Aug 17, 2026
5 mins read
AI Future of Work
I called my car dealership this morning to change an appointment. An AI agent answered and told me it was an AI.
That last part mattered more than I expected.
My request wasn’t standard. I was supposed to drop off my car and pick up a loaner. I wanted to flip it: have them deliver the loaner and take my car.
The agent offered three options. None of them were mine.
So it did the right thing, recognized it couldn’t help, offered to transfer me, and put me through to a live agent. No loop. No dead end.
The voice was natural, not robotic. It did ask me the same question twice. And I found I didn’t really mind, because it had been upfront about what it was. When a bot pretends to be human, a glitch feels like a con. When it’s honest, a glitch is just a glitch.
The interesting part isn’t that the tech works. It’s where it’s showing up. Not a tech company. Not an enterprise pilot. My car dealership.
Which tells you something about where this is heading.
Two years ago, a voice agent that could handle a service appointment was a demo. Today it’s a line item my dealership pays for and doesn’t think about. That’s the whole arc: impressive, then expected, then invisible.
And once something becomes invisible, the question changes.
You stop asking whether the technology works and start asking what it’s actually capable of doing.
Every category goes through it. The thing that felt like a differentiator becomes the thing you get right out of the box.
Working in flex space, I watch the same clock running.
Operators field the same three calls all day: is a room free Thursday, can I get a day pass, what does an office for six people cost.
The first two are scheduling questions. Any competent AI agent will eventually handle them, on every channel, and that capability will cost roughly nothing by this time next year.
And that’s not a bad thing. It’s progress. It means operators can spend less time answering repetitive questions and more time running their businesses.
But it also means the conversation itself becomes a commodity.
The third question is more interesting.
“What does an office for six people cost?” sounds like another simple request. An agent can answer it instantly. But it can only quote what it’s been told to quote.
If the answer behind it is a static rate card, you’ve automated the reply and left the RevPAR exactly where it was.
Because pricing isn’t really a conversation problem. It’s a decision problem.
What should that office cost on Thursday? Is demand already building for that day? How much inventory is left? How has this type of space performed historically? What are customers actually willing to pay? Should the price go up, or should you offer an incentive? What happens if you hold that inventory for another few weeks?
The agent can have a perfect conversation while still making a mediocre business decision.
That’s where I think the next phase of AI gets interesting.
The differentiator won’t be whether a business has an AI agent. Everyone will.
It will be what that agent knows, what systems it can access, and whether it can turn that information into a better decision.
Two operators might both have an AI agent that can answer the same question. One can tell you that an office is available and that it costs $28,000 for 12 months. The other can understand the demand behind that room and determine that, right now, it should cost $33,000 for twelve months.
Same conversation. Very different outcome.
That’s the shift from automation to intelligence.
The conversation gets cheap. The number in it doesn’t.