On the evening of April 24, 1980, eight helicopters took off from the carrier Nimitz to help free 53 Americans held hostage in Tehran. Six reached the refueling site in the Iranian desert that the planners called Desert One, between about 50 and 85 minutes late. Then one of the six had a hydraulic failure, leaving five against a firm minimum of six. The mission was aborted.
While one helicopter was repositioning so another could top off its tanks, it collided with a refueling C-130. Eight crew members died.
The review group the Joint Chiefs set up afterward found that the Desert One plan had not been fully rehearsed and “was soft.” There was no identifiable command post for the on-scene commander. In post-mission interviews, helicopter pilots said that in some cases they did not know or recognize the authority of those giving orders. James Locher, the Senate staffer who later led the work on the reform law, quoted the senior commander there, Air Force Colonel James Kyle: “four commanders at the scene without visible identification, incompatible radios, and no agreed-upon plan.”
On paper, the chain of command had been fixed long before.
A law nobody carried out
In 1958, Locher wrote, Congress removed the service chiefs from the operational chain of command and gave unified commanders “full operational command of assigned forces.” But “those provisions were not effectively implemented.”
The chart said one thing. Everything around it said another. Promotions were decided by each officer’s own service. In the Navy of the mid-1980s, Locher wrote, joint duty was considered the “kiss of death.” The head of Central Command testified that there had not been a single volunteer for any of the thousand billets on his headquarters staff.
A new org chart alone changes little. Strategy, process, pay and people have to move too.
Five points
The organization designer Jay Galbraith drew this as a star; his firm says he developed it in the 1960s. Strategy sets direction. Structure decides where decision-making power sits. Processes move information across that structure. Rewards cover salaries, promotions and bonuses. People covers recruiting, selection, rotation, training and development.
His complaint is about where the effort goes: “Most design efforts invest far too much time drawing the organization chart and far too little on processes and rewards.” Structure gets the attention, in his view, because it touches status and power.
Galbraith offers the star as a framework for design choices, not as a tested law, and this summary of the model offers no common scale for how aligned the points are; that still takes judgment. What it gives you is a question to ask of any reorganization: which points moved, and which stayed put? When the “Spotify model” spread, Henrik Kniberg, one of the coaches who described it, said some companies just renamed their teams squads and changed nothing else.
Pods that turned back into tiers
Imagine you run customer support at a software company. Tickets bounce from a first-line team to specialists, and customers tell their story three times. So you redraw the team into pods, one per product, each owning its customers from first message to fix.
Six weeks later the pods may well have rebuilt the old tiers inside themselves. Plenty pushes that way: the bonus still pays per ticket closed, so a hard ticket is bad business; the help-desk tool still routes by tier; every hire for two years was screened for speed on a script.
Move the bonus to the pod’s resolution rate and the routing to the pod, and the new chart at least has a chance.
“Systems designed for humans”
Bringing in AI tempts the narrowest move: buy the tool, add an AI box to the chart. In April 2025, Duolingo’s chief executive, Luis von Ahn, sent an all-hands email, posted on the company’s LinkedIn, saying Duolingo would be “AI-first” and that “making minor tweaks to systems designed for humans won’t get us there.” His “constructive constraints” read like the other points of the star. AI use would count in hiring and in performance reviews. “Headcount will only be given if a team cannot automate more of their work.” Contractors doing work AI could handle would gradually be phased out. He promised more training, mentorship and tooling.
The memo shows intentions, not results. And the star raises a question about one line: a review that counts AI use rewards an input. By the memo’s own account, the point was more teaching content for learners, sooner. A reward points at that only if it also counts what the work delivered.
If your team is adopting an agent, write the five points on a page and, next to each, what actually changed. If only the tool has an entry, the tool is what changed. Then read the review form: does it count use, or results?
The point Congress reached for
In 1982 the chairman of the Joint Chiefs, General David Jones, told a closed House committee session, in Locher’s paraphrase: “The system is broken.” The Goldwater-Nichols Act became law on October 1, 1986. It strengthened the unified commanders again and required the president to submit a national security strategy every year. And it reached into careers: “An officer may not be selected for promotion to the grade of brigadier general or rear admiral (lower half) unless the officer has served in a joint duty assignment,” with waivers.
In 2001 Locher, who had led the Senate staff work behind the act, graded the results. One critic he cited, Mackubin Owens of the Naval War College, called the act’s contributions “marginal at best.” And because it moved several points at once, it is hard to say which one did the work.
Locher gave joint officer management a C+. Qualifications had “improved dramatically,” he wrote, but not thanks to the Department of Defense, which had been “until recently indifferent in its implementation.” The officers had done it themselves, once they came to see joint experience as something that could promote their careers. In the mid-1980s, the same posting had been called the kiss of death.