Presenting 'Project Constellation' at the Predictive Aircraft Maintenance (PAM) APAC 2026 in Singapore today, Audrey Majors, director consulting, Americas strategy and transformation team at SITA, and Alex Bychkovsky, director of operational analysis at Air Transport Services Group (ATSG), outlined the results of a co-innovation project built around live operational data from two ATSG carriers.
The project tackles a familiar industry problem: maintenance, inventory and procurement data often sit in disconnected systems, making it difficult for airlines to see the condition, location and reliability history of parts across their operations. "The result can be stock sitting unused in one location while another team pays a premium to source the same requirement elsewhere," explained Majors.
Rather than beginning with a broad industry-wide concept, ATSG and SITA deliberately narrowed the initial scope to parts availability and inventory accuracy across the Boeing 767 fleets of two ATSG carriers. "The analysis covered 83 aircraft, more than 56,000 installed components and over 207,000 inventory units across 77 stations, using 12 months of production data," delegates were told.
What emerged was a sizeable opportunity hidden within existing data.
The analysis found that 20–30% of the excess and dormant rotable pool sat above historical demand, while nearly one in five installed parts returned within 90 days. Across the two carriers, more than 1,200 hours of aircraft time were also lost to parts-related disruption during the year, spread across 328 events.
Crucially, the two carriers did not display the same operational behaviour despite belonging to the same company and operating the same fleet type. Their relative performance differed across measures including early part returns, hours lost per disruption and inventory depth - differences that were not apparent when each carrier's data was viewed in isolation.
'Project Constellation' uses AI agents to bring that information together and recommend options, while retaining human control over the final decision. As the presentation stressed, “the agents recommend. The human decides.”
To test the potential impact, ATSG and SITA explained to delegates that they reconstructed a full year of parts movements and replayed the same demand twice - first as it actually occurred and then with agent-supported decisions. Depending on the level of adoption modelled, the exercise indicated reductions in serviceable excess inventory ranging from 4.5% to 9.4%, alongside increases in service-critical safety stock.
The longer-term ambition goes beyond optimising inventory within individual airlines. The audience was told that 'Project Constellation' envisages a federated model in which carriers can collaborate across inventory boundaries, replacing some of the telephone-based lending and borrowing that already takes place with decisions supported by joined-up data.
Majors revealed that a decision review is planned for October or November, with the next steps potentially including defining the production build and bringing another two or three airlines into the programme. ATSG is envisaged as the first client and research partner if the platform progresses to market.
