6 Lessons for Facilities Leaders on Rebuilding Around AI: Insights From Fexa's CEO, Kurt Smith

This piece is adapted from Kurt Smith’s opening keynote at Fexa Flex NYC, where facilities leaders gathered to share what they’re seeing and learning across the industry.

In 1930, the Indiana Bell Building was set to be torn down. Eleven thousand tons of it. Then the architects looked again and said: actually, there might be another way. They rotated the whole building 90 degrees, and rewired the utilities as it moved so the telephone operators inside could keep working the entire time.

Sometimes there’s another way of doing things.

Here’s my honest read on where we are with AI. The first chatbots were mostly a novelty – a faster way to search. Then came agents, and with them real personal productivity. A McKinsey study found 80% of people using AI at work say their own productivity improved, even though the organizational impact mostly isn’t showing up yet. 

What’s coming next are agents running in the background, always on, picking up work that used to sit on someone’s plate. That’s where organizations start to feel it.

Whether that fills you with dread or FOMO, I think the conclusion is the same: get curious, start learning, start experimenting. The cars have arrived. We’re not going back to horses.

At Fexa, we’ve spent the last 12 to 18 months on the leading edge of this, because as a software company, we were disrupted first. Now, nearly every team at Fexa is building its own agents, and in a recent survey, 98% of our employees said they use AI to change how they work, and morale and satisfaction stayed high through all of it.

We had to rewire the company to get there. Here were the six lessons from doing it.

Lesson 1: Why Is AI Adoption a Time Problem Before a Tool Problem?

I’ll raise my hand on this one. I resisted it, and I was wrong.

People can’t learn a new way of working in the margins of a full day. Things only took off when we carved out real time for people to experiment, build a first agent, look at it, decide it wasn’t very good, throw it away, and try again. Maranda Dziekonski, our chief customer officer, talked me into letting her team run a hackathon. It produced practical agents, plenty of dead ends, and a million lessons.

Lesson 2: Where Should You Look First for Work to Hand Off?

Look for the repetitive, administrative busy work hiding in the crevices of your day and the pockets of your team’s week. 

Kelli Brewer, a customer success manager on our Trakref product, took one recurring quarterly exercise down to 30 minutes.

And here’s a rule of thumb from Steven Blumenfeld (VP of Asset Intelligence, Commercialization, and Partnerships) that I love: do it manually a couple of times first. That’s when you actually know the rules and the steps well enough to build the agent.

Lesson 3: How Do You Make AI Part of How Your Team Is Measured?

We put our money where our mouth is. We told every person at Fexa that if we want to lead our customers in solving problems with technology, we have to be the best in the industry at it ourselves. Every role.

Then we evaluated people on it, with honest conversations about where they were struggling and where they felt confident. One of the key criteria was how actively people share what they’ve learned with their peers. We take that seriously, and we celebrate it when it happens.

Lesson 4: Who’s Responsible When AI Gets It Wrong?

You are. Because AI can do a lot, but it can’t take responsibility. That one’s on us.

It’s tempting, and I’ve felt it myself, to get so impressed by AI that you start outsourcing a little of the responsibility for an answer to it. That’s a slippery slope. We’ve made mistakes here. Trust takes a long time to build with customers and with each other, and it can be lost very quickly. So hold the quality bar high, and don’t allow “that was the AI” as an excuse. We don’t allow it at Fexa.

Lesson 5: How Do You Know When It’s Time to Stop and Build?

Ask yourself: is this a sandwich?

At our recent company retreat, Luke Delengowski, one of our engineers, told a story about his dad and brother racing to make 200 hoagies for a catering order at the family deli. About 90 minutes. An impossible task. His brother started making sandwiches one at a time, the normal way. His dad moved the cash register off the counter, wiped it down slowly and deliberately, rolled out 30 rolls, and went down the line. Meat, cheese, lettuce, mayo.

When you catch yourself doing the same thing over and over, that’s your cue to pause and clear the table. Create a rule, design a skill, build an agent.

Lesson 6: What Part of the Job Should AI Never Replace?

Give the same recipe and the same ingredients to a master chef trained in Italy and a Japanese chef. Does the dish come out the same? No. Can you explain why? Not really.

Every job has parts like that. We know more than we can write down. Think about onboarding a new employee: can you really sit them down and explain everything about the role? The intuition, the gut calls, the judgment built over years are precious. When you design how your team works with AI, those are the things to protect.

Will AI Replace Facilities Jobs or Change Them?

I don’t know the exact order or magnitude, but I’m confident the impact will look very different from role to role. The more physical the work and the closer it sits to the built environment, the harder it is for AI to touch directly. Office and administrative work, with all its repetitive tasks, is more exposed. Even in software, demand for programmers hired to write code is falling while demand for engineers hired to solve problems is growing.

So far, the fear has also outrun reality. A year ago, 32% of people expected AI to reduce headcount at their company. Only 14% actually saw it.

The distinction I’d encourage you to make is between the task and the job. A task is a unit of work you can write down and verify: clarifying a request, chasing a vendor, scheduling a visit. The job is the responsibility and the judgment: choosing the priority vendor, making a tough repair-or-replace call, deciding which stores get more resources during a storm. When we looked at more than 67,000 workflow interruptions in Fexa’s platform, they looked a lot like that list of tasks.

History doesn’t repeat, but it rhymes. Switchboard operators disappeared because the task was the job. Spreadsheets replaced clerks, and we ended up with more accountants and auditors. The barcode let grocery stores carry far more SKUs than anyone could ever track by hand. Technology can automate work, elevate the people doing it, or expand what’s possible at all.

What Does This Mean for How Facilities Work Gets Done?

As my good friend Steven puts it, facilities management has always been an orchestration problem. And that orchestration has always run on one source: people.

That model is under real pressure. From 2012 to 2019, AC and refrigeration costs rose 18%. From 2019 to 2026, they rose 62%, and in most cases prices haven’t caught up. Open skilled trades jobs far outnumber the people available to fill them. You know the limits of this better than anyone. Adding headcount isn’t always an option, and the model creates waste in time, spend, and energy.

It also hides a silent killer: latency. When people are the ones spotting issues and pushing work through every step, each handoff adds delay. That latency is what ultimately slows the impact we can have in the field.

What AI gives us is abundant intelligence, and that lets us rethink the model. We call it orchestrated intelligence: pick up the signals your facilities and assets are already sending, apply the rules defined by the operators who know the outcomes they’re after, and give facilities managers a command center to run things exactly their way.

But none of that matters unless what we build solves real problems and saves you time, money, or energy.

Frankly, you shouldn’t have to care about the technology. It should just feel like great software.

Instead of seven clicks to find an answer, you ask: why are my refrigeration costs up in the Southeast? Which 20 stores should I worry about this week? The store is hot, create a work order. Agents work quietly in the background, surface what needs your attention, and keep a human in the loop where it matters. Early on, when customers can choose, they pick FexaAI work orders 70% of the time, because the work gets done faster.

The foundation for all of it is the insight our founder, Marc Balzamo, built Fexa on: every operator wants to run facilities their own way, with their own rules. The software has to accommodate that.

What Comes Next?

Phase one of AI made each of us personally faster. The next phase is organizations moving faster. Our job is to support you on that path, understand what you need, and build the technology that lets you define the rules that get automated across all of your facilities.

We’d love to build that future with you.

Kurt Smith is the CEO of Fexa. Want to see what orchestrated intelligence could look like across your portfolio? Request a demo.