The training session in Knysna on June 20, 2010, was open to the public. Children from the nearby township had been invited to watch some of the most famous footballers in the world. Two days later France would play South Africa, needing a big win to have any chance of staying in the tournament.
In the middle of the pitch, the captain, Patrice Evra, got into a furious argument with the fitness coach, Robert Duverne. The head coach, Raymond Domenech, had to pull them apart. Duverne threw his stopwatch onto the grass and left.
Then the players went back to their bus and shut themselves in, curtains drawn. By Eurosport’s account, the children spent a good forty minutes looking at a bus.
The team director, Jean-Louis Valentin, walked off the pitch. “It’s a scandal,” he told reporters. “I’m resigning.” Then Domenech stepped out of the bus and read the press a statement from his own players: “The players are unanimously against the French Football Federation’s decision to expel Nicolas Anelka.”
The coach of France was reading out his team’s protest.
Three days of leaks
Go back three days. At halftime of France’s 2-0 defeat by Mexico, the striker Nicolas Anelka insulted Domenech in the dressing room. Two days later the insult was on the front page of the sports daily L’Équipe, and within hours Anelka had been sent home. Evra told reporters that a “traitor” inside the squad had leaked it.
So by the day of the bus, the captain’s public diagnosis was not that the team had played badly, or that a teammate had insulted the coach. It was that someone among them could not be trusted with what was said behind a closed door.
Patrick Lencioni’s model also starts with trust, though not the same kind. Evra was talking about secrecy and loyalty. Lencioni means being willing to admit a weakness or a mistake in front of colleagues. His model is still a useful way to ask about Knysna: could those players admit an error, or say they disagreed? His book The Five Dysfunctions of a Team (2002) is a fable: a fictional company called DecisionTech, a new CEO named Kathryn Petersen, and an executive team in such disarray that it threatens to sink the business. Out of the story comes a model of five dysfunctions, which Lencioni’s firm, the Table Group, draws as a pyramid numbered from the bottom up. In its own wording: fear of being vulnerable prevents trust; “the desire to preserve artificial harmony stifles productive ideological conflict”; without clarity or buy-in, people don’t make decisions they stick to; avoiding discomfort means nobody holds anyone accountable; and the pursuit of “individual goals and personal status erodes the team’s focus on collective success.”
Trust sits at the bottom. Results sit at the top, and they are what everyone sees.
Lencioni’s bet: when a team stops delivering, the trouble started lower down, with trust.
Read through his pyramid, Knysna shows several layers failing at once. A dressing-room row ended up on a front page. The rift between the star and his coach came out as an insult, then a leak, then a strike. The players’ answer was a unanimous statement against Anelka’s expulsion. On June 22 France lost 2-1 to the hosts and went home bottom of their group, with one point and one goal from three games. The story helps show what the layers look like. It does not prove that one caused the next.
A best seller with a thin file
The Table Group says the book has sold more than 3 million copies in more than 30 languages. Independent tests are far scarcer.
One of the few, a preprint posted in August 2026 and not yet peer reviewed, surveyed 40 student project teams at Insper, a university in São Paulo, on Lencioni’s five dimensions and compared the scores with their grades. The overall link was modest. Only one dimension was significantly associated with grades: attention to results. Trust, the base of the pyramid, was not, and the authors did not test whether the layers cause one another.
One small study of students settles little. It does suggest using the pyramid as a set of questions, not a law of how teams fall.
Reading the pyramid from the top
Say your team of five misses its quarterly target, and in the review every individual’s own numbers look fine. The designer shipped her screens; the engineer closed his tickets. The easy move is a tougher target next quarter.
Lencioni’s model suggests walking down instead. Was anyone held to the shared number, or only to their own? When the plan was agreed, did people commit, or just stop objecting? Did anyone argue at the kickoff? You might find that two people had doubts in January and kept them to themselves. If so, a harder target may simply get the same silence.
A company made of agents
In 2025 a team led by researchers at UC Berkeley studied what goes wrong when AI agents work as a team. In more than 1,600 traces from seven multi-agent systems, they found 14 kinds of failure. In one trace, the agent handling a phone app never told its supervisor that the username had to be the phone number, and the supervisor never asked. In another, from ChatDev, which simulates a software company with agents playing CEO, CTO, programmer and reviewer, a chess program passed the review’s checks, such as whether the code compiled. It still had bugs, because nobody checked it against the rules of chess.
The researchers’ fix was structural. They changed ChatDev so the work could end only when the CTO agent confirmed that every review had been satisfied. On 32 programming tasks, success rose from 25.0% to 40.6%. The authors argue that many failures come from how the system is organised, “rather than the limitations of individual agents.”
Here the parallel stops. The paper treats these as failures of design and coordination, says nothing about anyone’s feelings, and the fix was a rule, not trust. What carries over is the top of the pyramid: someone has to own the shared result and check it against the real goal. Even with the new rule, most of the programs still failed.