Two things tend to dissolve quickly when FM directors engage seriously with this question.
The first is anxiety about agents displacing people. When you map actual FM processes and identify the tasks that agents can carry out — ticket triage, alarm consolidation, context enrichment, meter data quality checks, predictive maintenance scheduling — you find that the work being automated ranks lowest on the list of what highly trained facility professionals find meaningful. The economics reinforce this: reactive maintenance typically costs three to five times more than preventive approaches, and unplanned downtime runs to tens of thousands of dollars per hour. The majority of that cost comes from the coordination, routing, and administrative work surrounding the repair itself — precisely the work that agents handle best.
The second thing that becomes clear is the scale of what becomes possible. When that work leaves the human layer, capacity returns. And what organizations choose to do with that capacity is where the real strategic question sits.
The capacity question
There are two directions FM organizations tend to move when they recover operational capacity from automation.
The first is insourcing. Tasks previously outsourced because the internal team lacked bandwidth — certain inspection cycles, preventive maintenance work, compliance documentation — become feasible to bring in-house. The cost reduction that follows is real and measurable.
The second is harder to quantify but arguably more significant. People freed from screen-bound, reactive work start doing the things that facility managers are actually trained and hired for. They are on-site more often. They build genuine relationships with tenants instead of communicating primarily through ticketing systems. They apply domain knowledge to problems that genuinely require it. As Erik Wallin noted in the Agentic Proptech Webinar Series, Session 3: “When manual processes are removed from your team, you can either insource work you currently outsource — saving costs — or your people work on much higher-value tasks. You leave the desk. You go talk to your tenants. That’s the real shift.”
The direction of travel here is away from FM professionals as ticket processors and toward FM professionals as operational decision-makers. The desk-bound, reactive version of the role gives way to something that better reflects what experienced FM expertise is actually worth.
What a mature agentic FM operation looks like
Buildings running structured preventive and AI-driven maintenance programs are documenting 30% lower total maintenance costs, 45% less unplanned downtime, and 38% fewer emergency capital expenditure requests compared to reactive baselines.
The organizations making the most progress are not trying to automate everything at once. They map their processes systematically and score each for automation potential against business value. They identify the subtasks that are high-feasibility and high-impact, and they start there. They treat the first deployments as a trust-building exercise as much as an operational improvement.
The standard approach is to run new agents in review mode for one to two weeks. The agent proposes actions but does not carry them out. The FM team reviews the proposals each day, building confidence in the reasoning before extending autonomous authority. When the agent’s judgment is trusted, its permissions expand — and the operational benefit begins to compound.
What emerges, over time, is a tiered model. Routine triage, enrichment, alarm consolidation, and recurring maintenance prediction run autonomously. Complex decisions — peak shaving proposals, equipment replacement recommendations, escalations involving tenant relationships — are surfaced to human operators with full context already prepared. The professional arrives at a decision, not a diagnosis.
Predictive maintenance represents perhaps the clearest illustration of what this model achieves. Rather than waiting for a component to fail and trigger an alarm, an agent monitors sensor telemetry and operational data to identify when maintenance is needed before failure occurs. Correctly implemented, AI predictive maintenance typically delivers a 15 to 30% reduction in overall energy costs alongside a 10 to 15% decrease in direct maintenance expenditure. Two hundred alarms generated by a deteriorating air handling unit become one scheduled preventive maintenance work order, planned at a time that minimises operational disruption.
The structural argument for moving now
Despite the proven ROI, 73% of facilities still rely primarily on reactive or calendar-based maintenance. The adoption gap comes from the perception that implementation requires major infrastructure changes and long project timelines — a perception that is outdated.
Once buildings are connected to a structured data foundation — such as the RealEstateCore ontology that underlies ProptechOS — deploying and testing a new agent can happen in days. The larger investment is in the discovery phase: mapping available data to use cases and defining permissions. For the first deployment, this typically takes a few weeks, not months.
The 42 open-source agents available in the ProptechOS Agency library — covering monitoring, triage, enrichment, and actuation — are deployable now, on buildings already connected to a structured data layer. The organizations that start mapping their own processes today will have a structural operational advantage over those that wait.
The strategic question worth asking now
If your FM team had 30% more operational capacity next quarter — not from hiring, but from removing the work that should not require a trained professional — what would you do with it? The answer to that question is where your agent strategy should begin. Not with a technology evaluation. With an honest accounting of where your people’s time is going, and what you would rather they be doing instead.
