The Instructor Is the Constraint: The Latest Heads of Training Report

Six topics were tabled at this year’s Heads of Training Meeting in Bangkok, across ten moderated roundtables and two rounds of discussion. They covered CBTA and EBT maturity, instructor standardisation, emerging technology, the instructor workforce, training capacity against regional growth, and the relationship between flight schools and airlines. Participants came from full-service and low-cost carriers, ab initio schools, independent and manufacturer-owned training organisations, technology developers, a pilot union’s training arm and, at several tables, civil aviation authorities. All of it ran under Chatham House Rules.

Read together, the ten reports point to one finding that no single table set out to make. Participants were asked to name the greatest risk in their training pipeline. At different tables, on different topics, in different markets, they gave the same answer. Not simulators. Not candidates. Not money. Instructors.

We have been standardising the wrong half

The sharpest point of the meeting came at one of the instructor standardisation tables. 

When organisations say instructor standardisation, they almost always mean grading standardisation. But grading is assessment, and assessment is only part of an instructor’s job. The instructor’s job is to instruct.

One participant explained why the balance has tipped that way. Very little published material teaches competency-based training. A great deal has been written about competency-based assessment. So that is where organisations start, and it is where airlines most often get stuck when they introduce EBT. The other half of the job is instructional skill: looking at one pilot, working out what is holding them back, and helping them past it. Several heads of training admitted they had rarely discussed that in their own organisations.

The effects showed up everywhere. One participant described a concordance exercise across eighty-five instructors at a large training organisation. For a single event, the grades ranged from two to five. He put the cause not on the instructors but on how they were trained. He compared a demanding instructor course from earlier in his career with what he is required to do now: an online module you can sleep through and still tick the box.

Systems get the behaviour they reward

One problem came up more often than any other: the middle grade. Instructors mark three across the board, explain nothing and write nothing down. It shows up later, when a pilot fails a check. The file is reviewed, there is no record of how it happened, and nobody can say whether the pilot or the programme was at fault.

What makes this useful is that several organisations, working separately, reached the same fix. They changed the system instead of asking instructors to try harder. Observable behaviour markers are now required at every grade, including three, with one positive and one negative. The reasoning is simple. A three means some expected behaviours were missing while the flight stayed safe, so there is always something to record. And the shortcut now takes more effort than a proper grade.

The same principle explained a second finding. At one airline, a low grade on an EBT day one bars a pilot from becoming an instructor later. So pilots learn the EBT by heart and manage the number. As that table put it, if there is always a consequence, it is not training. It is checking. Several participants noted that their operations manuals still carry rules written before EBT existed. One carrier now runs a third training day with no record at all — no hours, no grade, nothing in the file — built entirely around what the pilot wants to work on. Another offered a simple test: a failure rate of two thousandths of one percent does not mean all the pilots are excellent. It means the programme is not working.

Everybody has data. Few have it connected.

Asked how many connect safety data to training data, only a minority of operators said yes. The reason is structural. Safety is one department and training is another. But several working models were described. One safety department now sends every event with the competencies already identified, rather than as a raw report. One operator holds a weekly meeting bringing training, safety and flight operations around flight data. Another sends its data to a manufacturer benchmarking programme that reports back every six months.

At the technology table the warning was blunt. One organisation brought two years of training records and asked where its programme could improve. Every record was a three or a four, with no observable behaviours and a comment saying “good job.” The answer was not in the data, so no analysis could find it. Which brings the argument back to grading, and from grading back to the instructor.

That table also gave the clearest case for AI in training. Research on working memory suggests a person can hold about four things at once. CBTA and EBT ask an instructor to track eight or twelve, while running the exercise, playing air traffic control, coaching the crew and writing an assessment at the end. At four in the morning, the seventy-odd observable behaviours do not get recorded. Seen that way, AI in the simulator is not a way to save money. It is what makes the method workable for a human being. The return is not fewer staff. It is training data that exists at all.

What would actually help

The encouraging part of these discussions was how many of the answers cost nothing. One regulator extended the period in which a retired pilot can still qualify as an instructor, from three years to seven. The reasoning is simple: a captain retiring at sixty-five wants two or three years before coming back, and a three-year window loses them for good. The change recovers experienced instructors immediately, and any market could copy it.

Other answers were just as ordinary. One provider limits its simulator instructors to no double sessions and no more than three nights in a row, in a country with no fatigue rules for simulator work, because quality drops sharply without it. It also pays them to attend standardisation days, and says people are queuing to join. A large training organisation found more than twenty different ways of assessing instruction in use across the group. It adopted the published instructor competencies instead, partly because an international standard is harder to argue with, and now runs eight thousand assessment reports through one dashboard. The dashboard showed that facilitation was the behaviour instructors most often failed to demonstrate, so refresher training was aimed at facilitation. At the technology tables, the tool rated most effective by the people using it was the video library — named more often than VR as the thing that improved training most.

The uncomfortable part

One line from the CBTA maturity table deserves the last word. Participants were asked to place their organisation on a scale of one to five. 

The real danger, the moderator said, is not being a two. It is being a two and believing you are a four. An organisation convinced of its maturity, stops listening, and typically finds out the truth only when an incident forces an investigation.

The industry these ten tables described is not short of technology, data, or frameworks. It is short of instructors, short of the time to develop them properly, and short of an honest account of how well its programmes work. Outcomes, as one participant said, are easy to measure. Efficacy is the hard one. And efficacy is what maturity really means.

Upcoming Heads of Training Meetings:

EATS Heads of Training Meetings (Cabin Crew and Pilot)

Location: Cascais, Portugal

Date: 3 November 2026