RTX: AI cuts inspection time by over 30%

RTX's maintenance units have been moving from early-generation digital tools to models trained on a far more robust dataset, according to Nicole White, vice president and general manager for Connected Aviation at Collins Aerospace. That shift reflects years of investment in data collection. Aiir Innovations spent nearly a decade training its borescope product on one of the largest and most diverse real-world datasets in the industry, drawn from actual inspections across multiple engine families and operating conditions. Because engines wear differently depending on climate, region and how they are flown — dust in hot, dry places, salt over oceans — that breadth allows the model to recognize patterns across fleets and perform on platforms it was not originally trained on.

The company's predictive platforms have followed a similar path from skepticism to acceptance. Seth Babcock, general manager for FlightAware and Tech Ops at Collins Aerospace, recalled that some operators initially wanted to test the system by monitoring specified parts rather than acting on its recommendations. When those parts later showed the exact degradation the platform had identified, skepticism turned to trust. White said airlines now expect this level of predictive insight to be part of any major maintenance agreement.

The stakes for RTX rest on whether these tools keep delivering outside pilot programs and on how quickly they can be connected across the company. White described the next step as moving from prediction to prescription — not just knowing a part will wear out, but modeling its entire life cycle. For airlines, the value hinges on whether earlier warnings translate into repairs planned during scheduled downtime rather than during delays, and whether the inventory is available where and when it is needed.