How AI is impacting productivity in FM

Anthropic recently published a research note where they look at 100,000 real conversations with their model Claude and ask a simple question:

How much faster does work get done when people use AI?

Their answer is slightly uncomfortable reading if you work in real estate or facilities.

Across all those conversations, Claude estimates that the typical task would take around an hour and a half without AI ... and that AI cuts about 80 percent of that time. When they extrapolate this to the wider US economy, they find a world where current AI could lift labour productivity growth by 1.8 percent a year over the next decade if it were adopted broadly. That is roughly double recent trends.

So yes... some people really are getting a four day week out of this thing.

Anthropic’s workers vs our world

The really interesting bit is where the time is being saved. People bring weighty tasks to Claude:

• Legal and management work that would have taken nearly two hours
• Financial analysis, curriculum design, document drafting, all cut down to minutes
• A typical task worth around 55 dollars of expert time in the old world

This is AI aimed squarely at knowledge work that happens in documents, spreadsheets and code.

Now think about a day in a facilities operation:

• 200 new work requests from tenants
• 20 quotes to review, argue with and approve (or at least sense check)
• 5 compliance visits to schedule and document
• Nobody really sure how many assets are actually on site

On paper, it is exactly the kind of messy, repetitive, information-heavy work where an 80 percent time saving would be transformational.

In practice, a lot of FM teams are still printing PDFs.

The productivity story they are not quite telling

Anthropic are careful about caveats. They acknowledge that their method only sees what happens inside the chat window. It does not see:

• The time a surveyor spends visiting a property
• The back and forth with a contractor who still prefers WhatsApp to portals
• The hour spent explaining to finance why the budget line for “reactive plumbing” is on fire again

So their numbers almost certainly overstate real productivity gains today.

But you do not need the full 1.8 percent to change an FM business. If you can:

• Cut quote preparation from 40 minutes to 8
• Generate compliance packs in seconds instead of evenings
• Have an AI watch your jobs and suppliers for you, and only nudge you for the exceptions

... you are not just “slightly more efficient”. You are running a different operating model.

Bottlenecks and boiler rooms

One of my favourite bits of the paper is their discussion of bottlenecks. AI turbocharges some tasks and barely touches others. For software developers, AI speeds up coding, testing and documentation, but it does not attend the stand-up or negotiate priorities. For teachers, it helps with lesson planning, not playground duty.

In facilities, the pattern will be even sharper:

• AI can write the method statement and chase the expired gas certificate
• It can analyse a year of job data and tell you which supplier is quietly eroding your margins
• It cannot yet crawl into the ceiling void to find the leak
• It cannot convince the board to fix the ageing plant instead of patching it again

It is worth adding a single word to the last two points ... 'yet'.

If we are not careful, AI will make the digital parts of FM beautifully fast while leaving the physical and organisational bottlenecks exactly where they are. The paperwork will be instant ... the lift will still be out of service.

What this means for FM and real estate leaders

So what do we do with a research paper that says “AI might double productivity” while we are still arguing about whose job it is to update the asset register

A few pragmatic thoughts from what we see at Cleverly:

1. Treat AI as a junior ops manager, not as Google on steroids The Anthropic data is all about meaningful tasks, not trivia. If your team only uses AI to tidy up emails, you are aiming far too low. Point it at workflows like work order triage, quote review, supplier performance, compliance pack generation.
2. Use AI to make the invisible visible Our sector is full of work that is done in email, in heads and in “how we have always done it”. Getting a clear view of where the time actually goes is half the battle. The nice thing about automating workflows is that you finally get data about them.
3. Expect the bottlenecks to move, not disappear If you take 80 percent of the admin out of a job, you discover what is really slowing you down. Often it is approvals, data quality or a single overloaded person in the chain. AI will highlight those edges brutally. That is a feature.
4. Design around the site visit, not instead of it For all the talk of digital twins, someone still has to turn up, climb the ladder and sign off the works. The smart play is to wrap that visit in automation: pre-visit packs, route optimisation, automatic report generation, triggered follow ups. Let humans do the bits that require judgement and a hi-vis.

Buildings as part of the productivity story

Anthropic’s paper is ultimately about macro numbers and growth rates. Hidden inside it is a quieter claim... that current AI already delivers huge time savings in the right hands.

Real estate and facilities have all the ingredients to benefit:

• High volumes of repeatable processes
• Expensive labour and ever-thinner margins
• Customers who notice when you get it wrong

We built Cleverly on the assumption that buildings should participate in this productivity story rather than sit outside it. That means AI woven directly into work orders, quotes, assets and suppliers... not yet another dashboard nobody logs into.

If AI is capable of adding 1–2 percent to national productivity by helping people write documents and code, it will be very interesting to see what happens when we point it squarely at blocked drains, fire doors and maintenance backlogs.

The research labs have done their bit. The question for our industry is whether we are ready to let our buildings catch up.

#workingcleverly