Elisa-Christin Müller

For thirteen years, I have guided organizations through change, often where business units, IT, leadership, and works councils meet: reorganizations, agile scaling, post-merger integrations. The occasions have changed over that time, the dynamics have not. When work shifts, roles, status, and belonging shift with it. That was true before AI and it is true with AI.

Access and training are necessary and easy to measure. What they don't solve never shows up on an attendance list: the expert who has to rethink her professional identity before she accepts a tool. That is not a training question but identity work.

Still, AI adoption is mostly treated as a tooling question: more licenses, more training, more access. That gets the curious on board. What happens underneath, once a process is rebuilt or an agent is set up, often goes unaddressed. The long-standing expert who falls silent because their experience suddenly seems worth less. The leader who demands adoption and sets no example. The process that should be redesigned but stalls because no one has decided who owns it afterwards. The usage that moves into the shadows because there is no official path, and that actually shows where the need already is.

Where adoption is already running, a different question comes up. When business units build their own agents, activation is no longer the problem. The question becomes how many good individual solutions turn into changed processes rather than sprawl, and what happens to a role whose core moves from doing the work to reviewing AI decisions.

None of these patterns goes away with the next wave of training. It takes both at once: broad enablement and work on roles, leadership, and processes.

Translating AI strategy into measures that land in the day-to-day work of business units. Reading resistance rather than breaking it, because it usually carries information. Helping leaders model AI use rather than just demand it, including with what they can't do yet themselves. Involving the works council and those responsible for governance early, so they build trust rather than showing up as a hurdle shortly before go-live. Measuring adoption by what changes in the process (cycle time, error rate, quality of outcomes) rather than by logins.

This includes the groundwork that rarely has a cost center: clarifying whether a use case solves the right problem at all, and whether data maturity, process readiness, and role clarity can support what it promises. Distinguishing where a problem can be automated and where it needs a structural or political answer. And aligning capability building with the real use cases of business units, from foundational AI fluency to domain-specific training.

Finally, the path from pilot to standard practice: the point where a handful of enthusiastic early users becomes everyday work, with clear accountabilities, multipliers, and governance that grows with it. This is where many pilots get stuck, because they work under lab conditions and not in line operations. With agents, this becomes more pressing: when systems act on their own rather than just respond, decision rights shift, not just tasks. Then someone has to settle who decides, who reviews, and who is accountable when the agent gets it wrong.

In many of my projects, introducing new technology was what set change in motion. I managed the rollout of an HR ERP system across almost 70 sites of a research organization from the project management office. In a group-wide ERP program, I advised on change, communications, and enablement and supported leaders in new roles. Add to that large-group work with up to 180 participants and confidential coaching up to CEO level.

Beyond that, the transformation work itself: a post-merger transformation with leadership development from the executive board to team leads, carrying full responsibility for planning, resources, and budget. Guiding more than 30 international, cross-functional teams through reorganizations and the setup of new business units. A company-wide OKR process, redesigned and piloted. A multi-day onboarding format on ways of working and culture, co-developed with colleagues. Before that, organizational assessments covering over 500 employees and more than 50 structured interviews with senior leaders.

The parties involved were almost always the same: business units, IT, HR, legal, risk management, and co-determination. AI is the new occasion. The work behind it is familiar.

I write on Substack about the human side of AI adoption and about the work that begins after training and access.

• Alle machen Breitensport. Die Rolle, die KI-Adoption wirklich braucht, existiert nicht.14.07.2026
• Verantwortung für KI wird nicht vergeben. Sie wird geduldet.29.07.2026
• Der Betriebsrat sitzt schon am Tisch20.08.2026