Sent to the newsletter list on . The Travel Tech Podcast newsletter goes to 30,000+ people in travel and aviation technology each week.
Listen to the episode · Ep 41AI Knows Your Bag Won't Make It: Predicting Connections Before Takeoffwith Ismael Pacheco, Co-founder, Out of LimitsFrom the Travel Tech Podcast newsletter for episode 41, AI Knows Your Bag Won't Make It: Predicting Connections Before Takeoff.
The best test of an airport system, my latest guest argues, is whether the phone rings within a minute of it going down. Founders rarely set that bar for themselves, and for the 30,000+ of you across aviation, travel tech and AI, it's a sharper filter than any demo. I wanted that perspective on the show.
He started with a university simulation project inside Lisbon Airport's baggage hall in 2002 and co-founded a consultancy four years later. Then he turned it into a product company that builds baggage and turnaround software on site with airports and airlines rather than in a lab. Its turnaround system now runs at Lisbon, Porto and Faro, with Budapest and Lyon in deployment. Listen on Apple Podcasts, also streaming on Spotify and YouTube.
The thread running through our conversation: turning baggage and turnaround data into knowledge early enough for people to act on it, with a human still taking the last decision.
From black box to prediction
Data is where it starts, and for a long time it was hard to reach. A few large suppliers build and run baggage handling systems, and for years the operational data sat inside their proprietary systems, where finding anything meant digging through raw files, which Ismael called a nightmare. At some airports the problem remains. Ismael's team finished a South American baggage project this year in which the handling supplier either resisted sharing data or quoted the airport a steep price for access to data from its own operation.
Prediction runs on that data. Out of Limits says its system works out at check in, before an eight hour inbound flight, whether a transfer bag will make its connection, and keeps updating as taxi times and baggage hall lead times change. On the stand, Ismael says, every event in the turnaround has to be timestamped before AI can help predict whether an aircraft will really leave at its target off block time. The payoff compounds, he argues: two or three minutes saved per turnaround, at an airport with more than 800 turnarounds a day, becomes a huge improvement within a few months. Autonomy is where the line still holds. Out of Limits' new agent, Captain, monitors flights and raises warnings, but humans still make every decision.
I spoke with Ismael Pacheco, co-founder of Out of Limits, about the baggage data black box, setting alarm severity for crowded control rooms, computer vision on the stand with NeuroStand, why no two airports deploy the same way, and why being a painkiller beats being a vitamin.
Key Takeaways
- Severity must track consequence: A low scanner read rate on a sorter induction line needs action now, while the same drop after check in can wait.
- Small outages cascade: Five minutes of baggage system downtime can take 60 to 90 minutes to recover and push delays onto departures.
- Ready is a prediction problem: A ready message can go out while the stairs are still on and the door is open, and the aircraft then holds a slot another flight could use, which is why predicting the real off block time matters.
- Every deployment is local: Computer vision models trained in Portugal still need retraining for Budapest's and Lyon's stand layouts, equipment and lighting.
- Operator expertise is the fallback: Systems may now hold more knowledge than experienced controllers, but those controllers are what keeps the baggage hall running when the software fails.
🔧 Three to Build
This week's episode makes the case that baggage data only earns its keep when it reaches a person early enough to act on it.
Build a transfer bag predictor that scores every connection at check in, hours before the inbound flight lands.
Ismael's example: a passenger checks in in South America for an eight hour flight to Lisbon, and the system is already working out whether the bag will make the onward flight. The score keeps updating as taxi times and baggage hall lead times change. For a very short connection it recommends moving the bag tail to tail, straight from one aircraft to the next, because a trip through the main terminal would miss the flight.
Catch: the prediction is only as good as the baggage system data behind it. Ismael finished a project this year where the handling supplier quoted the airport a very high price for access to data from its own operation.
→ Baggage routing and prioritisation for connecting flights
https://airsidelabs.com/use-cases/intelligent-baggage-routing-and-prioritization-using-ai-powered-decision-support?utm\_source=linkedin&utm\_medium=social&utm\_campaign=ep41
Build control room warnings ranked by what the fault will cause.
A baggage control room already has SCADA screens, CCTV and dozens of signals competing for attention. A useful warning weighs the consequence. A falling read rate on a sorter induction line needs someone there now, while the same drop on a scanner after check in can wait. A full chute is critical when bags are on their way to it and a minor note when none are.
Catch: those severity rules live in the heads of experienced operators. Ismael's team learned them by spending whole days in the baggage hall with the people who run it.
→ Anomaly detection for baggage handling issues
https://airsidelabs.com/use-cases/utilize-iot-sensor-data-and-anomaly-detection-to-proactively-identify-and-resolv?utm\_source=linkedin&utm\_medium=social&utm\_campaign=ep41
Build a shift brief agent that watches while the operator decides.
Out of Limits launched its agent, Captain, at World Aviation Festival. An operator starting a shift asks where to look, gets a summary, then tells the agent which flights to monitor. The agent raises warnings and takes no action, and the human makes every decision. The same pattern would suit a baggage control room.
Catch: Ismael says operators have mixed feelings. Some fear being replaced, while others are starting to see the agent as the best support they can have. Those operators are also the fallback when the software goes down, so their trust decides whether the tool gets used.
→ Real time recommendations for baggage handling operations
https://airsidelabs.com/use-cases/leverage-multimodal-data-fusion-and-decision-support-systems-to-provide-real-tim?utm\_source=linkedin&utm\_medium=social&utm\_campaign=ep41
Safe travels!
Alex Brooker, Host of the Travel Tech Podcast, Founder of Airside Labs

Alex Brooker
Founder & CEO, Airside Labs
Alex Brooker brings over 25 years of experience in aviation, defense, and safety-critical systems. Former VP of R&D at Cirium (RELX PLC) and Principal Engineer at BAE Systems. Award winner (IHC Janes Innovation 2016, Technology 2015). Alex founded Airside Labs to ensure AI systems truly understand their domains through rigorous evaluation and testing.
Connect on LinkedInEvery episode, in your inbox
One email per Travel Tech Podcast episode: the conversation, the key takeaways and the use cases worth building in aviation and travel AI.
Give your agent aviation tools
General assistants guess at aviation codes and get them wrong. Agent Tools give Claude, ChatGPT or your own agent checked answers instead. Free plan, no card.
Connect your agent