The Messi Test and how the Spaniards messed with it: what the World Cup final taught me about the jobs AI can't touch | Fortune

In the thrill of the game, you may be forgiven for overlooking how much of this tournament was run by tech. FIFA’s semi-automated offside system used dedicated tracking cameras mounted under stadium roofs to follow up to 29 points on every player’s body 50 times a second, fused with a sensor in the ball reporting its position 500 times a second; tickets were priced by algorithm; highlights were cut by AI; fans entered stadiums by biometric scan.

And yet, where it truly mattered, humans decided the game.

My team at Digital Planet at Tufts University’s Fletcher School and I spent this year mapping which American jobs AI could hollow out. Given the steady march of tech onto one of the planet’s most worshipped human endeavors, I watched the 2026 FIFA World Cup final through a particular lens to find the lines where man passes to machine and where the machine passes back. Let’s just call it the Messi Test.

To be clear: none of us will ever be Messi, or come close. But the human attributes he draws from live, in some measure, in all of us; strikingly, they even mark the boundary of what machines can take from us. Consider these four:

First, human work is safer when it rests on tacit knowledge. This is at the heart of what the philosopher Michael Polanyi meant by “we can know more than we can tell,”and an argument MIT’s David Autor makes to explain why judgment-heavy tasks resist codification. Messi created more chances for his team largely by walking, averaging around five kilometers a game. He found higher valued or “elite space” on the field while barely moving, reading patterns others can’t see by scanning the field. The very best scanners can get up to 0.8 scans a second. Translating the scans into positioning oneself cannot, as yet, be written into an algorithmic playbook.

Second, human work is safer when it calls for action and reaction in non-stationary, open-ended environments. Messi specializes in a game with adversarial settings with no single optimal play. He must improvise each time. AI can even conquer Go, a complex but closed game with fixed rules, but it is less capable in a game that’s open-ended in so many ways as soccer or a negotiation with an endless menu of pathways.

Third, human work is safer when human presence is key to the value. This might sound like a truism, but it is inescapable that extraordinary performers like Messi bring in the viewers. They command outsized markets, on the field, in corporate sponsorships and in inspiring new generations of players. Employers, consumers and other stakeholders want the real thing, not a machine trained on Messi’s past plays. Even OpenAI paid for Messi’s authentic face as its first global sports ambassador for a worldwide campaign promoting ChatGPT.

Fourth, human work is safer when there is a trust gap. By its own admission, FIFA’s high-tech approach to catching offsides is only “semi-automated”. The system generates an alert, but a human must validate it before the referee acts. Both humans and algorithms can make mistakes, but society does not yet trust the algorithm with the final word.

Prior to July 19, I thought I had the key attributes for a Messi Test; then Spain rewrote it. And they sharpened it. The greatest player the beautiful game may have ever produced was beaten not by a better player, but by a better system. Messi was choked by the Spanish team, finishing the final with zero touches in the opposition box and one blocked shot. This was the triumph of collective tacit knowledge over individual genius: a pressing structure, positional rotations, an unspoken sense of when to squeeze and when to hold. The task was distributed and the necessary skills were resident in no single head.

This brings me to the final attribute of the Messi Test:

Fifth, human work is safest not when the machine tries to replicate the superstar; it’s when it tries to devise a system that can suffocate the superstar. Rodri, Spain’s midfield captain, who won the tournament’s Golden Ball, was the embodiment of leadership of such a system. He didn’t score a goal, but he completed 101 of 105 passes and won 80% of his duels and across the tournament completed more passes and covered more ground than any player, thereby anchoring the entire Spanish team. He helped make the system work.

To cap it off, the Messi Test also takes the air out of a prevailing myth, which is an oversimplification of a more nuanced reality. This is a narrative that says physical work is safest from AI; after all, no matter how much better they have become, robots trained by AI still can’t fold laundry.

Consider this for a moment: Messi is among the least physical of the World Cup greats. He spends 63% to 64% of his match time walking, less than almost any other outfield player in modern soccer. What protects him are the attributes I noted earlier and not his ability to run, defy gravity or physically intimidate his opponent with his size. The implications cut both ways: not all deskwork is unsafe just as not all physical work is automatically safe ( think: warehouse sorting, farm harvesting, long-haul trucking). Our American AI Jobs Risk Index, covering 784 occupations across 530 metro areas, finds the safe/unsafe line runs through task attributes, not through the physical jobs versus cognitive jobs boundary alone.

The Messi Test should help those who are asking the question about what skills and occupations they should prepare for in an AI-saturated future. It should also help employers and executives navigate their own decisions and plan the composition of their workforce and avoid many of the traps they seem to be falling into.

For example, while Messi may be today’s superstar, he was overcome by a coordinated system; so investing in the organization is critical. A critical member of that Spanish “organization” was a teenager, Lamine Yamal, who is the product of roughly a decade in Barcelona’s La Masia academy. This means that even though it might be expedient to replace a junior associate with AI and saving money on the junior hires, as many companies now seem to be doing, ask: where does the next Yamal of the org chart come from?

Also, companies need to invest in the trust roles. This means training the humans who confirm the call and are the co-pilots. Much of the investment has gone into training AI – the pilot. As AI generates more of the decisions, the most valuable job may be of the human who validates them.

Companies must retire the reflexive “let’s re-skill the displaced workers” orthodoxy: most organizations don’t have a clear idea of what are the new jobs they will be preparing them for and the experience of re-skilling during an earlier era when automation was eating manufacturing jobs suggests that it doesn’t work particularly well.

***

While FIFA’s machines tracked 29 points on every player’s body, what made the 2026 World Cup such a thrill was how an individual like Messi brought more than an athletic body to the game. What won the World Cup was the invisible connection between the bodies that ultimately beat him. No camera has been built that can track that. So far.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.