Predictive maintenance
AI is increasingly being used to support predictive maintenance in commercial property by identifying potential issues before they occur, helping to reduce costs and downtime. These systems work by analysing data from various building sensors and devices, monitoring factors like temperature, energy use and equipment performance. Some commonly used sensors and devices include smart thermostats, vibration sensors, energy meters, humidity sensors, and equipment-level diagnostics integrated into lifts, HVAC units, or boilers.
These sensors and devices are often part of Internet of Things (IoT) networks, which connect physical devices to digital systems to enable real-time data collection. Machine learning algorithms, analyse current data to detect patterns and anomalies that highlight inefficiencies or risk of failure.
For example, it might flag a lift motor running hotter than usual or spot irregularities in HVAC performance. This allows facilities managers to plan maintenance activities proactively, reducing unexpected breakdowns and extending the life of assets. As the model learns from past data, it becomes more accurate overtime, enabling better forecasting and management of the property overtime.
However, the effectiveness of these systems depends heavily on the quality of the data collected. Inaccurate or incomplete data can lead to false predictions—either missing real issues or triggering unnecessary maintenance. There are also risks of over-reliance on the technology, which may lead to a false sense of security or neglect of routine inspections and preventative maintenance. In addition, integrating predictive systems with existing workflows can be complex, often requiring coordination across departments and ongoing investment in training, support, and infrastructure.
In addition, predictive maintenance systems often involve the collection and transmission of large volumes of operational data—some of which may be sensitive. For example, IoT devices can capture occupancy patterns, usage data, or behaviours that, may be considered personal or identifiable. If not properly managed, this could raise concerns around data privacy, cybersecurity, and regulatory compliance. Organisations must ensure that data is anonymised where possible, stored securely, and only used for its intended purpose. Where third-party vendors are involved, it is essential to have clear data processing agreements and safeguards in place, particularly if data is stored or accessed outside the UK.
To mitigate the associated risks, predictive maintenance should be used as a decision-support tool alongside professional judgement. Facilities teams should ensure sensors are regularly maintained and that data is accurate and complete. Staff should be trained to interpret system outputs, and traditional preventative maintenance routines should continue alongside AI-enabled strategies. It is also important to secure connected devices against cyber threats, plan for system downtime, and ensure compliance with any relevant data protection laws and industry regulations. Ongoing monitoring and review of system performance will help ensure that predictive maintenance continues to deliver value over time.
Principal author: Aaliyah Pollock