Ai + FM = FAiM ? ? ? ?

We have all been there in life (and Facilities Management)… It comes along, promises the world, then we try and apply it and it is not so great afterall, or at least - it doesn’t live up to the hype used to sell it… 

Most “AI in Facilities Management” I see right now falls into one of three orange B&Q buckets:

1. Overpriced dashboards

2. Rebranded automation

3. Slick PowerPoint fairy tales

And yet… quietly, in the background, there are actually a few AI use cases that genuinely save time, money and headaches for our estates teams.

As FMs, some (an ex-colleague of mine) of us are stuck between pressure from above (not me)..  

“Why aren’t we using AI like that School case study I saw?”  

…. and, pressure from below; (also not me).. 

“Please don’t give me another system I don’t have time to learn anymore.”

So let’s be honest for a moment: We are in the hype phase of Ai in general, but especially in my opinion in FM. (And probably most other applications). 

In researching recently I have seen lots of Dashboards/Ai products available that promise efficiency and the world, but in reality, when tried and tested - are not there yet. 

I have tried to date (not exhaustive): 

ChatGPT - which made up some Working at Height regulations, fortunately I found it out - it then very schmoozily declared; “Oh yes, that is right, you noticed, you are right to correct me on that, it was a mistake”. This behavior stems from the model's design, which prioritizes sounding coherent and helpful over being factually accurate. 

CoPilot - which uses the underlying models from OpenAi - basically ChatGPT, this - when tested with an accident report spreadsheet, failed to give accurate calculations of counts in columns. Apparently, large files can cause performance issues, time out, or crash the application - all of which I have found testing it. Even the simple efforts I made didn’t work out particularly well, so the idea of complex assimilations are hard to trust. “What percentage of accidents are caused in the classroom environment?” …Ai:”thinking”..[some time later]…. (yawn).... Maybe it is because so many people are using it I think to myself…. ((more spinny wheel thingy)).....CoPilot: something along the lines of..”I can handle simple calculations but fail at replicating the complexity, accuracy, and auditability of enterprise spreadsheets.” Me: Oh… 

Perplexity - In my experience, less prone to error at research, but still not what is promised by the hype merchants of AI. Also, can be costly in the long run. 

I have no doubt of the potential of Ai and have found some great uses for it, but don’t think we throw caution to the wind and find actual use cases that it excels at and stick with those, whilst testing the actual limitations - as any system. Until we see reliable evidence and POW. 

What the “fad phase” looks like in reality.. 

In management theory they call this the “fashion” phase, ideas spread faster than the evidence, somewhat like a ‘90s tribute night in a dual-carriageway Harvester… tragic, but inexplicably well attended..

In FM, it looks like:

• Every CAFM suddenly becoming “AI-powered”

• Shiny “smart building” screens that don’t change a single PPM task resolve time

• Demos that focus on 3D spinning buildings, not on your leaking roof or your legionella logbook compliance

But it’s not fair to say it is all smoke and mirrors.

There are 3 areas where AI is already quietly useful – even in a typical school, hospital, university or office portfolio like yours.

Where to focus in the fad phase

If you’re an FM or estates leader now, looking at AI to apply, I’d suggest three priorities:

1. Fix your data basics - Train it with Clear building / room structure lists (also helps you to update your visio/plans). Provide it with a sensible asset hierarchy that YOU define. Next, teach it consistent job categories and priorities, don’t let it decide to run you and tell you that the crack in a window is more hazardous than the water leak pouring through a ceiling onto a plug socket in a classroom. Without getting too philosophical, the moment we hand epistemological authority over to the machine mind in its infancy, we will lose more of our mind than our money.. You wouldn’t trust a child's view on how to run your business if you manufactured guns, would you?

2. Pilot test 1/2 very specific use cases - I am training CoPilot to Job triage, using a sharepoint list, this has the potential to save hours - but must be done carefully as ascribed above. I am also testing and playing to one of Ai’s strengths - document extraction, chunking down (NLP style) large documents (my own ideas/waffle mainly) to the essence can save time, increase reader engagement and transmit info’ efficiently, but again - the caveat is: you, must do the reading, do not rely on it being accurate, not “hallucinating” (industry term) or making up laws/regs/practice. It does and will do. And finally, a simple trend analysis on a high-risk systems, again, the ground work of fixing the data is vital to glean meaningful analysis assistance.

3. Longer Term: Train your team for “AI-assisted work”, not “AI replacement”Position AI as a tool in your arsenal not a replacement for reading/writing or even, for members of your team. It can take away some admin, but be careful what... It can help you recognise and surface patterns and can help your staff (when applied correctly) to do the skilled, human bits better.

Final thought

Ai in Facilities Management right now is a bit like BMS was in the 2000’s: Lots of expensive box-ticking installs, very few people using them properly but clearly it is a massive untapped value, in its early days. 

The winners in this phase won’t be the ones who buy the flashiest system early riddled with issues, they’ll be the ones who know their estate, know their data and use Ai as a spanner in the toolbox, not some sort of grimoire powered magick wand.

As always: 

• Views my own - not my employers. 

• Many Ai’s were harmed in the making of this.