Most companies approach AI transformation the wrong way around. They start with the technology, which model to use, which tools to deploy, which vendor to partner with. They build a roadmap. They present it to leadership. And then, six months later, they wonder why adoption is stuck at 20% and the problems they were solving are still there. 

Julian Hagenschulte, Chief Digital Officer at CargoBeamer, has spent nearly 20 years leading digital transformation programmes across industries. Speaking at railXchange 2026, he put it plainly: in his experience, the technology choice is almost never the deciding factor. What determines whether a transformation succeeds or fails is the human and process component, and in complex, asset-heavy industries like rail, that gap between strategy and reality is where most initiatives quietly die. 

The seductive clarity of a roadmap

There is something reassuring about a well-built digital strategy. Ambitions defined, stages sequenced, dependencies mapped. CargoBeamer’s own strategy, which is to become a Native Digital Intermodal Champion by 2030, is genuinely ambitious: AI-based operating models, digital terminal twins, automated disruption management, agent-based decision support across the transport chain. 

But Hagenschulte was candid about the distance between a strategy on a slide and a strategy that lives in an organisation. When he asked the room how well-known their own digital strategies were among their employees, the honest answer from most was: not well enough. 

This is not a CargoBeamer-specific problem. It is the defining challenge of digital transformation in industries where processes have grown over decades, where the asset base is old, where regulatory frameworks are slow, and where the people closest to operations have been doing the same job for twenty years. In rail, that description fits almost everyone in the room. 

The technology arrives. The strategy exists. And then it meets the organisation. 

Disruption is the norm, not the exception

CargoBeamer’s core business is moving truck semi-trailers over the Alps by rail, a high-frequency, terminal-intensive operation where disruptions are not occasional events to be managed but a near-daily reality to be absorbed. 

That context shapes everything about how Hagenschulte thinks about AI. The question he keeps returning to is not “how do I automate this process?” but “how do I make better decisions faster when things inevitably go wrong?” That reframe matters. Optimising for a world that runs to plan is a different engineering problem from optimising for a world that constantly deviates from it. 

This is also why data sharing across operator boundaries is not a nice-to-have for CargoBeamer, it is an operational necessity. A disruption at a terminal in Zurich affects a departure in Lyon. If the infrastructure provider’s data stays in their system and CargoBeamer’s stays in theirs, both sides are making decisions blind. The first project where Hagenschulte sees genuine cross-operator collaboration paying off is exactly this: feeding real-time infrastructure data into terminal processes early enough to act on it rather than react to it. 

The technology for this exists. What requires active work is the willingness to share, and the trust that sharing doesn’t mean giving something away. 

The AI Agent Factory: two weeks from problem to prototype

One of the more concrete things Hagenschulte described was CargoBeamer’s internal approach to scaling AI without scaling bureaucracy. They call it the AI Agent Factory, a structured process with a deliberately provocative value proposition: from a clearly defined business problem to a working agent MVP in two weeks. 

Not every problem fits. Not every two weeks delivers. But the intent is important. It forces three things that typically don’t happen together in large organisations: a specialist who owns the problem, a digital solutions team that can build the concept, and IT that ensures security and integration. The failure mode it’s designed to avoid is the one where AI use cases sit in a backlog for months because no one owns the transition from idea to implementation. 

The deeper point is about culture as much as process. Hagenschulte described opening a channel where any employee can speak in a business challenge via voice input, and have it logged, reviewed, and followed up by a solutions team. The message that sends to the organisation is not just “we have AI tools.” It is “your daily frustrations are worth solving, and we have a mechanism to do it.” 

That is a different kind of adoption lever than a training programme or a mandated tool rollout. 

What actually needs to change

Hagenschulte closed with the observation that rail, as an industry, is not uniquely bad at collaboration or digital transformation. Other asset-heavy sectors, construction, real estate, infrastructure, face the same incentive structures, the same silo tendencies, the same gap between what is technically possible and what organisations actually do. 

What is different about the current moment is the pace. Earlier digital waves gave companies years to adapt. This one doesn’t. 

The implication is not that every company needs to immediately have 80% of employees using AI tools daily, or a fully mapped agent strategy to 2030. It is that the companies spending time now, building the cross-functional habits, creating the low-friction channels for bottom-up ideas, doing the uncomfortable work of sharing data with partners they also compete with are accumulating an organisational capability that compounds. 

Technology can be procured. That capability cannot. 

Julian Hagenschulte is Chief Digital Officer at CargoBeamer, an intermodal operator specialising in semi-trailer transport by rail. He spoke at railXchange 2026 in Frankfurt.