Expert Reveals The Pros And Cons Of Maintaining Buildings Using AI

AI has triggered a shift in the facility management and built environment sectors, offering opportunities to streamline operations and enhance maintenance strategies. However, despite the promise of increased efficiency and cost savings, AI adoption comes with significant risks – especially when it comes to building maintenance schedules for high-risk buildings.

Mike Talbot, Chief Technology Officer at SFG20, the industry standard for building maintenance, has shared his insights on the dos and don’ts of using AI in facilities management.

The Dos Of Using AI

Using AI tools for deep research

AI is an invaluable tool for conducting in-depth research. Whether you’re assessing new maintenance technologies, gathering market intelligence, or researching evolving regulations, AI can quickly process vast amounts of data and extract relevant insights. In facilities management, where decisions often depend on the latest information about equipment performance, AI can significantly enhance the speed and quality of research. By automating this process, AI frees up human resources to focus on more strategic activities—such as adapting the findings to your unique operational context.

Leveraging AI to review and challenge

AI can be a powerful tool for reviewing and challenging your work. From checking for compliance issues in contracts to identifying inconsistencies in documents, AI-powered tools can scan large volumes of information quickly and flag potential risks or errors. In facilities management, AI can identify gaps in existing workflows, prompting facilities managers to revisit and refine their strategies.

Maximise asset lifespan

AI-powered solutions can extend asset lifecycles by predicting potential failures before they occur. By recommending the best timing for maintenance (not applicable for statutory requirements), AI can prevent both over-maintenance, which can cause unnecessary cost, and under-maintenance, which can lead to premature failure.

The risks of using AI

While AI offers clear advantages, its over-reliance can expose organisations to risks that could undermine maintenance effectiveness. One of the main risks is using it to create maintenance tasks and schedules, which may miss critical regulations or fail to take specific context into account.

Lack of human judgment when creating contracts or building maintenance schedules

AI-generated schedules and contract clauses can be useful tools, but they are no substitute for the judgement of experienced engineers and facilities professionals. Competent practitioners draw on years of technical training, practical experience and industry knowledge to interpret legislation, understand risk, and apply standards in context.

Without this human input, AI is prone to treating guidance as static and absolute, missing nuances that are critical to statutory compliance. An engineer or FM professional will recognise when a new code of practice affects an existing regime, when a manufacturer’s recommendation should be challenged, or when building-specific factors demand a more conservative approach. AI, by contrast, may uncritically reflect incomplete or out-of-date legislation, regulations or Approved Codes of Practice, creating an impression of compliance that does not stand up to scrutiny.

Human judgement is also essential in understanding how assets behave in the real world. Experienced practitioners know how changes in occupancy, usage patterns, environmental conditions or operational strategy will alter maintenance requirements. While AI might propose a schedule based on assumed or “ideal” usage, a professional can see when a change in activity renders that schedule inadequate, and adjust it to prevent missed maintenance, equipment failure and potential non-compliance.

Inability to adapt to unforeseen circumstances

AI systems are typically designed to follow patterns, a main reason they are beneficial in predictive maintenance. However, these systems may not respond well to unexpected changes. If a facility experiences a sudden surge in activity, a breakdown, or an emergency situation, AI-generated schedules might fail to adjust quickly enough to accommodate the new conditions. This lack of adaptability could result in improper or delayed maintenance, putting equipment and operations at risk.

Relying on outdated content in AI-written maintenance schedules

A key risk in using AI to generate maintenance schedules is that it may not keep pace with changes in legislation, regulations and standards over time. Maintenance obligations in the built environment are inherently longitudinal: statutory duties, Approved Codes of Practice and industry standards evolve, and maintenance strategies must be updated accordingly.

If an AI system is trained on, or continues to draw from, incomplete or outdated legal and technical sources, it may generate schedules that reference superseded standards or omit newly introduced requirements. This can create a false sense of compliance while exposing duty holders to enforcement action, contractual disputes, insurance challenges and increased liability – particularly in higher-risk settings such as healthcare and residential buildings.

The risk is not only about immediate accuracy, but about maintaining a defensible compliance position over the life of an asset. Where AI-generated schedules are not explicitly grounded in current legislation and updated guidance, organisations may struggle to demonstrate that their maintenance planning has followed the law as it has evolved, undermining both safety outcomes and legal compliance.

Mike Talbot, CTO at SFG20, says:

“AI, like any professional tool, delivers its best results when guided by people who understand both the technology and the realities of managing complex building assets. It offers genuine value by speeding up early research, helping teams explore maintenance options in more depth, facilitating longer asset lifespans, and increasing overall efficiency. But even with these strengths, it still benefits from clear human direction and careful review.

“SFG20 recently tested creating a maintenance schedule for an L1 Fire Alarm system within a Large Language Model (LLM) and encountered a glaring error within seconds: the reference standard the schedule was based on was out of date. The problem here is that it could potentially put building occupants at risk and leave building owners exposed to legal consequences.

“When used in a controlled, informed way, AI can enhance efficiency and help organisations get more from their assets over time. Its role is to support professionals in making better decisions, not replace the experience and judgement that ensure maintenance strategies remain accurate, safe, and fully aligned with current requirements.”