What can Illinois Basketball teach us about AI?
Adoption Isn’t Strategy
On Saturday night, the Illinois men’s basketball team punched its ticket to the final four for the first time in 21 years.
Their secret? A group of players from eastern Europe nicknamed the “Balkan Bloc”: Petrovic, Mirkovic, Stojakovic and the Ivisic twins. While other schools spent millions bidding on players in the transfer portal, coach Brad Underwood built a Balkan recruiting pipeline the others missed. Mighty Kentucky spent $22 million and failed to make it out of the first round.
Just as colleges scrambled to respond to new NIL rules, business leaders are racing to adopt AI. But they’re all running the same plays.
When everyone gets the same answer
Business success comes from doing things others don’t: different products, operating models, brand promise.
AI models do the opposite, producing the most statistically average answer from their training data. That’s why chatbots write bland, wordy prose like “Let’s delve into an ever-evolving business landscape that AI is reshaping.”
Studies consistently confirm the effect known as “flattening”: In one, 25 AI models were asked to write a metaphor about time and produced 25 versions of the same two ideas. Another found that humans using AI had fewer novel ideas months after the tools were taken away.
If every company uses AI models trained on public data, they’ll all end up with the same ideas.
Bland and predictable has its place in business. You don’t want your customer service agent getting creative with refunds or your legal AI agent inventing new contract language on the fly.
But you do want to launch products the market hasn’t seen yet, target customers others neglect and write advertising copy that catches people’s imagination.
An AI toothbrush?
Yep. The new Oral-B iO toothbrush tracks usage across 16 mouth zones and uses AI to improve people’s brushing. And it works: 82% of users found it improved their oral health.
The product idea came after six years of research and 250 patents. The critical insight was what P&G calls the “say-do” gap: In surveys, consumers claimed to brush for 2 minutes but in reality only 47 seconds.
Most people think of P&G as a marketing company. But Tide, Crest, Pampers, and Ivory were all products that redefined their categories.
Now P&G is using AI to upgrade its product development engine. They built an “AI factory” that generates new concepts and tests them with “digital twins” and synthetic research panels.
This engine is distinctly P&G’s: it sits on top of decades of proprietary data and is integrated into their human workflows.
The Advantage that Compounds
Every use of P&G’s AI factory generates new data that is fed back into the platform. Better data drives better insights, better insights drive better results, and better results generate more data. This is known as the ‘Learning Loop’, where model quality and impact improve over time.
Domino’s does the same thing, using order history from 80 million customers to predict the odds of completing an order. If they are high enough, your pizza goes in the oven before you pay for it. Later, computer vision scans every pizza before it goes out the door to ensure it has the right ingredients.
Almost any company can do this. For example, a custom manufacturer can build a model that uses order history, inventory availability and cost data to produce faster, more accurate quotes. And it improves over time as it learns which deals are won and lost.
Don’t be Kentucky
Illinois basketball has built something special. Other schools are stepping up overseas recruiting, but can’t offer what Illinois has in Champaign.
The same is true with AI. The basic use cases have positive ROI but are becoming table stakes. Companies shouldn’t mistake them for a winning AI strategy.
Remember the Kentucky Wildcats: Doing the same thing everyone else does will land you squarely in the middle of the pack.
Illinois won because they played a different game.
Dad Joke: Why do coaches hate AI referees? Because they call too many technical fouls 😂