AI in facility management works excellently. At least as long as you don't look too closely at what data it's actually based on.
Welcome to the German CAFM market 2026. A market where artificial intelligence is now ubiquitous – at least in presentations, product brochures, and strategy papers. Buildings are no longer operated there, but "intelligently controlled." Systems no longer document, they "learn." And data no longer simply becomes information, but immediately "insights." How nice...
In actual operations, however, the day often begins with a completely different question:"Which of these three area specifications is actually correct?"
Between vision and reality
The German building stock is not a cleanly modeled system. It is the result of decades of renovations, extensions, changes in operators, and improvised solutions. What is on the plan and what actually exists are two things that can resemble each other, but are not necessarily identical.
CAD plans are available in many places, but rarely fully consistent or up-to-date. BIM, if it exists at all, is mostly in the form of individual projects or an ambitious target state. Or in impressive new buildings with ample funding and high standards. And even where models exist, their informative value often ends precisely at the point where operations begin (one cannot emphasize enough the importance of clean data handover for operations – but this is often not done or done too late).
The idea of using AI based on this is not fundamentally wrong – but it presupposes something that is often missing in reality: a reliable data basis.
Data quality: The elephant in the server room
AI needs data. Not many, but good. And this is precisely where it unfortunately becomes uncomfortable. Because in everyday German CAFM, people are often already satisfied if data exists at all. Whether it is consistent, unambiguous, and up-to-date is a completely different question.
Assets are named differently, areas calculated differently, maintenance documented differently. Histories are incomplete, responsibilities not always clear, and somewhere there is almost always an Excel file that is "unofficially" closer to the truth than the system (because people don't want to give up their own little kingdom for a system without personal benefit). When AI models are trained on this basis, something fascinating happens:
The results suddenly look very professional. They are just not necessarily correct.
When 'AI-powered' is mainly marketing
The German CAFM market has quickly understood how to deal with the topic (I know this for sure, believe me...). AI is no longer a feature, it is increasingly a mandatory component of External image. Those who don't have it appear backward. Those who have it appear future-proof. So suddenly they all have it.
Reports become "AI analyses," trends become "prediction engines," and known regulations become "learning systems." The wording has evolved, but the underlying logic has not always kept pace. This is less deception than market dynamics. Providers react to demand, and demand reacts to what is considered innovative. AI is currently the currency of future viability.
The crucial question for users is therefore no longer whether a solution contains AI, but how much of it is actually substance – and how much narrative and PowerPoint.
The dangerous leap of faith
An observation from self-reflection: The more intelligent a system appears, the more willing we are to believe it. Right?
This is human – and potentially problematic in facility management. Because buildings are not standardized datasets, but complex, grown systems. Many things work not because they are perfectly documented, but because people know how they actually run..
The AI does not know this reality. It only knows what is in the system.
When it suggests optimizations based on this, they often sound plausible. And that is precisely what makes them so convincing. Plausibility quickly replaces verification.
"It comes from the AI" becomes a shortcut for "It will be fine. And disaster takes its course...
Organization beats algorithm
The biggest hurdle for AI in facility management is not the technology. It is the organization.
Because suddenly different rules apply. Data maintenance becomes critical, processes must be followed, naming must be consistent. What was previously compensated for with experience and improvisation now becomes measurable – and thus visible. This sounds like progress, but in everyday life it often feels like extra work. And this is exactly where many initiatives fail. Not because the software doesn't work, but because the organization is not prepared to build the necessary Discipline .
AI is not plug-and-play. It's an amplifier for what's already there – for better or worse.
The underestimated effort
A particularly persistent myth in the market is the idea that AI will quickly bring efficiency gains. In the long term, this is true. In the short term, the opposite usually happens.
Before systems can learn meaningfully, data must be cleaned, structures standardized, and processes stabilized. Interfaces must function, responsibilities must be clarified, and usage must be anchored in everyday practice. This takes time. And it ties up resources.
The actual AI is often the smallest part of the overall project. The largest part is preparation. Or, to put it less kindly:
You only automate after years of tidying up.
Where AI really works
With all irony: AI has its place in facility management. Where data is consistently available across many locations, where energy consumption is comparable, and where processes function stably, real added value is created. Anomalies are detected faster, decisions are made more soundly, and priorities can be set better. But this is not a starting state, but rather a Maturity Level. AI does not work well where it is supposed to solve problems that are actually organizational in nature. It works well where these problems have already been solved.
Perhaps the most important question in the current AI hype is an uncomfortably simple one: How well is our facility management really organized? Not strategically, not conceptually, but operationally.
Is data consistent? Are processes followed? Is the CAFM/IWMS system the central truth – or just one of several? The answers to these questions will determine the success of AI more than any software feature.
Conclusion without a slide (sorry)
AI in facility management is neither nonsense nor a miracle solution. It is a tool – and a very demanding one at that.
In the German CAFM market, it encounters a reality that is often less digital than strategy papers suggest. And it is precisely from this tension that the current hype arises.
- The technology is ready.
- The presentations are too.
- The data basis... is still working on it ;-)
And until then: The probably most important 'Single Source of Truth' in building operations continues to bear a very familiar name: 'final_final_v6_now_really_final(b).xlsx'. But it doesn't have to stay that way forever.
