What AI Actually Does in FM Right Now — and What It Still Can't Do

Maxcene Crowe
Apr 8
5 min read

Most AI in FM conversations are either five years ahead of reality or five years behind it. Here is where we actually are.

Suppliers are pitching AI. LinkedIn is full of claims about self-healing buildings and autonomous maintenance. Meanwhile, the FM manager sitting in a mobilisation meeting is asking a far more practical question: what does this actually do for me, today, on a live contract?

That is the question this post answers.

The Reality of AI in FM Right Now

AI in facilities management is not a future concept — but it is also not the revolution some vendors are selling. As of 2025, only 28% of organisations have embedded AI solutions in their FM operations, according to Dexterra Group. The majority are still evaluating, piloting, or being pitched to.

What AI is actually doing right now, at scale:

Analysing work order data to identify recurring failure patterns before they cause disruption — ServiceChannel's 2025 FM Trends report describes AI-powered analytics flagging issues from patterns across thousands of work orders
Optimising cleaning and energy schedules based on occupancy sensor data rather than fixed timetables
Automating dispatch — using technician skills, availability, job complexity, and priority to route work orders more efficiently than manual allocation (City FM, 2025)
Parts prediction — processing equipment symptoms and error codes against large databases to recommend parts before a technician arrives on site, reducing repeat visits

None of this replaces the FM manager. All of it requires clean, accessible data to function.

Predictive Maintenance — What Works and What the Data Shows

Predictive maintenance is the most mature AI application in FM, and the data is credible. According to Dexterra Group, approximately 35% of facility managers report that IoT-driven predictive maintenance has reduced downtime by more than 20%. In one healthcare case study, vibration analytics reduced diagnostic equipment maintenance costs by 25%.

IFMA's 2025 analysis describes the shift clearly: AI analyses historical and real-time sensor data to predict equipment failures and schedule maintenance during planned downtime — moving teams from reactive firefighting to proactive asset management.

The mechanism is straightforward: sensors monitor temperature, vibration, and performance metrics continuously. AI detects deviation from baseline. A flag is raised before failure, not after.

The constraint is data quality. Predictive maintenance AI only works if sensors are installed, calibrated, and feeding clean data into a unified system. On a contract mobilisation, this means asset data must be collected and validated early — not three months into delivery.

Space Utilisation and Occupancy Analytics

Space utilisation is the second area where AI is delivering measurable results. AI analyses occupancy trends in real time to identify underutilised areas, suggest layout adjustments, and — critically — stop conditioning, cleaning, and securing space that nobody is using.

Dexterra Group cites retail mall operators using AI-driven footfall and dwell-time tracking seeing around a 30% revenue lift through optimised layouts and adjusted service schedules. IFMA notes that AI can detect underutilised areas and provide recommendations for desk-sharing strategies — directly relevant to post-pandemic hybrid workspace management.

For FM teams managing corporate estates, the practical value is in aligning service delivery with actual usage rather than scheduled assumption.

What AI Still Cannot Do

Here is the part the vendor deck skips.

AI cannot manage a TUPE transfer. It cannot read the room when an inherited workforce is resistant to a new service model. It cannot negotiate with a client whose expectations were set by the incumbent. It cannot tell you whether a subcontractor relationship will hold under pressure.

IFMA is explicit: "temper enthusiasm — ensure GenAI is well-behaved and transparent, validate data origin and governance." The effectiveness of every AI tool in FM is determined by the quality of the data it runs on and the human judgement applied to its outputs.

During mobilisation and transition, where data is sparse, systems are being integrated, and relationships are being established, AI adds limited value unless the groundwork has been done. It cannot substitute for a properly structured mobilisation plan, a validated risk register, or a mobilisation lead who understands what the contract actually requires.

AI is a force multiplier. It multiplies what is already there — good data becomes actionable insight; poor data becomes confident nonsense.

How to Evaluate an AI Claim from a Supplier

When a supplier tells you their platform uses AI, ask these five questions:

What data does it require, and who owns that data? If it needs sensor data you do not yet have, the tool cannot function at mobilisation.
What does it actually automate, versus what does it flag for human decision? The distinction matters — flagging is useful; autonomous action needs governance.
What does the output look like in practice? Ask for a live demo on real (anonymised) contract data, not a slide deck.
Has it been independently validated? Case studies from the same vendor are marketing. Third-party or client-reported outcomes carry more weight.
What happens when the AI is wrong? Every predictive system generates false positives. Understand the failure mode before you commit to the workflow.

IWFM's Market Outlook 2025 confirms that technology investment is increasing across the sector — but investment and effectiveness are not the same thing.

Saveable Framework: The AI Readiness Check for FM Contracts

Save this for your next supplier AI pitch or contract mobilisation.

Is asset data clean and accessible? — AI needs structured, validated data — not spreadsheets from the incumbent

Are IoT sensors installed and calibrated? — Predictive maintenance requires live sensor feeds

Is there a single data repository? — Siloed systems prevent AI from functioning at contract level

Who owns the AI output governance? — Define accountability before deployment, not after

Has the AI been tested on comparable contract types? — Retail AI and healthcare AI are not interchangeable

Does the mobilisation plan account for data migration? — Data gaps in weeks 1–8 will undermine any AI tool

Develop the Skills to Lead AI-Enabled FM

AI tools are only as effective as the people deploying and managing them. These two courses give you the frameworks to make better decisions in complex FM environments:

MCFM00203.2 Advanced Continuous Improvement and Adaptability — £895 Build the operational adaptability needed to integrate new technology, manage change, and sustain performance improvement across a contract lifecycle.

MCFM00202 Developing Problem Solving Strategies — £695 Structure your approach to complex FM challenges — including evaluating technology claims, managing transition risk, and building resilient service models.