AI isn’t really about technology. It’s about behavior.
When I get home from travel, Juan — my son, who is on the autism spectrum — will sometimes come to me saying this:
“Besito…” (Little kiss. I’ve learned this means he wants me to give him a kiss.)
“Do you like Spaghetti?” (He says this with the rhythm of the children’s song Do You Like Broccoli Ice Cream? — what he really means is: “Please ask me this so I can answer.”)
I answer: “Do you like Spaghetti?”
He immediately shouts: “YES I DO!!!” with a big smile on his face. That’s how he asks for his favorite pasta.
This is Juan’s way of making a request. And it’s also how I know he truly understands the meaning behind it — because his brain learned it through training with that song. If instead I ask, “Juan, are you hungry?” he might say yes… but I don’t always know if he’s connecting the words to the feeling, or simply repeating what reinforcement taught him. At home, our goal is always understanding.
And that’s the same goal the OpenAI team has with their massive pre-training and reinforcement loops. Of course, they face one big disadvantage: Juan has a human brain.
Juan is nine now. He was completely non-verbal for at least two of those years. During that time, I immersed myself in PECS (Picture Exchange Communication Systems), Applied Behavior Analysis, ABC Data, speech therapy, occupational therapy, and more. Every year, I’ve been part of negotiating his Individual Education Plan, and sitting in urgent and non-urgent meetings when behaviors at school needed a full team response. He has learned so much — and I’m proud of his progress — but the journey continues.
Here’s what struck me: the fundamental explosion of this current “AI spring” happened in roughly the same decade as Juan’s journey. As I watched the exponential growth of AI investment and capability, something clicked. The most powerful methods today — reinforcement learning, deep learning — echo the same principles I’ve lived with my son. Machines don’t just “know.” They adapt, adjust, and respond to feedback. Just like B.F. Skinner described decades ago: learning shaped by reinforcement.
That perspective shaped how I think about AI in organizations. Too often the first question is: “How advanced is this tool?” The better one is: “What behaviors will it reinforce in our teams and culture?”
Because if we don’t design for behavior intentionally, AI will shape it anyway.
I believe the future isn’t humans serving machines or machines serving humans. It’s both learning together — a kind of collective intelligence, as Thomas Malone puts it.
And honestly, I didn’t learn that in a lab or a classroom. I learned it from Juan.
That’s why I believe the real story of AI is not just about machines becoming smarter — but about how both humans and machines learn through behavior.
PS. This is the first post in my new series, The Behavior of Intelligence. In the next one, I’ll explore how behaviorism directly shaped the foundations of AI — and why that matters for the way we adopt it in LATAM — in business and in life.
Note: This post reflects my personal views. It is not an official statement from Salesforce, nor does it claim to represent the perspectives of the neurodivergent community. I share this as a father, drawing from my own journey with my son’s autism. This writing is my own; I simply used AI tools to help polish grammar.
