Facilities Management (FM) has never suffered from a lack of technological promises, writes Mike Gillespie, of Advent-IM, the information security and physical security consultancy.
Over the years we’ve been told that CAFM platforms would revolutionise estates management, that Integrated Workplace Management Systems would deliver complete visibility, that smart buildings would optimise themselves and that digital twins would transform operational decision-making. Each generation has undoubtedly delivered value, but rarely to the extent promised. Artificial Intelligence feels different.
Not because the technology is necessarily more advanced than previous innovations, but because for the first time we’re seeing systems move beyond supporting decisions and increasingly towards making them. The emergence of agentic AI represents a significant shift in how operational choices are generated, prioritised and executed across the built environment.
For FM leaders, this is no longer simply a technology discussion. It is becoming a fundamental question of governance, accountability, risk and organisational control. The real challenge is not whether AI works. The challenge is understanding what happens when software begins influencing outcomes that were previously determined by human judgement.
Most AI deployed within FM remains relatively narrow in scope. It analyses data, identifies patterns and recommends actions rather than operating autonomously. Yet even in this assistive form, AI is already reshaping organisations. Predictive maintenance systems identify potential equipment failures before they occur. Energy optimisation platforms continuously balance occupancy, weather forecasts and tariff data to reduce consumption and support sustainability objectives. Fault detection systems filter thousands of alarms to identify genuine performance issues, while service management tools increasingly use AI to prioritise tickets, route work orders and anticipate service failures. These capabilities are usually presented as efficiency improvements. That description is technically correct, but strategically incomplete.
Every system that determines which maintenance issue is addressed first, which asset receives investment, or which operational anomaly is prioritised is already influencing organisational risk. Decisions that were once made by experienced managers are increasingly being shaped by algorithms operating behind the scenes. The technology may still require human approval, but it is already framing the choices humans make. That distinction matters.
Despite rapid progress, most AI implementations remain constrained by traditional organisational structures. Energy systems optimise energy use. Maintenance systems optimise maintenance activities. Security systems optimise security outcomes. Workplace management tools optimise space utilisation. The problem is that buildings do not operate in silos. Facilities leaders are constantly balancing competing priorities: cost against resilience, efficiency against comfort, sustainability against operational continuity, and increasingly, security against convenience.
Reactive AI
Improving one area in isolation can unintentionally create risk elsewhere. Reducing ventilation to save energy might affect occupant wellbeing. Delaying maintenance to improve short-term budgets could increase long-term asset risk. Restricting access may improve security while negatively impacting business operations. These trade-offs have traditionally relied on human judgement. Agentic AI aims to change that.
Most AI systems are reactive. They analyse information and produce recommendations. Agentic AI is different because it is goal-driven. Rather than simply identifying issues, agentic systems pursue defined objectives and take action within the authority delegated to them. They assess conditions, determine responses, execute actions, evaluate outcomes and adapt their behaviour over time. The difference may sound subtle. It is not.
This is the point at which automation stops being a tool and starts becoming a participant in decision-making. In practice, organisations are unlikely to deploy a single, all-knowing FM intelligence. More likely, multiple specialised agents will emerge, responsible for areas such as asset performance, energy management, compliance, workplace utilisation and physical security. These agents will increasingly interact, negotiate priorities and coordinate responses. At that point, leadership attention must shift from managing operational activities to governing decision frameworks The question becomes less about what the system does and more about what authority it has been granted.
Impact on physical security
One of the most important implications of agentic AI is its impact on physical security. Historically, FM and security functions have often operated alongside each other. They share buildings, infrastructure and budgets, but remain organisationally distinct. Agentic systems have the potential to blur those boundaries significantly. Access control, occupancy monitoring, energy management, workplace analytics and CCTV data are all becoming increasingly interconnected.
Imagine a building where occupancy data identifies low utilisation in part of an estate. An agent automatically adjusts lighting, cleaning schedules and HVAC settings. Simultaneously, access permissions are restricted, security patrols are adjusted and contractor access windows modified. From an operational perspective, this appears efficient. From a governance perspective, it raises important questions.
Who authorised those decisions?
How are competing priorities balanced?
How are unintended consequences identified before they become incidents?
The more systems become interconnected, the more security becomes embedded within every FM decision rather than existing as a standalone discipline. This convergence is likely to become one of the defining characteristics of next-generation facilities management. The technology itself is not the biggest obstacle. Governance is.
As organisations grant increasing autonomy to AI systems, accountability becomes more complicated rather than less. When an autonomous agent delays maintenance, restricts access, modifies environmental controls or prioritises one risk over another, leadership must still be able to explain why that decision occurred. Regulators, auditors and stakeholders are unlikely to accept “the AI decided” as a satisfactory response. Explainability therefore becomes critical. Organisations must understand not only what decisions were made, but also how and why they were reached.
Privacy creates another challenge. As FM systems consume larger volumes of behavioural, occupancy and operational data, the line between intelligent management and excessive surveillance becomes increasingly blurred. Maintaining trust will require clear ethical boundaries alongside technical controls. There is also the issue of systemic risk. Traditional failures tend to be localised. A faulty sensor affects a single asset. A maintenance error affects a particular system. Agentic failures can propagate at scale. Incorrect assumptions, poor-quality data or poorly designed objectives can influence hundreds of decisions simultaneously before anyone realises something is wrong. That changes the nature of resilience planning entirely.
Dashboards to decision engines
The direction of travel is becoming increasingly clear. Facilities Management is moving from dashboards towards decision engines. From periodic assessment towards continuous optimisation. From disconnected operational domains towards integrated ecosystems where FM, security, safety and resilience operate as parts of a larger system. That evolution is unlikely to slow. The organisations that benefit most will not necessarily be those that deploy the most advanced AI. They will be those that combine technological capability with strong governance, clear accountability and effective leadership.
The first priority is to treat AI as a strategic governance issue rather than an IT project. Technology teams cannot solve accountability challenges on behalf of leadership. Second, organisations must recognise that data quality is no longer simply an operational concern. Poor data feeding autonomous systems becomes a strategic risk with potentially significant consequences. Third, physical security, cyber security and facilities management should be considered as increasingly converged disciplines. The future operating model is likely to depend on close integration between all three.
Finally, FM leaders need to develop new capabilities. The role is evolving beyond operational oversight towards the governance of increasingly autonomous systems. That requires a different mindset. Agentic AI does not reduce responsibility. It concentrates it.
The future Facilities Management leader may spend less time directing individual activities and more time defining objectives, establishing guardrails and managing exception-based decision making. In many respects, that is a more strategic role than the profession has traditionally occupied.
The question is not whether AI will become more autonomous within facilities management. That trajectory already appears set. The more important question is whether governance, leadership and accountability frameworks can evolve quickly enough to remain in control. Because if they do not, decisions will still be made. The only difference is that someone else, or something else, may be making them.
