The Rational Actors
Why the aviation industry looks stupid from the outside and makes perfect sense from the inside
Joan Prince Crandall retired on May 30, 2026, after 66 years as a flight attendant. She started in 1959 on a 24-seat Douglas DC-3. CNN ran a full feature. The aviation world celebrated.
Nobody mentioned that the systems scheduling her replacement were designed in the same decade she started flying.
Delta’s crew scheduling infrastructure runs on mainframe architecture whose lineage traces to the 1960s. The Kyndryl contract maintaining it was renewed in 2023. It is, by Delta’s own engineering leadership, “definitely with us for the foreseeable future.”
She got a story about legacy and grace. The mainframe got a five-year services agreement.
I have been inside 130 Operations Control Centers across 80 countries over the past 20 years. I have reviewed more than 600 RFPs. I have watched industry conferences where 18 of 45 sessions were sponsored by vendors selling to the same practitioners sitting in the audience. I have welcome 1,130 members of an aviation operations community to talks by some of the sharpest minds in the field and watched the attendance rate sit at 0.18% of the membership.
The temptation, after accumulating that much evidence, is to reach for a simple explanation. Intellectual insolvency. Wrong personality types. The industry doesn’t understand its own problems.
That explanation is wrong. And it leads to the wrong interventions.
Here is what I actually observed across those 130 visits, those 600 RFPs, and four years of running a professional community that refuses to engage with its own future.
The people inside these systems are not stupid. They are rational actors in a structure that rewards the wrong behaviors. The distinction matters because if the problem is people, the solution is different people. If the problem is structure, the solution is different incentives. The industry has been trying the first solution for 40 years. It has not worked.
The Gantt chart was invented around 1893. It is still the dominant planning artifact in airline crew management. Not because no one ever thought of something better. Because no vendor in 130 years has produced a planning interface that is as legible under pressure, on a degraded system, at 03:00 in the morning, by someone who has been awake for ten hours and is managing a disruption that will cascade if they misread a single row. The Gantt chart stayed because it is the only tool in the room that everyone from the CEO to the overnight shift supervisor can read without training. The moment something better arrives that meets that standard, the Gantt chart disappears in a single contract cycle. It has not disappeared because the replacement has not arrived in a usable form.
The 500-requirement RFP is not evidence of organizational stupidity. It is a scar map. Every requirement in that document represents a previous failed implementation, a vendor promise that dissolved into middleware, a “seamless integration” that took 18 months and a $40 million integration tax. The airline that writes “must integrate with legacy 1990s crew system” writes that because the last vendor who promised a clean migration delivered complexity instead of clarity. The procurement theater that observers mock from the outside is institutional PTSD wearing a Word document.
The dispatchers on JetCareers who pushed back on automation in 2025 were not defending ignorance. They were defending expertise. My eye-tracking research across those 130 OCCs showed controllers spending 60 to 70 percent of their cognitive capacity navigating systems rather than making decisions. Those controllers have spent 20 years developing navigation expertise. Eliminating the navigation burden does not elevate them. It makes their expertise irrelevant and their seniority meaningless. Nobody makes that trade rationally when their mortgage depends on it.
Carlo Cipolla, the Italian economic historian, defined stupidity precisely: actions that cause losses to others while producing no gain and often incurring losses for the actor. That is not what airline operations employees are doing. They are doing something different and more interesting. They are acting in ways that are locally rational and systemically destructive.
The procurement officer who writes a 500-requirement RFP protects herself from the career risk of a failed implementation. Locally rational. The requirement that the new system integrate with the 1990s crew management platform ensures that no vendor can promise a clean break, which means no vendor can deliver one, which means the 1990s platform survives another cycle. Systemically destructive.
The NACU conference attendee who accepts the sponsored breakfast and declines the collaborative workshop protects her company from the liability of surfacing sensitive operational data in a room that includes competitors and vendors. Locally rational. The collaborative workshop that never happens means the industry never builds the shared vocabulary for what good looks like. Systemic knowledge deficit compounds. Systemically destructive.
The senior controller who does not engage publicly with ideas that challenge the current architecture protects herself from the organizational cost of being seen as the person who said the system was broken. Locally rational. The 1,130-member community that never debates its own future means the vendors designing the next generation of systems are building to the wrong requirements. Systemically destructive.
Every actor in this system is making reasonable choices. The aggregate of those choices is what looks, from the outside, like an industry that cannot think.
There is a specific moment in every RFP process where this becomes visible. It is the requirements document review.
A procurement team at a mid-sized carrier will spend six months building a requirements document. They will consult the operations team, the IT team, the finance team, and the legal team. The document will grow to 400 or 500 requirements. It will include requirements that contradict each other. It will include requirements copied from the previous vendor’s proposal, sometimes with the competitor’s product name still in the text on page 47. It will include requirements for features the airline used once in 2017 and has not needed since.
What the document will not include: any requirement that the new system reduces cognitive load. Any requirement that the interface is legible at 03:00 after a ten-hour shift. Any requirement that the system supports pattern recognition transfer from retiring controllers to new ones. Any requirement related to the Knowledge Cliff the industry is walking toward as the Joan Prince Crandall generation exits the operation.
Those requirements are not in the document because they are not measurable in the procurement format. The procurement format rewards specificity and punishes ambiguity. “Must support IATA pairing rules” is a requirement. “Must not exhaust the person using it” is not a requirement. So the exhaustion gets excluded from the evaluation, and the industry buys systems that are technically compliant and operationally punishing.
The industry conferences that should correct this have their own structural problem.
I watched a panel in Doha where the most important question in the room, does any of this actually work when you need it at 02:00 with three crew members unavailable and a slot restriction at the destination, was asked once and not answered. The session was rated moderately well. The majority wanted more audience interaction. The conference survey came back satisfied.
The sponsors paid for the conference to happen. The conference needed sponsor revenue to fund itself. More sponsors meant a better hotel and a richer social program. A better social program attracted more attendees. More attendees made the sponsorship more valuable. The technical program quality was largely irrelevant to this loop. What mattered was that the program looked credible enough to justify attendance, and attendance justified the fees.
The result is a conference ecosystem where vendors present solutions that have not been tested in production, academics publish papers that will not be deployed, and airlines attend for the food and the networking and return to operations unchanged. Everyone’s incentives are satisfied except the one that should matter: whether the industry’s actual decision-making infrastructure improves.
The NACU pattern is the same dynamic in a smaller room. Airlines prefer sponsored meals to collaborative workshops because the sponsored meal has no political cost and no competitive exposure. The collaborative workshop requires sharing operational data with competitors and vendors. The airline that participates and nothing comes of it has spent political capital for nothing. The airline that attends the breakfast has eaten well and committed to nothing. In an industry where operational data is treated as competitive intelligence, that calculation is made the same way every time.
The Knowledge Cliff is the version of this problem that cannot be deferred.
Joan Prince Crandall’s retirement is a data point. There are thousands of them. Every senior controller who exits an OCC this year takes with them pattern recognition that no system was designed to capture and no training program was designed to transfer. The MIT research on tacit knowledge is precise about the mechanism: experts develop unconscious gaze patterns that direct attention to the informative parts of a complex visual field. In a crew management Gantt chart during a disruption, the experienced controller’s brain has learned which rows carry the binding constraints before the system flags them. The novice controller scans everything. The system logs outcomes. The judgment evaporates.
The industry treats this as a staffing problem. It is an architectural problem. The systems running today were built to execute against pre-defined rules, not to learn from human judgment in real time. The judgment that walked out the door with every retiring controller was never treated as data worth preserving. The architecture was not designed to receive it. The vendors who built the systems were not asked to capture it. The procurement documents did not include it as a requirement.
The Kyndryl contract was renewed for five years. The Knowledge Cliff is accelerating. Those two facts are not in the same conversation anywhere in the industry.
I am not arguing that the people in these systems are blameless. Rational behavior in a broken structure still produces broken outcomes, and at some point the people inside the structure bear some responsibility for the structure.
What I am arguing is that calling it stupidity produces the wrong diagnosis and the wrong intervention.
The Cipolla framework is useful here precisely because it distinguishes between the bandit, who causes harm to others for personal gain, the helpless, who causes harm to themselves without benefiting others, and the stupid, who causes harm to everyone including themselves for no benefit at all.
Most of what I have described in this piece is bandit behavior operating at an institutional scale. The conference ecosystem extracts value from the attendees. The procurement format protects careers at the cost of better outcomes. The sponsored breakfast substitutes for the collaborative workshop that might actually change something. These are rational extractions, not irrational failures.
The intervention for bandit behavior is different from the intervention for stupidity. You do not fix it by finding smarter people. You fix it by changing what the smart people are incentivized to do.
The airline that breaks this pattern first will not do it by hiring different people. It will do it by measuring different things. By including legibility requirements in procurement documents. By running conference sessions where the question “does this work at 02:00 in a real disruption” is not just asked but required to be answered with evidence. By treating the judgment of retiring controllers as data worth preserving before they leave, not as sentiment worth celebrating after.
The systems are not stable because they work well. They are stable because the incentive structure rewards stability and punishes the transition cost of replacement. That is a solvable problem. It requires someone to absorb the transition cost and measure the outcome on the other side.
A few airlines are starting to do this. They are not announcing it in conference presentations. They are doing it quietly, in contract cycles that will not surface publicly for two or three years.
By the time the rest of the industry finishes the current RFP cycle, the gap will already be structural.
Joan Prince Crandall flew her last flight. The mainframe is on contract through 2028. The next generation of controllers is already at the desk.
The architecture that receives their judgment has to be built before the senior controller beside them hands in their badge.
Daniel Stecher is Vice President Business Development at IBS Software. Over 20 years in aviation operations. 130 Operations Control Centers visited across 80+ countries. Founder of Airline Crewing and Operations Enigma, a community of 1,130+ members across 261 airlines. Thinkers360 Global Top Influencer. All views his own.