FlightSafety Brings AI To Training


FlightSmart has teamed with IBM to create a “fully integrated training tool … designed to increase training effectiveness and enhance safety through automated, intelligent and objective training.” Translation: Artificial Intelligence is going to analyze your training mistakes and develop a personalized remediation program.

“FlightSmart will set a new higher standard for personalized, highly effective pilot training,” said David Davenport, president & CEO. “This revolutionary new approach employs Artificial Intelligence and Machine Learning to evaluate a pilot’s ability while performing critical tasks and maneuvers during all phases of flight. FlightSmart is then used to create a customized corrective action training path that addresses any identified deficiencies in order to increase proficiency and mastery.”

FlightSmart will give instructors “a comprehensive understanding of a pilot’s strengths and weaknesses through access to real time data, which objectively measures performance against the highest standard parameters,” which in turn allows them to “proactively address any deficiencies by optimizing the training curriculum and focusing on areas that need improvement.” FlightSafety points out that this is a better process than to continue training by repetition. Moreover, FlightSmart can use “advanced analytics and data collection from virtually any training medium.” You can run but you can’t hide, apparently.

FlightSmart won’t be appearing at your local FBO right away. For now, it’s being rolled out for the U.S. Air Force’s T-6A training program at Columbus AFB. It will be used on “16 T-6A training devices, including Initial and Operational Flight Trainers.” From this test run, “FlightSmart will collect and analyze pilot performance data in order to improve instructor utilization and student proficiency advancement. The contract includes options to expand to other AETC bases that operate the aircraft.” FlightSafety says the program will expand into training of maintenance technicians and drone operators.

Marc Cook
KITPLANES Editor in Chief Marc Cook has been in aviation journalism for more than 30 years. He is a 4000-hour instrument-rated, multi-engine pilot with experience in nearly 150 types. He’s completed two kit aircraft, an Aero Designs Pulsar XP and a Glasair Sportsman 2+2, and currently flies a 2002 GlaStar.

Other AVwebflash Articles


  1. Sorry, not all that excited here. There is already far too much automation in the aviation training world. Reliance on CBT over classrooms with a live instructor gains momentum leaving fairly large holes in the gamut of training. I do not at all contest CBT and automation for many routine or repetitive tasks like, say, how to run a new FMC, or adding CPDLC, or some system changes crews need to know about. Items that can be clearly and readily taught via a computer screen. I do feel that the kind of automation addressed in this article is nothing much more than the beginning of taking the instructor out of the cockpit.

    A machine cannot do anything more than evaluate and report based upon a series of statements in lines of code. It cannot evaluate the response of the actual student only the results of the students inputs to the machine. It cannot see that, for example, the student remained focused on an event that is past and was no longer following the current event(s). It will accurately report what parameters were not met or exceeded but it surely won’t have any clue as to why. In the Instruction arena, I have found that “why” can often be a far more relevant question than “what”. A machine cannot evaluate and work on adjusting a mindset. No one can write enough statements to ever cover all of the occurrences or non-occurrences in flight training. This is the same as not being able to ask a CBT device “why”, or “how”, or “what”.