Do You Believe in This Thing or Not?
Every company has the same AI. Very few have decided whether they believe in it.
AnnotatedThe margins are mine: where an idea first showed up, where it went next, and the essays it talks to.
It’s past midnight, and I’m watching Moneyball. Again. I’ve seen it so many times I can mouth the dialogue. Tonight one line sounded really different to me, because it’s a question I’ve been circling for a year without the words for it.
Do you believe in this thing or not?
The line is not about baseball.
Everyone had the numbers
The story makes it easy to forget that nothing Billy Beane used in 2002 was secret. Bill James had been publishing his work since the late 1970s. Every front office in the league had the data. On-base percentage wasn’t hidden in a vault in Oakland.
The Yankees could have used it. The Red Sox could have used it. Every team had the same numbers.1
Oakland had the smallest budget, roughly a third of what New York spent. What it had that the richer teams didn’t was a general manager who had decided. The others had the information and the money. Oakland had conviction, and conviction mattered more.
Walk into almost any leadership team today, and it’s the same room. Enterprise licenses bought. Training rolled out. A center of excellence with a nice logo. Pilots in progress, some since last year. The same models are available to everyone for roughly the same price. What’s missing is the answer to Beane’s question.
Tools, training, budgets
Organizations buy tools, training, and budget lines to avoid making a decision. They prove activity, not belief. You can buy them without ever answering the question, and most companies do.
The questions asked in AI steering meetings are the scouts’ questions. What’s the ROI? Which use cases? Which vendor? How do we govern it? Those questions are reasonable. They’re also what you ask when you haven’t decided yet and want the decision to arrive by itself.
It won’t. The scouts in Moneyball didn’t lack data. They lacked willingness to let data overrule their identity. Forty years of “I know a ballplayer when I see one” doesn’t disappear because someone brought a spreadsheet.
Most resistance to AI in organizations works the same way. It isn’t about the technology. It’s about what the technology means for the people who earned their seats by being the expert in the room.
I know this because I spent years doing it to myself.
For most of my career, my headline was “Storyteller.” It was accurate. I tell great stories. That was never in question. It was also a comfortable place to hide. It got me into rooms, got nods, and asked nothing new of me.
Meanwhile, the work I actually did had changed. I built workflows, systems, and production pipelines, and figured out problems nobody had solved before. I had the tools. I had the training. I had a Midjourney account before most people knew the name. None of it mattered, because I kept asking, “Who am I?” - an identity question without an end.
It took Piotr Jakubowski to tell me I was asking the wrong question.2 I have known him for a decade, worked for and with him. He is more brother than friend, which gave him permission to challenge me.
“I don’t think that is the right question to ask. I’ll be honest. I think you are selling yourself short because you are positioning yourself as a storyteller who happens to know AI really well, versus an operator who builds systems and transforms businesses and processes.”
No one had ever challenged my label as a storyteller, something that’s pretty much written on the tag of my underpants, in case I get lost. His challenge forced me to take a giant step back, to recognize that he was right: I am a builder who happens to give great story. What was left was whether I believed it enough to shed the rusty armor of storyteller and expose the coveralls underneath.
When I said yes, everything sort of fell into place. The confidence, direction, and focus that had been in question came back. So did the appetite for the work. I saw more clearly, because I stopped filtering everything through a job title I’d outgrown.
None of my tools changed. I did.
Organizations work the same way, just with more people who have to say yes.
Belief is not faith
The AI industry isn’t short on believers. It’s full of them: vendors, evangelists, people who add “agentic” to everything and call it a strategy. That isn’t the kind of belief I mean. Blind belief is just another way of avoiding the work.
Beane’s belief was disciplined. He believed because of the evidence, and then he did the harder part: he acted on it when everyone around him had reasons not to. Belief without evidence is hype. Evidence without belief is a pilot program that never ends.
The honest ending hurts, too. The 2002 A’s never won the World Series. Belief didn’t guarantee a trophy. It changed the game, and within a few years the richest teams in baseball were copying Oakland. The believers don’t always win. The people who hold out usually lose.
The losing streak
The 2002 season started slow for Billy Beane. The system got blamed before it could prove itself. The loudest voices in the building wanted to go back to the old way. The manager kept playing his own lineup, so Beane traded away the players he preferred. Then came twenty straight wins.
That dip is where most AI transformations die. The pilot stalls. Adoption plateaus at the enthusiasts. Someone senior says the output isn’t good enough and gets agreement from people relieved to have the excuse. Six months later the company says, “We tried AI.”
They didn’t try. They tested.3 The difference is belief, and it only shows when things aren’t working yet.
The question
Every company has access to the same models, vendors, and consultants selling the same frameworks. The playing field is flat, and it won’t stay that way for long.4
What separates companies now isn’t what they bought. It’s whether anyone at the top has decided what they believe and is willing to stand by it through the losing streak.
The tools are the same for everyone. Belief is the one thing you can’t buy.5
A postscript.
The conversation with Piotr didn’t stay in my head. It became a new RockPaperScissors, rebuilt from the headline down: thirty years making the work, and now rebuilding how it gets made. Workflows, systems, and the part most AI rollouts skip, which is getting people to believe the new way is theirs. If your company bought the tools and is still waiting for something to change, that’s the conversation I want to have.