She had been using ChatGPT for three years and she still could not get started.
That is the part of this interview I keep coming back to. Chrissy McDannell is not a beginner and she is not shy about technology. She runs The Magnolia Firm, an M&A advisory that sells digital and tech-based businesses, and she has been building and selling her own companies for 23 years. She was answering her father’s HVAC business line at eleven years old, using a grown-up voice because he could not afford a secretary. She started with a snow cone stand while everybody else was doing lemonade. She sold candy bars out of her backpack. This is not somebody who waits for permission.
And for something like six months, by her own count, she saved things she never opened.
“I would save all these LinkedIn posts. I would save all these best prompts to use, and all these YouTube videos. Okay, great, I’ll save it, and this weekend I’ll go to a coffee shop and I’ll watch it and I’ll learn it and do it. And that was not happening.”
Sunday would arrive. She would look at the list. She would not go.
I have run into this in every room I have been in for two years, and I have never heard anyone with her résumé say it out loud. Usually the people who admit to being stuck are the ones you would expect to be stuck. She is the opposite case, and that makes her admission worth more than any amount of advice.
Learning is not doing, and the difference is a calendar
What she did about it was not a better system. It was other people.
“I started the club because then there’s this accountability, and you’re doing it with other people.”
The AI Implementation Club meets in person in Austin every Sunday afternoon, one to four, and it has run every week since March. Sundays on purpose: no email, no clients, no team interrupting. Three hours where the only thing on the table is building something.
Her diagnosis of the whole market fits in one sentence, and it is the sentence I would put on the wall of this summit:
“Everybody was, like, talking about AI, learning AI, talking, learning, but nobody was doing AI.”
The room she built is worth describing, because it is the thing most AI meetups get wrong. It is about half women and half men, which she notes is unusual for any AI or tech club. It is half regulars and half new people every week. And the skill range is enormous. “You will have a Google software engineer,” she said, “and you will have these beginners that are just now getting on to Claude.”
Nobody in that room is there to hear a talk. They are there because they blocked the afternoon.
What her assistant does before she opens her laptop
The workshop she is teaching at YouXAI builds the thing she uses every morning. She calls it a chief of staff, or an AI executive assistant, and what makes it worth two hours of a Saturday is that it is not a demo. It is running.
Here is what lands in front of her each day. It reads her calendar. For every person she is meeting it goes and researches them and writes her a briefing. It reports what she spent yesterday and what revenue came in yesterday. And it drafts her email.
Then there is the piece that is genuinely new, and that she set up shortly before we spoke.
Anyone who works inside these tools daily has the same quiet mess: dozens of open threads across two platforms, half-finished, pinned so they will not get lost, until the pins themselves are the problem. “My pins are so long on both platforms,” she said. “Is there a way my chief of staff can go every morning and check them out and be like, hey, this is important, you jumped to another task and you forgot about this?”
So it does. It sweeps her open threads across both platforms every morning and surfaces the one thing she dropped.
“Very accurate. Like, hey, this is number one priority that you didn’t finish yet. And it was the number one priority. I was so impressed today.”
That last line is the whole pitch for the workshop and it is not a claim about capability. It is a person being surprised by something she built.
The thing she will not let it do
Here is where she separated herself from most people who teach this.
“I don’t have mine hit send. I don’t have mine, like, booking stuff necessarily.”
She drafts, she reads, she sends. And when I described my own setup, where I open my drafts folder in the morning and find emails already written, and I asked whether she was heading the same direction, she gave the honest answer instead of the impressive one.
“Once it gets better and you’re not really needing to jump in to edit, then you can flip that switch to say, okay, I trust you, you can send or you can comment or you can post. I’m not there yet either. I’m still, like, tweaking it.”
Her frame for the whole thing is a comparison I have not been able to shake:
“It’s kind of like back 20 years ago when we were scared to put our credit card number in. That’s what’s happening.”
That is exactly the shape of it. People will not connect their calendar. Then they connect the calendar but not the email. Then they connect the email but only let it draft. Then one day they let it send. Nobody skips a rung, and there is no version of this where you argue somebody up the ladder. They climb it by watching it not go wrong.
If you have been feeling behind because you are still on rung two, she is on rung three and she runs a club about this. That should settle it.
Screen record the task. Get the manual back.
The most immediately usable thing in the interview is small and easy to miss.
You can screen record yourself doing a task, hand the recording to Claude, and get a written procedure back. She uses it both ways: to create a new one, and to clean up the enormous playbook her firm already had. “We have a big huge playbook, SOPs, that we just load in, and then it cleans it up and it updates it for us.”
But the version of this that matters goes one step further, and it changes what an operations manual even is.
“Once you really dial your process in, you say, hey, great, can you now create a skill MD file based on everything we just did? And then in the future it could update it. You could just say, hey, can you go update that skill. And so it’s almost like these skill MD files for all these different tasks are essentially now your SOPs.”
Think about what that does to the oldest problem in a small business. The reason you cannot hand off your work is that the procedure lives in your head, and writing it down is a project nobody ever starts. The traditional answer is to sit down and document everything, which is why it never happens.
The new answer is that you do the task once while recording, and the documentation writes itself into a file the AI will then execute from. The procedure and the worker are the same artifact.
What this does to the price of your company
Chrissy is an M&A advisor, so she sees a second-order effect most people building automations never think about.
Her own firm went first. “I cut my payroll about 83%,” she said, and she is specific about the scope: a small team, admin and assistant roles, done last summer, off ChatGPT alone. The reason was not strategic. It was that explaining a task to somebody took longer than doing it in GPT.
Then she turns it around on the owners she works with, and the argument is arithmetic.
A business is priced on a multiple of profit. If a company has already moved real work onto AI but kept the payroll, the cost sits in the profit line, and the multiple magnifies it. “Let’s say you’re getting a 3X multiple on your exit, but you should have cut out about $100K worth of payroll for the year. That’s $300,000 that just went, pshh, out of your pocket, gone.”
She is blunt about the cost of that advice. “It never gets easier letting people go.”
I want to be careful with this section, because it is the part of the interview that cuts against what a room full of Waco business owners wants to hear, and against what this summit argues. YouXAI’s whole premise is that the human is the multiplier. Chrissy is not making a case about people being replaceable. She is making a case about what a buyer pays for, which is a narrower and colder question, and one every owner planning an exit eventually has to answer.
Worth knowing either way. The multiple moves with automation, and it moves in both directions.
What to do with this
Two things, and only two.
Block the time. Not “learn AI this month.” A specific three hours, on a specific day, repeating. That is the only mechanism in this entire interview, and it worked on somebody with 23 years of building companies who could not make herself start. If it took a club to get her there, a note in your calendar is not going to do it on its own. Find the people.
Record yourself doing the thing you hate. One task, the one you would hand off first if you could afford to. Screen record it once. Ask for a skill file. That is a fifteen-minute experiment with a permanent result, and it is the closest thing to a free lunch in the whole conversation.
Chrissy is leading a hands-on workshop at YouXAI on September 12 in Waco. Not a talk and not a demo. You bring a laptop and you build the assistant during the session. If you have used Claude Cowork even once, you are qualified.
Some of the club is planning to drive up from Austin together. Somebody drove from Waco to one of her master classes already, which is how she found out the event was on her map at all.
Register at youxai.live. Tickets are $79 through August 31, then $99.
See you there, Fernando
P.S. She built a valuation calculator the week Fable came out — not the free kind every advisor has on their site, but one where you move the gauges and watch what cutting a line of payroll does to your multiple. She mentioned it in passing as a thing she made because she was curious. That is the posture the whole summit is about.