On February 7, 1994, Michael Jordan announced that he was going to play professional baseball. He would turn 31 that month. He had not played the game since he quit two games into his senior year of high school.
On April 8 he made his debut for the Birmingham Barons, a Double-A minor-league team. For a few weeks it looked as if talent simply travels: he got a hit in 13 games in a row.
Then it stopped. In May he batted .165, about one hit in every six at-bats.
His manager, Terry Francona, and hitting coach, Mike Barnett, now had a beginner failing in public, and his name was Michael Jordan.
Closer to the plate
Barnett started with the feet, and rebuilt the stance Jordan had brought from spring training: “I moved him closer to the plate and got him to stride forward into the ball and use his hips and those long arms of his.”
Jordan took it. “Michael would go after it five times a day,” Barnett said. He was in the cage before breakfast and again after games, until his hands bled.
The hits still didn’t come.
Francona recalled a stretch, a couple of months in, when Jordan got “real frustrated.” So the coaches changed what they were doing. “Mike Barnett and I sat down and talked to him,” Francona said. “I told him, ‘Just try to enjoy getting better,’ and he did that.” They pointed to what was improving: Barnett rated his throwing arm 20 on the scouts’ 20-to-80 scale in spring, 50 by August.
That autumn, in the Arizona Fall League, against better pitching, he batted .255, about one hit in four. “I do think with another 1,000 at-bats, he would’ve made it,” Francona said.
Two dials
Different people need different handling. So does one person on a new task.
Paul Hersey and Ken Blanchard built a model on that idea while writing a textbook together at Ohio University; Hersey’s company dates it to 1969. It asks you to read a person on one specific task for two things: can they do it, and how much do they want to? Then you set two dials, how much direction you give and how much support.
In Blanchard’s version, an eager newcomer who hasn’t shown skill yet gets directing: clear goals and close watching. The next stage, in his words, “usually sets in after an individual has been unsuccessful at a new task,” when enthusiasm “turns to discouragement and insecurity.” That person gets coaching, which keeps the direction and adds explanation and encouragement. Later come supporting (decide together) and delegating (step back).
One remembered season can’t confirm a stage model, and his coaches never said they were using one. What Jordan’s summer does is show the first two styles in action: exact instructions while he was eager, then a talk, on top of the instruction, once the slump set in.
The word doing the work is “task.” Hersey’s company says the approach judges readiness task by task “versus typecasting employees.” Whether the rest of the model holds up is another matter.
Thinner than the posters
The model fills training rooms. The evidence is patchier.
Robert Vecchio tested it in 1987 with high-school teachers. According to a 2000 review, he found support for what the model prescribes for the least ready followers, partial support in the middle, and none at the top. A 1992 replication with nurses leaned the same way, but not significantly. In 2018 Geir Thompson and Lars Glasø still called it “a very popular but as yet under-researched theory.” In their Norwegian data, its prescriptions held when the boss and the employee agreed on how ready the employee was.
So the firmest part is the part Barnett used first: people new to a task tend to do better with clear direction than with freedom. The neat four-stage choreography rests on less.
The second migration
Say one of your engineers, five years on the team, ran last month’s database migration, and it took the site down for an hour. Another migration is due this week. She knows the job better than anyone you have. At your one-on-one, she asks whether someone else could take this one.
Hand it to someone else and she learns that one bad night cost her the job. Hand her a step-by-step checklist and she hears that you think she forgot how. Her skill is fine. What has dropped is her confidence, and that’s the support dial.
The supporting version takes half an hour. You ask what she would do differently this time and mostly listen. The two of you agree on how to roll back if it goes wrong. Then you tell her you’ll be on the call when it starts, and that the plan is hers.
Readiness in minutes
Machines have the same split, and someone has measured it. In 2025 the research group METR timed skilled people on software tasks, then gave the tasks to AI models. The models of the day had “almost 100% success rate on tasks taking humans less than 4 minutes,” and succeeded “<10% of the time on tasks taking more than around 4 hours.” The same model was an old hand on short jobs and a beginner on long ones.
Anthropic’s guide for its coding agent, Claude Code, sets the direction dial by task in just this way. For a small, clear change: “If you could describe the diff in one sentence, skip the plan.” When the change spans several files or the code is unfamiliar, it suggests having the agent explore first and write a plan you can edit before it touches anything. That is directing.
The support dial is where the analogy stops. Francona’s talk was aimed at Jordan’s commitment, which had sunk with his average. METR’s success rates turned on task length, and nothing in them looks like a slump in morale, so with an agent the support dial has little to turn.
A one-line fix gets no plan. A four-hour change in code the agent has never seen gets told where to stand.