How the World Cup can help you master AI
You can run the best models but still produce slop
Is this a forced trendjacking of the World Cup so I can get more views for my AI content?
Probably.
But as a football fan since 2006, I can’t help but think how managing a football team and AI models are exactly the same.
And when I saw some of the favourites like Portugal and Spain fumble games that they should easily win, it made me realise the reason why so many are struggling with AI.
The best players (models) won’t give you the results
Both Portugal and Spain had some of the best players in the world. And yet, they struggled to score against opponents that ranked way lower than them.
If you just looked at the teams on paper, it should have been an easy win for both (hopefully, you didn’t bet too much on that result).
But what happens when the tactics don’t seem to work, or bring the best out of these players?
Talent means nothing without direction.
If the manager can’t find the optimal way to use the players he has at his disposal, they will fail to win.
We have seen teams that had all the talent, but still came up short. I remember when England tried to make playing both Lampard and Gerrard work, but it didn’t produce the results because they had the same play styles.
The same happens when you’re using AI in your workflows:
You could have the best models in the world. GPT-5.5, Opus 4.8, Fable 5. But if you don’t know how to direct them to do work for you, it produces slop.
You spend more time trying to troubleshoot and correct the output than generating anything meaningful.
You need to be a better manager (director of AI)
While the manager doesn’t kick the ball during the match, they play the most important role.
They come up with the tactics and clear instructions for the players to perform on the pitch.
They tell the players exactly what they need to do in every scenario they can think of when playing against the opponent.
And in a similar way, AI needs direction from you (its manager) on how to beat the opponent (the bottleneck you’re facing at work).
Every model is like a world-class player that has all the talent in the world, but it can’t perform without your vision.
AI is excellent at execution. When you don’t give it your tactics and instructions, it starts underperforming by giving average outputs.
It’s impossible to just tell a football team with the best players in the world to just ‘score more goals than the opponent’.
And likewise, you can’t just tell AI to ‘make me 6-figures a month’ and hope that they can perform magic for you.
So if you want to master AI, you have to become a better manager:
Tell AI what outcome you want to achieve
Tell AI the steps you need to achieve the outcome
Tell AI the bottlenecks you face when executing your workflow
Start by building a system
Lionel Scaloni won the last World Cup because he built his team around Messi.
And if you want to succeed with AI, you have to build a system that lets you direct your players (the models) in the most efficient way.
Yesterday, I shared the system that I use to work with AI effectively:
I’m not locked in with one model provider and I can switch models anytime
I have full access to all my data instead of having to export my data
I create more meaningful outputs with my system (instead of just words through a chatbox)
I showed how I built this system in a livestream, and here’s the recording if you missed it: