AI stands for Artificial Intelligence. The term refers to computer systems that perform tasks which previously required human judgement, such as recognising patterns in data, making predictions, processing language or preparing decisions. Most applications today are based on machine learning: the system derives its behaviour from example data, rather than having every step hard-coded.
What is AI?
The umbrella term ‘AI’ covers several methods. In machine learning, a model identifies patterns in large amounts of data and applies them to new cases. Deep learning is a variant of this that utilises multi-layered neural networks. Generative AI, such as large language models, generates new content, such as text, images or programme code.
The EU AI Regulation defines an AI system as a machine-based system that operates with varying degrees of autonomy, is capable of adapting after it has been started, and derives from its inputs how it generates outputs such as predictions, content, recommendations or decisions.
What distinguishes AI from rule-based automation?
In building services engineering in particular, many systems are described as ‘intelligent’ even though they do not constitute AI in the strict sense. A nighttime temperature reduction that lowers the room temperature from 10 pm, or a rule that closes the heating valve when a window is open, follows strictly defined ‘if-then’ conditions. Such systems are reliable, transparent and easy to understand.
The AI Regulation makes it clear that systems which act automatically solely on the basis of rules laid down by humans do not fall under its definition of AI. AI comes into play when a system learns from data itself – for example, how quickly a particular room heats up or when, based on experience, it is occupied. Decision-makers would therefore be well advised to ask every supplier which method is actually used, what data it works with and how traceable its results are.
Where is AI used in building operations?
Typical areas of application include forecasting and anomaly detection. AI methods can use weather data and historical consumption figures to predict heating demand for the coming hours, identify occupancy patterns or detect unusual consumption trends that may indicate a fault. In the field of maintenance, such methods form the basis for predictive maintenance. In all cases, a sufficiently large and clean database is a prerequisite. Without continuously recorded measurement and operational data, even the most powerful model has nothing from which to learn.
What does the EU AI Regulation cover?
Regulation (EU) 2024/1689, often referred to as the AI Act, came into force on 1 August 2024 and regulates AI according to risk levels. Certain practices are prohibited; high-risk systems are subject to strict requirements; others are subject to transparency obligations; and low-risk applications remain largely unregulated. Since February 2025, the prohibitions have been in force, as has the obligation for organisations to ensure their staff have sufficient AI expertise when using AI.
The majority of the remaining provisions have been in force since August 2026, with specific obligations for high-risk systems to follow later. High-risk systems include, amongst others, AI systems used as safety-critical components in heating, gas, water or electricity supply networks. Whether a specific system falls under this category must be assessed on a case-by-case basis. In Germany, the Federal Network Agency is the central market surveillance authority.
Related terms
Quellen & weiterführende Informationen:
- Betterspace: Schnittstellen für mehr Effizienz und Vernetzung
- Betterspace: better.energy radiator control
- EUR-Lex: Regulation (EU) 2024/1689 of the European Parliament and of the Council
- Bundesnetzagentur: Informationen zur KI-Verordnung und KI-Kompetenz