In 1996, Pixar started on a sequel to Toy Story while every one of its proven filmmakers was busy finishing A Bug’s Life. The sequel went to a creative team that had never led a movie. “We felt this was OK,” Ed Catmull, Pixar’s cofounder, wrote later. The people who made Toy Story hadn’t led a feature before it either.
The idea was sound. Disney had first encouraged a cheaper direct-to-video sequel; Pixar wanted one quality standard, and Disney agreed to put the film in theaters. But a Pixar film is built from story reels, rough storyboards cut together with dialogue and music, and each version typically gets better than the last. On Toy Story 2, Catmull wrote, “we had a good initial idea for a story, but the reels were not where they ought to have been by the time we started animation, and they were not improving.”
When A Bug’s Life was done, John Lasseter and his team took over the sequel. “Given where the production was at that point, 18 months would have been an aggressive schedule,” Catmull wrote, “but by then we had only eight left to deliver the film.”
Pixar didn’t go looking for a better idea. It changed who was in charge of the one it had.
The studio head at lunch
Catmull opened his account of all this with a lunch. The head of a major motion picture studio told him that his central problem was not finding good people. It was finding good ideas. When Catmull later asked audiences whether they agreed, he wrote, they almost always split 50/50.
He didn’t. With the new leaders, and with every other production shut down, the crew pulled it off, and Toy Story 2 became a critical and commercial success. What Catmull took from it: “If you give a good idea to a mediocre team, they will screw it up; if you give a mediocre idea to a great team, they will either fix it or throw it away and come up with something that works.”
Choose the people before the plan. The right ones can fix a weak plan.
Jim Collins gave the idea its name in Good to Great in 2001: first who, then what. The leaders he studied, he wrote, got the right people on the bus, and the wrong people off it, before they worked out where to drive. His best-known example was David Maxwell, who became chief executive of Fannie Mae in 1981, when it was losing $1 million every business day. Asked what he would do, Maxwell, in Collins’s telling, called it the wrong first question; 14 of his 26 executives left before the “what” got settled.
A good recipe in the hands of a cook who can’t taste is a gamble. A cook who tastes as she goes can rescue a weak recipe, or quietly swap it for a better one.
What the bus can’t prove
That Fannie Mae example has aged badly. In 2008 the economist Steven Levitt noted that Fannie Mae, one of Collins’s eleven great companies, looked set to need a federal bailout, and that its stock had lost over 80 percent since the book came out. A portfolio of all eleven, he wrote, looked like it would have underperformed the S&P 500. His deeper point was about method. Collins had picked his eleven out of 1,435 companies by their results, then looked back for what they shared. Books like this, Levitt wrote, are mostly backward-looking, with an implicit promise that the same habits will keep a company winning.
Catmull’s story has limits of its own. It is one film, told by the studio’s own president, and the rescue had a price he names himself: the crew worked “inhumane hours,” and lots of people suffered repetitive stress injuries.
What survives is narrower than a law. When a plan is likely to break, the people holding it decide whether it gets fixed.
The project with whoever was free
Imagine you’re told to launch a customer portal by March and handed whoever is free this quarter. The scope is already written. You start assigning tickets.
Six weeks in, the requirements turn out to be wrong in a way only someone who has talked to customers would spot. Nobody on the team has. The plan rolls forward as written because nobody holding it is in a position to notice.
Spend the first week differently and the same March deadline looks another way. Two people who know the customers, chosen before the tickets are split, can rewrite a wrong scope in week six. A perfect scope gives you no such protection.
An agent in the seat
In April 2025, Shopify’s chief executive, Tobi Lütke, posted an internal memo on X. “Before asking for more Headcount and resources,” it said, “teams must demonstrate why they cannot get what they want done using AI.” Then a question: “What would this area look like if autonomous AI agents were already part of the team?”
That puts software into Collins’s question about who is on the bus. The memo states a policy; it doesn’t report how the agents performed.
Catmull’s test still applies to whoever, or whatever, fills the seat: given a flawed brief, does it notice? Try it before you rely on it. Give the agent a real task from your own work with a mistake planted in the brief, a figure that contradicts another or a requirement that can’t be met. Check whether it points out the problem or builds around it. Run it a few times, since the same prompt can give different answers, and judge each result against what the task actually needed.
Pixar drew its own conclusion. After Toy Story 2, by Catmull’s account, its development department stopped coming up with ideas for movies. Its job became finding individuals who would work effectively together, and helping directors improve ideas of their own.