AI Gets You in the Game. Practices Keep You at the Table.
The buy-in is cheap. Staying in the game is not.
Happy Friday friends, this week from Miami Beach!
You may have noticed there was no letter last week. I’ve been bouncing all over North America, and the writing chair never stood a chance. I’m curently delayed in FLL airport, and I owe you one. So, here we go…
Picture a card table. Not the glamorous kind from the movies. The kind in a buddy’s garage, with folding chairs and somebody’s cousin who swears he’s “feeling lucky tonight.”
Anyone can sit down. The buy-in is twenty bucks.
And for the first few hands, you can’t tell who’s good. The cards do the talking. Somebody’s cousin wins a big pot on a lucky river and starts giving lessons.
Come back at 2 a.m. and look at who still has chips.
That’s AI in business right now. The buy-in has collapsed. Everybody’s at the table. Everybody’s feeling lucky.
So if everyone can afford the buy-in, the buy-in can’t be the advantage.
Let’s break it down.
The buy-in gets you a seat. That’s all it gets you.
Two years ago, having AI in your business made you interesting. Today it makes you average. The chatbot your ops manager uses is the same one your competitor’s ops manager uses. Same model, same price.
The numbers tell you how this is going. MIT’s GenAI Divide report looked at enterprise AI pilots and found 95% deliver no measurable P&L impact. That’s after $30 to $40 billion invested. A Writer’s 2026 survey found 75% of executives admit their AI strategy is more for show than for actual guidance.
Three out of four executives know their own AI strategy is theatre.
They bought a seat at the table and called it a winning hand.
The boring work makes you good at the start.
Here’s what separates the players from the cousins in the first hour: NOTHING FLASHY! Knowing the rules cold. Folding the garbage hands. Counting what’s in front of you before betting what isn’t.
In business terms, that’s the unglamorous foundation work. One documented process instead of five vague pilots. Clean data the tool can actually use. A baseline number, hours per week, dollars per month, written down before the AI touches anything. Read-only access first. A human reviewing every output before it reaches a customer.
I wrote a whole issue about this back in February (2026.05: AI Infrastructure Beats Model-Chasing) and the thesis hasn’t aged a day: the real work isn’t in the shiny capabilities, it’s in the boring foundations. McKinsey’s data agrees. Redesigning the workflow is the single change most correlated with bottom-line impact from AI, and only 21% of companies have done it. The other 79% bolted a faster engine onto a cart with square wheels.
The boring work is how you stop donating chips. It wins you the early hands.
But it doesn’t keep you at the table.
The practices keep you there
A long session doesn’t reward the player who learned the rules once. It rewards the player who makes the same disciplined decisions at 2 a.m. as at 8 p.m. Hand after hand. Especially when tired, especially when up big, especially when the cousin just sucked out on the river again.
That’s a practice. Not a setup task you finish. A behaviour you repeat until it compounds.
BCG puts numbers on it with their 10/20/70 rule: AI success is 10% algorithms, 20% technology and data, and 70% people and process. Most companies spend in exactly the reverse order. The foundation gets built, the press release goes out, and then nobody shows up for the repetitions.
What do the repetitions look like? Reviewing the exceptions every week, not just at launch. Tracking that one baseline number every Friday and making the call: expand it or kill it. Keeping a person at the checkpoint for anything that touches a customer or your books. And keeping the judgment calls. The ones you get paid to make. I wrote about that in May (2026.17: Let the Machine Lift. You Lead.): AI compresses time, it does not compress responsibility.
It’s circular but true: we become what we do. A company that reviews, measures, and decides every week becomes the kind of company that can trust its AI. A company that bought licences and moved on becomes the cousin, telling stories about that one good hand in February.
This week, ask yourself one question: which of the three are you actually doing?
In the game? That’s the licence you’re paying for.
Good at the start? That’s the one process you documented and baselined.
At the table at 2 a.m.? That’s the weekly review you’ve run for 90 days straight.
Most leaders I talk to are honest about it once they see the three stages laid out. They bought in, they did some setup, and the practices quietly died by week six.
Hit reply and tell me where you are. In the game, good at the start, or still at the table. No judgment. I’ve been the cousin too.
See you next week.
JT