The Unique Challenges of AI Adoption Within FM

Unless you've been hiding under a rock for a few years, you'll have read articles about how AI is going to take over the workplace, everyone's job is being replaced, and that it's the only future. But how much has AI actually changed your job today? Surely we just need to buy a service, and your business is 1000% more efficient right?

I've worked in IT for over 20 years, and whilst AI is a big thing, it's just another in the latest line on technology revelations we've had. Whilst it pains me to say, I'm older than Google, and 20 years ago the internet wasn't mass market - today you'll struggle to find any business without an online presence.

So with this in mind, why isn't AI revolutionising the FM industry today?

That's right, even with investment into enterprise systems - data is largely manual, unstructured data. You'll have a work order, it's on a CAFM system that has a certain amount of structured data (customer & site details, perhaps asset details), but most of the detail is in a large text box.

Why is this important for AI usage? Think of AI as an employee, a very logical employee. If they're presented with spreadsheets of structured data, they'll work out patterns and be able to manipulate the data quickly. If the data is unstructured, it's more difficult to use, and more difficult = more errors.

Now the good thing is AI can be used to quickly structure data, but unless you have a data structure in place in your CAFM system - AI is trying to structure your data for you. It'll work, mostly, but it makes the job harder.

If there's one thing I've discovered in my time in employment - people don't like change. Engineers want to, and quite rightly so, focus on their discipline - installing and fixing things, physical things. However we need to caveat that in 2026, in that practically everything is a computer now - even my washing machine has hundreds of sensors and is internet enabled.

To implement change, the replacement system or process needs to be better than the one before, and people need to see and understand that. People are slapping AI labels on all kinds of software, but when you scratch the surface - a lot of the time there's a reason people use the process they have done for years. It works.

The majority of claims you'll hear from AI revolution are from companies with little link to the physical world. Tech companies that work in software, lawyers who work in knowledge - but FM is a different beast, it's physical.

When you're transforming knowledge work, it's all on a computer. Naturally physical transformation lags behind, the cost of a new AI enabled BMS system might not ROI for a decade, putting in 1000 new temperature sensors in a warehouse might not make a saving.

AI can do a lot of things, but in 2026 it's not able to rewire a building, or fix a roof.

FM is fairly unique in that is has a physical liability. If gas boilers aren't serviced to compliance, that's dangerous. There's a reason standards like SFG20 are commonplace, and UK governing bodies such as Gas Safe & NICEIC exist - to keep us safe.

In a lot of industries, tech included, you can give AI a long lead and see what it does. If it gets it wrong, you try it again, and repeat until you see better results. The consequences of failure are a bit of lost time, and some disk space.

When the consequences are potentially fatal, guardrails need to be put in place, and implementation will naturally be slower. This is why we've not seen the healthcare AI revolution yet.

Whilst not unique to FM, for different contracts there are different locations, assets, SLA's, compliance standards, reporting requirements and more.

We touched on earlier that in SME's this is largely manual, the alternative being an enterprise system where as much time is spent on system setup as doing the work itself.

That's not to say AI won't help with this - in fact RAG models are perfect for extracting data into structures, but you need someone within the business to understand this. As you'd imagine the kinds of companies that can afford this are larger enterprises, but as the cost of AI comes down and it's implemented into CAFM this will start to drip down to SME's.

FM is a niche business, largely manual. AI models are trained using masses of data, that's why you'll find AI to be a font of knowledge at computer programming (as there's a wealth of articles online to learn from), but FM data is typically either behind a paywall or simply not publicly available.

Of course if you've someone in your SME that understands AI model training and coding, you can create an FM specific AI model that'd be incredibly accurate - but for most SME's that's completely out of reach. In an era what AI & data are king, in FM, with typically low margins, you probably can't afford to hire an AI data engineer for six figures! As FM is fairly niche, no-one else has hired that data engineer, unless they're building it into a CAFM system to sell back to you.

AI is everywhere, and FM, with its unique challenges, create a barrier for entry for implementing AI into all but the largest systems. As time goes on this will improve, likely drastically, but in 2026 if you're working with AI in an FM SME, you're leading the way.

This will change, likely very quickly. The heart of every revolution is getting the right people involved to build a project, and with AI these people are currently in demand at a premium, likely pricing out most FM companies.

That's why FM has always bought software, rather than built. Ideas aren't the problem, it's getting the staff to build the platform, and implement the ideas, that's the challenge. Without the right people, there's no option to build bespoke systems to FM's unique processes, only buy systems, or change processes. As clever as AI is, there's still the 'person' aspect to technology.