CRM scenarios pilots don't roll their eyes at.

Every pilot knows the canon. Tenerife, 1977: two 747s and a communication breakdown. United 173: a flight deck so focused on a landing-gear light that nobody flew the fuel state. The accidents that created CRM are real, the lessons are real, and after the fifteenth retelling in a hotel conference room, the eyes glaze anyway. Not because crews don't care. Because hearing about someone else's worst day is not practice for your own.

The industry knows this, which is why the regulatory ground has moved under CRM training over the last decade, and why "we did the annual CRM day" is quietly becoming an obsolete answer.

The bar moved: competencies, not attendance

ICAO's competency-based training and assessment framework folded CRM into nine assessed pilot competencies, and evidence-based training (ICAO Doc 9995) rebuilt recurrent training around data from real operations instead of repetitive set-piece maneuvers. The definitions matter, because they describe something a slideshow structurally cannot deliver:

Observed through behaviors. You cannot observe a behavior in a classroom lecture, and the same IATA paper says the quiet part plainly: CBTA "uses more scenario-based training for more realism." The framework is, in effect, a mandate for rehearsal. What remains open is where crews get enough rehearsal to matter, because full-flight simulator hours are scarce and expensive, and the competencies that decay fastest are the non-technical ones.

Why this is worth solving

The case for drilling the human layer hasn't weakened since 1977. Across decades of Boeing's commercial jet accident summaries, more than 70% of hull-loss accidents carry human error as a primary factor. The machine got safer faster than the conversation in front of it. And the general training research points the same direction: Sitzmann's 65-study meta-analysis found simulation-game trainees 20% higher on post-training self-efficacy and 14% higher on procedural knowledge than comparison groups. Confidence to act and knowing-how under pressure: that's CRM's entire job description.

What a scenario needs before crews respect it

We built a flight-deck concept called The Gradient around the single hardest CRM behavior to train: the first officer's challenge that actually gets heard. A few design rules from that build generalize to any CRM scenario worth running:

• Score the speak-up, not the recall. If the scenario rewards naming the threat-and-error-management category, it's a quiz in costume. The Gradient scores whether the challenge changed the captain's behavior, including how hard the player pushed when the first attempt got waved off.
• Make the social pressure real. The authority gradient is the scenario. A scripted captain who accepts the first polite query teaches nothing; the dozen variations where the pushback costs something are where the instinct forms.
• Map every scenario to the competency framework. Each run should produce observable-behavior evidence against the nine competencies, so the output feeds your EBT program instead of sitting next to it.
• Let crews replay it. The expensive sim slot happens twice a year. A scenario that runs on an iPad in the crew room can run twenty times, and rep count is what turns a lesson into a reflex.

The eye-roll test

A simple way to evaluate any CRM scenario, bought or built: does it ask the crew to remember an accident, or to make a decision? The first produces nodding. The second produces the small, useful failure, the challenge made too softly, the monitoring lapse nobody noticed, that a debrief can actually work with. Crews respect training that lets them fail safely at something hard, and they roll their eyes at training that asks them to admire how someone else failed.

There's more on how we build aviation scenarios, from CRM and automation management to ramp and cabin decisions, on the aviation page, and on the simulation approach generally under safety training.