full exampleThe first 90 days
People first: five steps.
Many AI projects start with a list of use cases. I start with how the work really runs today. The first cycle usually takes about 90 days, depending on approvals in your organisation. Then at least one AI-supported process is running that your people actually use, with a measured result—and the next one follows, with what the team has learned. IT security, legal, data protection, and architecture join as the risk requires, with specialist partners where needed.
- Step 1
Set the goal
Conversations with leadership: what should get better, and where are we today? Outcome: goals, baseline, and measures, in writing.
- Step 2
Understand the processes
With the people who run them every day: workflows, handoffs and exceptions, blockers and friction, good and bad moments. Outcome: a clear picture of where it really gets stuck.
- Step 3
Choose where to start
Rating by value, feasibility, and risk. Outcome: three initiatives in a fixed order, and a list of what we deliberately will not do. Not every answer needs AI: at ChurchDesk, days of phone tag became a booking capability for every customer, without any AI.
- Step 4
Design and build
One important process is redesigned with the people involved and built as a working solution. Built so your team can run it themselves a year from now. Plus two quick improvements people notice in daily work right away.
- Step 5
Enable and measure
Guidance on real work, AI champions with real time allocated, an AI guideline, AI literacy under Article 4 of the EU AI Act. At the end: a measured result and the next initiative.
Working together
How we can work together.
It always starts with a free conversation (30 minutes). Then you decide which form of collaboration fits.
AI assessment
In two weeks you know where AI should start in your company, and where it should not. I talk to leadership and to the people who do the work. Then you decide whether and how to continue.
- Goals, baseline, and measures, in writing
- A clear picture of where your processes really get stuck
- Three initiatives in a fixed order
- What you deliberately will not do
Also possible: one bounded process we rebuild together.
Interim mandate
I take ownership of AI for a defined period, following the 90-day plan. Afterwards I hand over to someone in-house I have onboarded—and can help with the role profile and selection. Or we extend. Or you fill the role permanently, with me if you like.
Permanent role
Want to fill the role permanently from the start? Then let’s talk about a position where strategy, delivery, and capability building stay in one pair of hands.
Guidance and sparring
For leaders and AI owners who do the work themselves and want an experienced sparring partner. One-to-one coaching, fixed office hours for your team, and training on real work instead of in a classroom.
Discuss guidanceWhat I need from you
- A clear mandate from leadership.
- One accountable person in the business team.
- The willingness to stop an initiative when the value isn’t there.
About
I lead—and I build myself.
I have led product and design teams, served on a Senior Leadership Team, and designed software for banks and insurers, media companies, and platforms. Today I build AI-supported processes myself: at Bibel TV, development builds directly on my prototype code. That lets me talk priorities with leadership, real work with business teams, and the actual system with IT. Product design taught me to take people seriously: their context, their goals, and what slows them down every day. An AI rollout needs exactly that, inside the company.
Leadership
Product and design leadership, Senior Leadership Team, distributed teams in Germany and New York, mentoring.
Product and process
Product strategy, process design, prototypes in real code, automation with APIs and n8n.
Enabling people
Leadership workshops and one-to-one guidance inside daily work. Ten years mentoring career changers at CareerFoundry, guest lecturer at Spiced Academy, design-thinking training at Hypoport, keynote at ICCM Europe in Brussels. Today, my own programme, AI Design Mentor.
Questions
Frequently asked questions.
Why someone with a product and design background?
Because AI rollouts rarely fail on the model—they fail on adoption. Shaping processes that people actually want to use has been my work for more than 15 years. Many AI leads today come from product management, process management, or business functions, not from research.
Why not a large consultancy?
I build myself, work directly with your teams, and hand over to your people. For specialist topics such as legal, IT security, or complex data architecture, I bring in the right partners.
How do you handle the works council and data protection?
Early and openly. The works council, data protection, and IT security are involved from the start; a clear AI guideline and firm data rules are among the first outcomes. Where AI prepares texts or decisions, a person always gives the final approval.
Do you know our industry?
Your people bring the domain knowledge; I bring the pattern. Complaints, quotes, or order handling—these are processes full of expertise, exceptions, and handoffs. I have redesigned exactly such processes at a media company, on software platforms, and in a listed financial group.
What does it cost?
It depends on the format. The intro call is free. After it, you receive a concrete proposal with scope, outcome, and price.
Interim or permanent?
Both are possible. A good path is often to start with an interim mandate and decide after 90 days whether and how to fill the role permanently—with me if you like.
Which process in your company should AI make noticeably better?
Book a 30-minute call directly or write to me. I reply personally within two working days.