Cancel the Flights So Nobody Misses Connections

From the Travel Tech Podcast newsletter for episode 37, Cancel the Flights So Nobody Misses Connections.

When Kevin and I first tested an AI on a real ops problem (three inbound flights landing together, protect every connection), it answered with flawless logic: cancel the flights. It satisfied the goal exactly as written and destroyed the goal as meant, which is the failure every team deploying agents in operations will eventually meet. Kevin has spent more than two decades building systems where that answer would be dangerous, and for the 30,000+ of you across aviation, travel tech, and AI, his view comes from operations, not demos. I wanted that perspective on the show.

He started on Lockheed Martin’s oceanic air traffic control program, where a new system replaced paper strips and HF radio relays and let controllers handle 30 aircraft instead of a busy 7 to 15. From there he rose to CTO roles in the US and UK, built the platform that connects retail partners to Google’s drone delivery network, and now runs the innovation lab at the aviation data company where he previously led product. Listen on Apple Podcasts, also streaming on Spotify and YouTube.

The fault line running through our conversation: AI has made building and analysis nearly free, but aviation’s data is still fragmented and its decisions too interdependent to automate end to end.

Behind the dip, where nobody has time to look

That tension has a structural cause. Aviation data is a graph that starts with a tail number: who ordered the aircraft, who leases it now, the schedule it was planned against, the hundred changes that schedule goes through before departure, and the tail swap on the day. Most of it lives in separate systems, and the standard fix is a multi year, eight figure data lake program that ends with everything in one place and still nothing to act on.

AI changes the economics on top of that foundation, not underneath it. Kevin estimates a typical aviation decision runs about eight steps, and AI can now carry four or five of them: analyzing every turn instead of one, spotting that inbound delays held under 10 minutes resolve at a given station, or flagging a part failing faster than planned before the next maintenance visit. The last step is where it breaks, which is why he expects flow and efficiency tools to accelerate while separation stays firmly human in the loop.

Kevin Hightower, Chief Innovation Officer at Cirium, joined me to talk about what oceanic ATC taught him about automating the routine, why prototypes should now be cheap enough to throw away, the aviation knowledge graph Cirium is launching, why AI may make the passenger experience worse before it makes it better, and the data interoperability problem he thinks nobody has solved.

Key Takeaways

• Automate the routine, not the judgment: Oceanic controllers went from a busy 7 to 15 aircraft to a comfortable 30 once the system handled conformance checks and altitude probes, freeing them for problems only a human could solve.
• Kill prototypes faster: When a build costs days instead of months, the discipline is discarding work rather than force feeding an idea customers have already rejected.
• Investigate the dip before anyone sees it: The real value is AI that analyzes every turn behind a KPI drop, benchmarks it against competitors, and surfaces a decision.
• Efficiency is the safety lever: Better sequencing and time based flow management upstream reduce the load on safety critical systems that will stay human controlled for years.
• Interoperability beats consolidation: The next unlock is combining internal and third party data with access controls intact, without a five year data lake program.

Alex Brooker, Host of the Travel Tech Podcast, Founder of Airside Labs