My AI Agent Scheduled a Meeting With an AI Founder — And He Had No Idea

A few months ago, I had a call with Lucas Siegel. He’s the co-founder of Yuna, an AI mental health platform with a $30M valuation and over 50,000 users across 155 countries. He builds AI products for a living. He thinks about this stuff constantly.

Near the end of our call, I told him something that surprised him.

“The email coordination we did to schedule this call — that wasn’t a person on my team. That was an AI agent.”

He paused. “I had no idea. I just figured it was someone you worked with.”

The agent had replied to him within 60 seconds. The total cost to coordinate the meeting: 11 cents.

This is the story I tell when someone says “but AI can’t handle the human-feeling stuff.”

The Setup

I’ve been running an AI coordinator called Linda for months. Linda is built on Lindy — a no-code AI tool I use for several automations. The instructions I gave Linda were written out once, in plain language: here’s how to check my availability, here’s how long meetings should be, here are the times I prefer, here’s what to do if someone wants to reschedule.

That’s it. From that point, Linda handles the back-and-forth.

When someone emails asking to set up a call, I CC Linda. She checks the calendar, replies with available options, confirms when they pick a time, sends a calendar invite, and handles any changes if something comes up. The whole thing runs without me.

What Used to Take 20-30 Minutes

Here’s what meeting coordination actually involves: reading the initial email, checking your calendar across days and time zones, writing a reply with options that work, waiting for their response, confirming, sending a calendar link, and sometimes going back-and-forth another two or three times.

For a typical meeting, that’s 20-30 minutes of email overhead, spread out over a day or two. Multiply that across a week of scheduling and it adds up fast.

Linda does all of that in under a minute, for 11 cents per meeting.

The number that gets people isn’t the 11 cents — it’s that Lucas didn’t notice. He runs a company built on AI. He understands the technology. And the experience still read as human to him. Not because it was tricky, but because it was just… fine. Normal. Exactly what you’d expect from a competent assistant.

The Real Lesson

I used to think the benchmark for AI handling a task was “can it do this perfectly?” That’s the wrong benchmark.

The right benchmark is: does the person on the other end have a normal experience?

Linda makes minor errors sometimes. Occasionally the phrasing is slightly off, or she suggests a time I wouldn’t have picked. But in practice, none of that matters. People just treat it like communicating with a competent assistant. They reply, they confirm, they show up.

The bar for replacing calendar coordination isn’t perfect — it’s good enough that nobody notices. Linda cleared that bar a long time ago.

One More Detail

For a while, I had Linda reply as fast as possible. Then I noticed something: when she replied in under 60 seconds, some people got suspicious. The turnaround was too fast. Nobody expects a human assistant to respond in under a minute.

So I added a three-minute delay. It now waits a bit before sending, which makes the response feel more natural. That’s the whole fix.

This is a good example of how AI implementation usually works. The first version works technically. The second version works humanly. The adjustment is almost always small.

What This Means for Your Workflows

The question I get most often is some version of: “where should I start with AI automation?”

Calendar coordination is one of the highest ROI places to start, and here’s why:

It’s well-defined. The instructions are the same every time: check availability, send options, confirm. There’s no ambiguity about what success looks like.

It’s repetitive. Most people schedule dozens of meetings a week. Every one of those is 20-30 minutes of admin that can be automated.

It’s already good enough to pass the test. Lucas didn’t notice. That’s the whole argument.

If you’re not using AI to handle scheduling yet, this is a reasonable place to start. Set up an agent with your availability preferences and scheduling rules. CC it on emails when someone asks to meet. See what happens.

The 4-Day AI Sprint walks through exactly this kind of setup — starting with high-ROI, low-risk workflows like this one and building from there.