Airside Labs
    Cancel the Flights So Nobody Misses Connections
    Newsletter

    Every 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.

    Episode 37
    October 15, 2026

    Cancel the Flights So Nobody Misses Connections

    Kevin Hightower, Chief Innovation Officer, Cirium

    What if the smartest way to ensure no passenger misses a connection is to cancel every flight? That was the logically flawless and completely useless answer an AI produced when given the wrong objective. Kevin Hightower, Chief Innovation Officer at Cirium and leader of its Propulsion Labs innovation group, unpacks what that moment reveals about AI in aviation, and where the technology will genuinely reshape how the industry operates.

    Show Notes

    Kevin Hightower is Chief Innovation Officer at Cirium, where he leads Propulsion Labs, the company's dedicated innovation group, after previously serving as its head of product. His career spans systems engineering on oceanic air traffic control at Lockheed Martin, where a new FAA system let controllers comfortably handle 30 aircraft instead of a stressful 7 to 15 by automating conformance checks and altitude probes, through CTO roles in both the US and UK, to four and a half years at Google's drone delivery program, where he built the platform that partners such as Walmart and DoorDash use to connect to a service that delivered about a quarter million packages across Australia, the US, and Finland. In conversation with host Alex Brooker, Hightower explores how AI is reshaping product development when build cycles collapse from months to days, shifting the key discipline from shipping to knowing when to throw work away. He introduces his three hour decision rule, explains how most aviation decisions run roughly eight steps with AI now capable of handling four or five before a human intervenes, and describes Cirium's aviation knowledge graph, which links aircraft orders, lessors, schedules, and tail swaps into analysis no single dataset can produce alone. He also addresses why multi million dollar data lake programs so rarely deliver usable value, and why AI will accelerate flow management and sequencing long before it touches safety critical separation systems.

    Read the takeaways

    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.

    Read the newsletter issue

    Full TranscriptClick to expand (10,872 words approx.)

    Kevin Hightower: It might suggest a decision based on whatever, but you still have to — that human in the loop has to be there to really understand what the options are. I think you actually ran one of our first tests with this. We're like, hey, I've got three flights, we're all in at the same time, what's the best way to not have any passengers miss their connections? And it said, cancel the flights. Nobody will miss their connections.

    Alex Brooker: Right, yeah, it's a logical solution.

    Kevin Hightower: Correct.

    Alex Brooker: Kevin Hightower, welcome to the Travel Tech Podcast.

    Kevin Hightower: Hi, great to see you again, Alex.

    Alex Brooker: It's good to see you too, and I'm interviewing you this time rather than any other kind of meeting setup.

    Kevin Hightower: Awesome, I love it.

    Alex Brooker: Full disclosure to the audience, Kevin and I worked together for a few years and had a great time at Cirium. Just a couple of — seems like yesterday.

    Kevin Hightower: Yeah, and worked together even before that, right? When you were Snowflake and I was Lockheed Martin. Worked together with NATS.

    Alex Brooker: Absolutely. Tried to get them moving in the right direction.

    Kevin Hightower: Long time. Yeah.

    Alex Brooker: Yeah, fantastic. Maybe let's start — perhaps even before then, with some of the other guests, we sort of started with: how did you get into aviation? You've spent much of your career in and around and deep within aviation, so I'd love to hear you tell your story.

    Kevin Hightower: Sure, sure thing. So we'll go way back. I turn 50 next year in August. So when I was in school, way back when — before we had... I think my first job, I had two 20-inch CRTs on my desk. That's how far back I go, and that was when I was actually working. But even before then, in the 90s, computer engineering and software engineering was big. Everybody hiring was fantastic. You had your choice of jobs.

    When I finished high school, I wanted to do software engineering, and there were lots of choices. And Embry-Riddle reached out, and I love airplanes, so I was like, let's go see what this is about. My uncle was actually a private pilot — occasional private pilot — and I lived in South Florida. Embry-Riddle is in Daytona Beach, about 200 miles north, and he flew me there in a rented Cessna.

    Alex Brooker: Wow, that's cool.

    Kevin Hightower: We landed at Daytona International Airport, and Embry-Riddle is connected to the Daytona International Airport — their flight line's right there. Did the tour, did lots of other places, but Embry-Riddle is where it's at. And it was a way to specialise, right? Everybody's doing software engineering. How could you be unique about it?

    And so I went to Embry-Riddle, and during the introduction to the school — when I started, the president said there are two types of people in this world. There's people that see a plane flying by low and go, "that's a plane," and then there's people that pull over and walk up to the fence and watch it land. And all of those people are in this room right now. And so I felt right at home, as I'm sure many of your listeners do as well.

    And so, yeah, I did Embry-Riddle. I did my undergraduate in aviation computer science and my master's in software engineering. When I started my master's, they started a five-year programme for what had been a six-year programme, so unfortunately I was too late for that and did six years overall. But it was great. I had a wonderful job running an IT department and software development for a small company that had 50 locations across the US. I dialled up ISDN when I started to get connected to these sites — that's how far back we're going.

    We had red phones — we called them the bat phones — that only worked over the ISDN to call the other facilities. So you had your phone to call customers and your phone to call the facilities. It was really cool for me to actually see this stuff in action. Anyway, I could diverge on that for a long time. Software development, had a five-person IT team, wrote a bunch of software — Visual Basic 6, I think.

    And then after I finished my master's, there were a lot of choices. A lot of people were hiring software engineers, and it sounds strange these days, right? They would fly you there for three days, take you out, put their offer to you. I interviewed with Lockheed in Texas, and all of the interviewees sat at a table and the managers came to you. They would come interview you and then leave, and another manager would come to you. It's just that whole flip of — you're the valued asset, not us. This is 2001, right?

    And so I had two job offers: one in Texas doing R&D — at the time before they had won the Joint Strike Fighter, but Lockheed was in the competition — and another one in DC doing air traffic control. I was really having a hard time deciding.

    Alex Brooker: That's a tough choice, Kevin.

    Kevin Hightower: There was something like 30,000 people there and they had a 15-person R&D department, and I would be in there straight out of school, which was amazing.

    I chose the ATC route, but my first manager — while competing with each other for me — Lockheed Martin said, "Do you want to kill people or do you want to save people?" I was like, well, I'm going to save people.

    Alex Brooker: That's how you ended up in air traffic control.

    Kevin Hightower: Air traffic control, yes. And so I joined Lockheed Martin just outside of Washington, DC as a systems engineer. Did requirements and design and got to do formal tests with the FAA — formal requirements specs that take forever. We launched a new ATC system in the ocean in about four years, from deciding to do it to actually making it happen, which is remarkable for any aviation facility. And not only that, it was really changing what controllers did in their job, and it had a really big impact on me.

    So if you're familiar with oceanic controllers — there's a saying: you could work a flight, leave, take your break, come back in eight hours for your next shift, and still be working the same flight. It's the Pacific Ocean. Some of the flights take 14 hours to cross.

    They had these big workstations with paper strips — 14 feet long, I think — one little screen, and pencil and paper. They'd call on a scratchy HF radio, change the times, "oh, you want higher? Let me move my strips. Okay, you can have higher." That would be relayed through HF, bouncing off the atmosphere to get to the plane.

    Because of that, separation was huge — a hundred miles in some places between airplanes. The new system was — and I still think it's the only system in the world where the system does the separation. If the computer doesn't find a conflict, the controller is not at risk.

    It automated all the routine. If they called within a minute or two of when they were supposed to, — and we're working in whole minutes out in the ocean — it would automatically update it, reprobe the flight, make sure everything was safe. They'd call for higher, push a button, and it would say green or red. They had situational awareness, but it was the system that would decide whether it was green or red. And because of that, automating the routine, they could do all this extra stuff.

    I got to sit there when they used it — watching people transition from pencil and paper to a computer and data entry, that kind of thing. The first week I was there, there were three flights coming from the West Coast down to the Australia and New Zealand area, all kind of next to each other. Of course, the plane in the middle wants higher than the two above him. And the controller said, "Last week, that's it for now. This week, everything else is being processed — I have a little time to look at this." And he figured out a solution: in 30 minutes, give them five degrees to the right, and then in 30 minutes they'll be able to climb. Set it up via data link straight to the cockpit.

    And because of that, controllers went from being busy with seven to 15 airplanes to not being busy with 30 airplanes.

    Alex Brooker: Oh, wow.

    Kevin Hightower: And not being busy let them do these other things. And I think that's where my drive comes from around AI — you automate the routine so you can move to the next thing.

    So I did that for a long time. Eventually became CTO in the US, then moved to the UK for two years as CTO there — that covered everything outside the US. And then it was time to go home from the UK, and the only way back was DC. We had the hard question of taking another year in the UK, going to DC with Lockheed, or looking for something else. And I joined a 50-person startup in Santa Monica, California, instead of a 100,000-person well-established company, as you do.

    Did that for a year. I tried to recruit a friend from Google, he tried to recruit me, and he won. So I joined Google's drone delivery for about four and a half years, and we delivered about a quarter million packages to real people using real drones across Australia, the US, and Finland in the time that I was there. It was an amazing time.

    I learned a lot coming from safety-critical, then moving into airports and airlines as part of Lockheed — which was not safety-critical but still had its structure — to "let's move fast," the Google kind of drive around product delivery and things like that. And then I did that for about four and a half years. My boss decided he wanted to go back to Google Maps or something, and it was just a catalyst for me to take stock of where I was.

    I was happy at Google, though while we were waiting on regulation early on, we had done a lot and were still doing a lot. I built a platform for — if you look in the news — Walmart and DoorDash using Google drones for deliveries. I built the whole platform for how they connect. That was about five years ago.

    And so I looked around. Ian Painter — I think he's been on this podcast as well — we're both good friends. I reached out and he said, "Oh, there's a job that's about to go up." And so I joined as head of product at Cirium about four and a half years ago. Did product for about two years, got our products headed in the right direction. Had Propulsion Labs 1.0 — the first version, which Alex ran — where we were figuring out, oh, LLMs or something, what should we do with this?

    Alex Brooker: Yeah, that was fun.

    Kevin Hightower: Yeah. And then I became chief innovation officer — left the product side and focused 100% on innovation. Had a 25-person team for Propulsion Labs 2.0, and that's where we really built some of the substantive stuff. That just wrapped up the 1st of October, and those 20-odd people have moved back into the core business to bring what we've worked on into the products and really try to drive change. And I've now moved to Propulsion Labs 3.0, which is a smaller team — just four people — but we're focusing on horizon-three stuff, the really out-there things. So you can't talk about that too much, but all this other stuff we'll hopefully be able to talk about.

    Alex Brooker: That was a long-winded way to tell you how I got here — no, I mean it's an amazing story, Kevin. And I think maybe the first question that springs to mind is: because you've seen how products and systems are put together, from safety-critical to business-critical to more cutting-edge innovation and fast-paced — not even to mention Google — what reflections have you had about building products in those different environments?

    Kevin Hightower: I think the biggest thing to me is the different people that are product managers and how they think. And this is where diversity and inclusion makes all the difference in the world. There are so many different ideas out there, but also so many different ideas of how to get people to move forward. Like, even in the ATC world — you're not going to have a pencil anymore. How do you talk to somebody about that? Even in a safety-critical environment. And then in the Google world there was a saying: sometimes you have to build what the customer needs, not what they want.

    But then you have this other part — some product managers that always want to build something that nobody needs. And so the different views of that, and getting those people together to balance all three of those things out, is really what I've seen as the magic way to do things. You could have an all-star doing one thing, but how you actually do that across products, across customers, across different contexts — it may be as simple as doing one little thing, but if you have a rockstar who wants to make everything complicated, that doesn't work.

    So this diversity of having different thoughts and bringing them together — I think there's no one right way to be a product manager. And if you're a product manager listening to this, your current role might suit you perfectly, but when you move to the next role it might require a completely different approach. So how do you take pieces of all that with you? That's what I've tried to do in my role at Cirium — I pull in some Lockheed pieces. Formal testing, for example. Some of those concepts stick with you: I don't need a 70-page document to test this, but I sure should write the requirements pretty strictly.

    Alex Brooker: And people don't — or write down the requirements.

    Kevin Hightower: Or write down the requirements to start with, right?

    Alex Brooker: Yeah, that's always a good start too.

    Kevin Hightower: And so, yeah, I think that's what I've pulled out of all of that — that generation of thought. And a lot of times I think I've learned that when I first started, I went way too fast: first thought, go do things. And that gets you so far. But as you grow as a leader — and I'm a big fan of Myers-Briggs and understanding what your first thought actually is and then what to add to it — pulling that in has really helped. Yes, most of the time first thoughts are amazing, but make sure you're working through them and pulling all that together, and having the diversity of everyone's first thoughts, second thoughts, and third thoughts all coming together.

    Kevin Hightower: But on the other end of that, I've had a lot of influence from one of my bosses from Lockheed. An orthodox boss, used to say that it doesn't matter what the decision is, it should never take more than three hours of thought. Buying a car, buying a house, deciding to buy a company — it's three hours of thought. It might be 30 minutes and you go, "I need more data." Data goes, data goes. But that's not thought of like the decision. Bring the data back and then you're in. And if you spent three hours really thinking about it, the rest is relatively detrimental. Now, new data comes in, new things come in — that's a new choice.

    But what furniture do you buy? What car do you buy? Give yourself those three hours. And I've taken that to heart too. My younger self would be, "Three hours? I need three minutes for that." But my older self is like, "Well, I really need three days or three weeks." But as I got around all of these different roles, it's really come back to that three hours of actual conscious thought about the decision.

    It should be all you need, and more than that is overthinking. Well, again, if nothing's changing, right? If you get two hours in and you're like, "Oh, if I only knew this data," I'm like, cool, pause your thought, get the data, but don't think about it while you're getting the data. Get the data, bring it in. Now I've got this.

    I've got an hour left. It's kind of like what we're doing with LLMs nowadays, like you have six hours of runtime to go fix this — don't overthink it, don't underthink it. And it's just funny how all those things kind of link together.

    Alex Brooker: Yeah. So maybe I wanted to take that thread of the pace of innovation and doing things differently, and the pieces that you've brought with you. If you were to reflect back — if you were to go back into the safety-critical side of air traffic management, or all those kinds of systems that live in that world — do you think that they could be built, designed, and tested quicker using some of the techniques and approaches that you've developed and learned at Google, Cirium, and elsewhere?

    Kevin Hightower: Absolutely. I mean, I'm old, right? So my first jobs could have been — and everybody, right? Like, the things we did in the 90s and 2000s and 2010s seem slow nowadays. But that's just because of what the demands are for growth and how technology has changed. But to answer your question directly, yes. I think even the simple things like the formal tests that we were doing — do I have coverage? Am I testing all of these things? Where are the use cases that I'm not thinking of? Give me the thoughts that I could go and do it. Now go run a formal test on the safety-critical system and tell me if it passed or failed. We're not there yet.

    But all of those steps help. There was a quote about thinking about what you're generating. And it should help with that, right? Not your thought — what are you generating? So in my case, I had a hundred-page formal test document, a mapping of requirements, design systems, code reviews, and all the stuff that we had to go and generate. But then separate the thought out of that. You drive the thought, let it do the generating. And that's where the experience, knowing what questions to ask — application of AI to the right things. How do you ask the right question?

    Alex Brooker: Yes.

    Kevin Hightower: And that comes from a blend of experience, but also I think there's a lot of value in the original computer science education, right? Because what used to happen — you go to your degree, you understand all these concepts, you know what a binary tree is, you know all these things. You might need it in your interview, and then you go to the real thing and you're like, "I'm not doing binary trees anywhere. I'm solving bugs with things not lining up right in CSS and stuff like that."

    And so you get that experience, but you tend to lose and forget some of that original foundational stuff. So important for directing AI, right? It's got that education too, but telling it exactly what you mean by things and how you want things to be built is really, really important. So I think people coming out of college actually have this really good strength. You know, remember the kid out of college who used to come in after you were ten years in? He's like, "Should we be doing a binary tree here?" And you're like, "We do it this way — this is reality." But I think some of that's coming back now where you could actually apply a lot of that in much of what you build.

    Alex Brooker: But do you think that — because you have that experience, you've built and assembled, tested, and got those systems accepted into production — you have that domain knowledge, you know how to direct an AI of whatever flavor or color in terms of creating software or running through some of those use cases. What about those people who... Because we're at an inflection point ourselves in that we sort of learned things the older way, and now we've got AI that can kind of power up what we are doing. But for the people just coming up, just starting their career, what do you think about that cohort who perhaps might be using AI from the beginning?

    Kevin Hightower: I think it accelerates the learning part. So, for example, my formal test days — that was a year and a half of my life. Writing formal tests, getting approved by the FAA, manually making coverage reports, running the tests, being in the lab. Now you can do that in five minutes. But if you don't actually read it and you don't actually understand it, then you're not learning from it to decide what's next and what coverage it has, right?

    And so if the job entails not thinking about it, sure, AI is going to automate it. But if the job entails, "Well, you have to learn about these things" — okay, well, we had this kind of error and we never really tested this, I see a pattern and there's a hole — being able to do that faster is the value.

    And I love this quote. Neil deGrasse Tyson was once asked, if you ever met intelligent life from other worlds, what would trigger him that he's in the presence of intelligent life? And he said, "If I went in their house and their kid's homework from first grade was on the refrigerator and it was quantum math." Like, you've compressed all that knowledge back in and now they're doing quantum in first grade, which means they can do so much more later. I think that's where the new hires are. If you use AI just to do somebody else's old job, you're not going to get far. But if you can use it to learn, iterate, and do those things that used to take us a year and a half — and actually learn from it, and if your brain can comprehend it out of all of this — then I think you're even better off.

    But you've got to do the research. You've got to dig in. You've got to try all these different paths and not just say, "Hey AI, it's not working." That's not very helpful, right? Understanding, I think, would drive it. And I think that's a lost skill at the moment. I have not seen people do it in that way.

    Alex Brooker: Okay, that's interesting. So let's bring it right up to date on that same thread of how do you decide what to use AI for and how do you decide to use a domain expert? Cirium has a lot of data. I'm sure AI agents and LLMs love looking through it all and trying to figure things out. But how are you deciding what an AI agent should take on?

    Kevin Hightower: I think one of the benefits of AI is something we learned. So when I was at Google, we were part of X and their R&D area, and it's the whole "fail fast and celebrate failures" thing, right? Like, when they had a big project that got canceled early in its stage, they would fly everybody to Hawaii to celebrate. Like, "Thank God we found it here, not five years from now when we spent a hundred million dollars on it." If it's a real reason to fail — not "I didn't have staff" or "the technology's not there yet, we'll look again in ten years" — celebrate it, right?

    And so I think that's where I look at it from. Maybe this has some legs. Maybe there's some data in here that does X, Y, Z. Let me go run something that I'm going to throw away if it doesn't work, and be okay with it. So much of the time we get caught up in, "Oh no, there's something here" — run something, it's not there, let me keep running, let me grow it this way, the customer doesn't like it, let me force-feed it in a different way. And you're just hugging on it, because of the sweat and time that you've put into building it.

    If you can say, "Well, that sweat and time's not there and I could throw it away" — do it five different ways. And so that's the trigger for what to go and look at. Something that you thought of in the past, something that you see in your customers, something you've explored and discovered with your customers, read in the news, or dug in deep and gone, "Oh, I think there's something there that could save that." I would use AI to go explore that and not lock into the results. I think that's a culture change, right?

    Little baby projects that you've worked on — you're like, "Oh, let me just keep doing it." No, throw it away, start over. And now it doesn't cost very much to throw it away and start over, where before it cost a whole lot of money to throw it away and start over.

    Alex Brooker: Yeah. The cost is coming down.

    Kevin Hightower: Yeah. So I think that's the drive with AI for us — to get ideas started, go build it, go explore it, go poke and research at it, but expect to spend a few days, not a few months. Take everything that would normally be 10x and divide it by 10 instead. Shrink it down, use AI to help you with it, and don't be afraid if it's not right. Work on the next thing.

    Alex Brooker: So when the cost of software creation and software dev has gone down by 10x, where's the new bottleneck? Let's say creating software is free — we can just ask Claude Code to create a million prototypes of different combinations — but where's the bottleneck? How do you decide what to kill and what to keep?

    Kevin Hightower: SVPG had a good piece, I think it was probably two or three months ago, kind of on this topic, where it was — just because you can build it fast doesn't mean you should be building it. And so I relate to that. But you could build it and throw it away — that's the difference. At Google, some prototypes, obviously you have to direct it into what you're trying to prove. And quoting more Silicon Valley product thinking — the four risks. We could do this, but it's not usable by anybody because it'll be too hard to comprehend. Well, probably shouldn't do that thing.

    So I think that's where your domain knowledge and your customer base and your research has to come in, and that's the value that we give as humans that AI can't do — at least not yet. Drive into the problems and go, "Well, I've got this data, I don't have this data, maybe I need a partner for this data" — that two plus two equals five thing. That's the value that you bring. Now you can augment that with AI. You go, "This would be a whole lot better with this data — who has that data out there?" Back to generation. Used to take me a week to go research everything, now it takes five minutes. Got to validate it, got to go look at their websites, got to dig in a little bit, but now you've got a little bit more. So again, it's just that compression of time.

    But yes, that's where the domain knowledge really has to come in — understanding your problems, understanding your customers' problems, understanding their pain, the empathy about what they're doing in their day-to-day job. That's all what a good product manager needs to come with.

    Alex Brooker: Yeah. I mean, ChatGPT is very supportive of all of my ideas.

    Kevin Hightower: Every single one.

    Alex Brooker: Every single one. Build it. It's always a yes. Have you heard the trend of replying to bosses like AI?

    Kevin Hightower: You send them something like, "Are you sure that's right?" — "You're great to question that. Let me get back to you."

    Alex Brooker: There's a whole — I forget, there's probably a term for it — but there is this theme going on where you just talk like AI. Like old AI, not like new AI.

    Kevin Hightower: "You're absolutely right. You're absolutely right."

    Alex Brooker: "You're right to push back on that."

    Just following on that theme then, of those different ideas and what AI is most suited for, and where do we keep the domain knowledge and application of experience — I want to ask you about the way the market is developing. Lots of people are creating AI agents across the whole stack of travel and aviation. How do you see the market moving forward? Cirium's a data company — lots of APIs and products for end users. But how do you see the interplay between these different AI agents, who have perhaps different needs to a UX for a human being, and the products that they might use? How do you see this sort of playing out?

    Kevin Hightower: Yeah, I think it's kind of back to how AI has also moved, right? The thing that used to take you five minutes takes you a few seconds — great, but not amazing. It's the things that take you five days that now take five minutes where you have the really strong value. And kind of back to the ATC world — if you can start automating some of that stuff, or I can do it faster, what am I going to do with my free time?

    And I think it was actually Neil deGrasse Tyson that said this too — you and I both remembered phone numbers when we grew up. You had to. You had them in your head, you dialed them when you needed to. And at some point you didn't have to learn phone numbers anymore. What did we do with that extra space? Did we watch cat videos, or did we learn something and fill up that space with something else?

    And I think that's where we are with AI. Yeah, you could automate your job and have 50% more free time — but what are you going to do with it? Are you going to do your same job more, or are you now exposed to all these new things you could go and do?

    We used to draft airplanes with 70 people and big blueprint paper. Now, since the 777 was the first computer-generated aircraft, what have we done? We've made all these variants, we've driven into engines, we've done all these other things that we just didn't know were available yet. We were capable of it. So I think that's where AI is really going to help.

    Kevin Hightower: And so specifically in our aviation side of things, you can write a SQL query that gets your data out better. But that's a small increment.

    The way we like to look at it is we want to power every decision connected to aviation. So if you think about those decisions — I'm going to buy an airplane, I'm going to cancel this flight, I'm going to serve more vegetarian meals on this plane — there are more or less eight-ish steps in there. Get some data, do some things, right? And the decision to not do anything is also a decision, which is down in that layer. We think specifically in aviation and hard questions, we can get AI to not just do that first step, but get down to like four or five. It's not going to automate the whole thing.

    But it's going to get you all that data to make that decision. It might even suggest a decision based on whatever, but you still have to — that human in the loop has to be there to really understand what the options are. I think you actually ran one of our first tests with this, where it was like, hey, I've got three flights, they're all landing at the same time, what's the best way to not have any passengers miss their connections? And it said, cancel the flights — nobody will miss their connections.

    Alex Brooker: Right. Yeah. Well, it's a logical solution.

    Kevin Hightower: Correct. And so I think that opening up to do more things shouldn't be just more of the same. Sure, that's also got benefit, but now what are those decisions you're not making? What is it you never even looked at? Stock trading used to take a day; now it's milliseconds. Okay, now what do you do with it? Same thing, I think, is in aviation. If we can start automating a lot more, it takes a lot of that knowledge to get the path right for the decision-making process. And then, okay, now what do I do with this? Well, if I did it this way, maybe I could cascade and make these three decisions together. And then those three decisions together are a much bigger impact, right? And today, kind of like the traffic control world, I don't have the time for it, so I'm just going to make the easy decision and move on. So I think that's really where we're headed.

    And so we try to focus on actually providing value by digging into the problem. You see a dashboard with a dip — it's the first thing you do. You go look at the dip, you go ask questions. AI should be doing that for you before you even see the dip. Get that information out, get all of those automated pieces out. How are my competitors doing this better than me? You would never do that before. One, you wouldn't have the data. But if you have the data, let me go find every 737 in a medium-sized airport over here, see how they've turned their 737s. Let's take out weather days or match weather. You could spend three days on analytics for one flight, one turn, one problematic turn that's happening a lot, and then you've done one. And that's where AI really helps. Okay, well, let me do 1,000 of them.

    Out of these 500, you have this kind of problem; your competitor doesn't have this kind of problem. Go talk to the airport, go do this. We don't know what's driving it. We have signals, like, oh, you brought that delay in and it got worse, and every time you do that it gets worse. If you just bring that delay in down to less than 10 minutes, it's resolved right at this location. That you can see. And when you dig into it, maybe it's because the fuel truck has to go from the E gates to the A gates to get fuel. But if you do it while the fuel truck's still there, you don't have to wait for them to come back. Or whatever it might be that you dig into.

    But the trends and the analysis and actually digging into the values — and not just giving you a table or a chart — is where the real value is. That's what Cirium really wants to focus on from an AI perspective. And one of the things we're just launching now is the linkage of all this data. We have data everywhere, but now we've created what we call the aviation knowledge graph. The data in aviation is a graph. It all starts with the plane. You're not doing anything unless you have the plane. When was the plane ordered? Who bought it? How's it changed? Who's the lessor now? What is their actual drive for performance and how often do they want to use it? And then the schedule it's going to plan. And then that schedule changes a hundred times before it actually flies. And then you have a tail swap, and all these things are just connected. And so now we've connected it. That also gives us that layer of: I don't want just a dashboard, I don't want just a table, I want answers, I want trends, I want to know what I can go make decisions on. That's what we try to have all of our AI deliver.

    And a lot of problems do not matter.

    Alex Brooker: No, it sounds fantastic. And I really like the way you explained the idea of going behind the dashboard. An old-school dashboard with a dip in it — what AI is now enabling us to do is go behind that little dip and analyze every single contribution to it, which previously was impossible, just too much work for a human, even with the tools from five years ago.

    Kevin Hightower: Right. And you still see it today. Your Monday morning session: how'd last week go? Oh, we had a problem here, there was some weather. Okay, are you going to do anything about it? Are you going to make a decision, are you going to change anything? No, we'll look next week to see if the dip changed. Not making a decision is still a decision, but we think there could be so much more that you could do if you had the time. And AI gives you some of that time.

    Alex Brooker: So will all of these dashboards go away?

    Kevin Hightower: I don't think so. We're still humans. We're still going to make something, make it pretty in Excel, take a screenshot of it, and put it in PowerPoint for somebody to go see. Because that's the way aviation will be for a long time. Is Anthropic doing that? No. Is Google? Well, Google's legacy, so they're probably still doing a lot of that. But when you think about some of these new startups and big companies, they're not wasting time with that. So I think, yes, eventually dashboards will go away — probably not in you and my lifetime. Screenshots and PowerPoint will be here for a while.

    Now, we're already doing a lot of automated PowerPoint and things like that. But again, back to where you bring in the expertise — you used to make a 15-page PowerPoint to explain a detail and get to a conclusion. Can you get it down, compress it? Here's the facts, here's the decision, let's keep going. That might be more of a personal thing than a trend I see out there, but there was this great comic — I think you were around when we had it — where it's just two guys, one's at a computer, one's looking over his shoulder. He goes, look, I put these three bullets in and it makes this nice long email. And on the other side of the comic: look, it takes this nice long email and turns it into three bullets.

    Why don't we communicate that way? Why do we need all this? And so I think there's a lot of momentum going in that direction.

    But the human element is still there. You still need that buy-in that what we're doing is correct, what the strategy for the company is, and how that relates to what you want to go decide. And that needs to happen somehow — whether that's a 15-page PowerPoint with screenshots in it, or a more automated way where it's like, hey, Kevin suggested this, you suggested this to somebody else, this is what's front and center that you need to go decide. And I think there are going to be a lot of great tools to do that.

    Alex Brooker: Yeah. I think what hasn't happened yet is some of that rethinking of how we do work.

    Kevin Hightower: Hmm.

    Alex Brooker: So at the moment, we're still in the stage of when electricity was first brought into factories — they took out the steam engine and put in an electric motor, but still connected everything up to the rubber bands that were linking people's sewing machines to the power. Power distribution was still using the old pulleys and levers. I think we're still in that phase with knowledge work, where even using AI tools, it's just doing the same thing, albeit quicker.

    Kevin Hightower: Yeah, definitely in aviation. In other areas, I think it's moving a lot faster.

    Alex Brooker: Yeah, which areas? Which areas is it moving faster?

    Kevin Hightower: So I think in a few — well, one, the people building the systems. The Anthropics of the world, it's just faster. You head it down a path like, okay, we're going to augment it with X, Y, Z features that will make it better, and it comes back with X, Y, Z and one, two, three features that made it better, and nobody came up with those one, two, three features. So that's the extreme on one end. The other extreme is the physical world — it hasn't been taken over by robots or anything. The plumbing and maintenance and putting stuff together. And we're somewhere in that middle area.

    And I think on the high end of that middle area — what's a good industry to reference? If you think about software development startups, obviously they've always been ahead of the curve as far as meetings and processes go. But now you have startups of two or three people getting billion-dollar valuations, right? Because the automation is helping them. They don't need large teams, and they're going to grow, but you can bet they're not going to start doing 15-page PowerPoints. They're going to take what they've learned at three and five people and start applying it.

    I think there are some parts of banking that are there too, from a decision-making process standpoint, if not from a control-my-money-and-make-these-transactions standpoint. And I think you're going to see more of it — remember the term digital natives versus digital immigrants? Internet natives, internet immigrants? Same thing with AI. We are immigrants to AI. We did not have AI growing up. But the people that just came out of college in the last two years are native AI. And what are they going to do that's going to be different? How are they going to drive us?

    Alex Brooker: That's really interesting. Yeah. It's scary.

    Kevin Hightower: It is, it is, it is.

    Alex Brooker: So in terms of that pace of change — which, I didn't plan it this way, but that feels like a theme through our conversation — looking ahead, will those safety-critical parts of aviation start to move a lot quicker? Or perhaps, to frame the question a different way, which part of aviation is going to move quickest, and is it the same part that was moving quickest before?

    Kevin Hightower: Probably so. And actually, you see this with the FAA — there was a program they called SMART, or something like that, where ANSPs and others are there to keep planes from hitting each other. And what's the safest way to keep planes from hitting each other? Fly one flight at a time. You only fly one flight, and then when it lands, the next flight could take off. Nobody would ever crash into each other. Obviously you can't do that. So then it turns from safety to flow and efficiency, because those both have to play hand in hand.

    And where they fit — I think you're going to see a lot more of that efficiency. Because we have seen, I've worked on programs in the past that make efficiency better. Time-based flow management with the FAA, where things kind of get synchronized before they make it to the TRACON. The TRACON is safer because they don't have as many planes — they've been organized a bit before they got there. And so I think you see that with the FAA program they just launched. I think it's SMART, actually. One of our friends is involved in that.

    Alex Brooker: Yeah.

    Kevin Hightower: The efficiency parts and making better decisions before it becomes safety-critical — I think that will move faster and faster. And the safety-critical layer that keeps metal from hitting metal, which is very important, will be very strict, very manually controlled, very human-in-the-loop for a long time to come. But the better efficiency you can bring to it, the less chance of metal hitting metal. And so I see that efficiency part growing — more stuff on the ground, more sequencing, more planning the day out.

    But, as everybody in the industry knows, aviation is a beast. You give me $100 million, I couldn't tell you the arrival sequence in Charlotte today. Doesn't matter. There are too many elements in play. I could get it kind of right, and then one flight gets canceled because it ran over a piece of metal on the tarmac. And that just happened. Or a passenger got sick, and the sequence is going to change. We don't live in a deterministic world where you can say, let's automate the heck out of this and we'll be done with it. You live in a very volatile, very interdependent world where there are five times as many flights as there are airplanes, and that causes different kinds of friction.

    So yes, to answer your full question, traffic flow, traffic management, and traffic planning are all going to be much, much more automated, with much better decision-making. And that will make safety better. But I don't think AI is going to be in control of air traffic anytime soon.

    Alex Brooker: No, that makes a lot of sense. Automate the mundane stuff, automate the things you can call routine.

    Kevin Hightower: Yes. Automate the routine and free up the human to focus on the complexity where it's needed.

    Alex Brooker: Yes. Absolutely. Back to air traffic control in 2001 in the ocean. Yeah. So now on the business side — because I know a lot of the work that you do is more on the commercial side of things in terms of transport and aviation — how do you see things changing and accelerating on that side?

    Kevin Hightower: I'm going to say this nicely. Having been in aviation for a long time, those things don't change fast. It's either the mindset of safety that just percolates up, or it's just aviation in general — we're just slow to adopt. And so I don't see anything dramatically shifting in the next few years, other than maybe around traffic and fares, and maybe more flexibility in scheduling if things can move around faster. There are some opportunities there. But I don't see aviation as a whole fundamentally changing in the next five to ten years. We were at a talk in Riyadh, in Saudi Arabia —

    And they had a nice fireside chat, three or four people talking about the future of aviation, future travel. And one guy's like, "I'm just going to throw a grenade in here just so we can chat about something different." And he's like, the biggest threat to aviation, commercial aviation, is virtual reality.

    If you could set up a room in your house that looks like a beach, that feels like a beach, that has an automated machine that gives you drinks like a beach, for the price of one or two tickets — that's a threat to where you're going. But that's how far out we're thinking: that VR in a room is a better chance of changing how travel is than anything we're actually doing in travel.

    Kevin Hightower: And I think I agree with that. The advent of low-price carriers was probably the biggest change we've ever had, and it's just led to a plane being a bus instead of a plane being an experience. And I don't see a path where AI makes that better. I think AI makes that worse, right? AI fares, putting things against each other, finding that last middle seat in the back of a 30-year-old airplane. Where AI may help — is that helping passengers have a different travel experience? I don't think it's changing that. And that might not be the question, but I'm putting it into the travel sense of the traveler.

    Alex Brooker: Yeah. You're the first guest who has made that point — that maybe in the short and medium term, AI being applied to a lot of those travel use cases will make the passenger experience perhaps worse. More complicated? Is that right — more complicated? Or what is your thinking?

    Kevin Hightower: I think it goes hand in hand, right? Yes, that will happen. But also more on-time performance, fewer flight cancellations for maintenance — less of those things will happen, and that will make travel better. And that's again that kind of offline stuff, the planning side. Oh, I have better understanding of all the parts in the world whereas before I couldn't. I couldn't even look into that. I couldn't manually try to create some Excel file that says, "Hey, this part's been failing more often than it was originally intended to, so maybe I should plan for that." But now with AI I can, right? And I can go, "Hey, you have that thing in for maintenance — you should replace this part instead of waiting for next time." And that would in turn make travel better.

    But back to a level of what you'd expect from travel — I get to my destination when I want to get to my destination. Not, "I've got $50 fares to go anywhere in the world because AI has made this big improvement." Like someday, yes, the engines will get more efficient and AI will definitely drive that. New fuels, new ways to do things, fusion engines — someday. And AI will be part of that, but not yet. Not in our lifetime.

    Alex Brooker: Okay. Interesting. Robotics is a whole other area we could dive into for days. I think Japan Airlines has a robot that helps load ULDs into their cargo bay.

    Kevin Hightower: Oh, wow.

    Alex Brooker: And it might just be publicity, but they actually have a humanoid robot moving these ULDs. And that could have some better effects on maintenance and turnaround times and things like that. But again, we're not fundamentally changing what aviation is going to be — other than it being more efficient.

    Kevin Hightower: I see what you mean. So even automated ground support, all that ground service equipment, those little vehicles — that's not fundamentally changing it. That's more incremental, just doing the job better. And what do you do with that extra time? Are we going to go watch cat videos, or are we going to go do something different about aviation to make it better? I don't know what that is, right? Robots in the middle aisle helping people get in — "Let me take that luggage for you and put it up." I don't know. We're still a long way from that too.

    Alex Brooker: Yeah, I'm waiting for the robot gardener. My garden is a disaster zone and I need a robot gardener. We have the robot lawn mowers, so we could just flatten it for you, but not really build a nice garden.

    Kevin Hightower: Yeah, yeah. I'm holding my breath for that to come out. That's actually a great topic too. If you had robots to do all this, have the robots do your work and you go out and garden. Don't have AI replace all our pictures and our music and everything else — have it do the work. Fold my laundry, do the dishes, do the stuff so that I can go out and enjoy life. But that's a whole other philosophy question.

    Alex Brooker: Yeah, that's a different topic. I would fully automate my garden so that I can spend more time coding aviation AI agents.

    Kevin Hightower: We've got to get your priorities in order.

    Alex Brooker: Kevin, this has been amazing. It's so interesting talking to you about all these topics. For people who want to reach out to you or to Cirium, how do they do that? How can they reach you?

    Kevin Hightower: Yeah, LinkedIn is great. That's the best place for business. You've got my contact info — feel free to share it. But yeah, reach out on LinkedIn is probably the best way to have more of these conversations.

    Alex Brooker: Thanks, man. Now, just before we close the episode — what should I have asked you? What did I miss from my prep?

    Kevin Hightower: Well, the Cirium side of me says you should have asked about our products and what we sell. But I don't like talking about that anyway. So I think the question is: what's holding people back? Especially in aviation — what's holding people from doing this today?

    I'm part of a bunch of different AI groups, airports and such, and the theme I hear over and over again is around data governance, data structure, understanding what the data is and what to do with it and how to get access to it. And everybody's got these really long programs: "Well, let's call Snowflake and let's make a data lake," or, "Let's call Microsoft, put a Fabric together, and let's spend $10 million to put all our data in one place." And then after all of that, you still don't have anything that's of value to go do. You still have to build something on top of that.

    So I think that's a really big problem — probably in a lot of industries — but we've finished our digital transformation, and anything that's digital is still not accessible. And we get that a lot too: "We have great amounts of data. How do I get your data and my data together?"

    And if you're advanced enough — if you have your own data analytics team, your own IT team — we sell cloud warehouses, we sell streams, we sell APIs that can bring it in. But even that's an undertaking. So how do you really get this data interoperable across boundaries and still maintain authority over it? How do you understand data for what it is, rather than spend five years trying to clean it up?

    If you can identify what's actually wrong with the data and not use that part, or augment it with other data to fill in those holes, I think there's a lot to that. And so that's something that kind of keeps me up at night in terms of what the next innovation looks like. Can you solve data interoperability without doing it the way we've always said you could? Multi-million dollar data lakes, giving everybody — you know, our airline customer wants all of our data. "Here's 50 terabytes of data. What are you going to do with that?" That's not a solution.

    So how do all these things come together so that they can make those decisions, and their AI can make those decisions, and their AI has access to that data — and it's protected? When you were here, we talked about a patent we have on how to protect your data while using a third-party AI service. That's the part where it's like, I've got data everywhere — how do I bring it together, and bring it together fast? This isn't a five-year program. Because you get to the end of the five-year program and you go, "Well, great, now I have all this data, but there are holes in it, I'm still missing things, and I still can't do what I want to go do."

    And typically you end up working the other way around: "I have this problem. Let me get this set of data and this little set of data." Now you've got another silo somewhere else, and you do it again over here, and again over here, and you're just making the problem worse. And so that's something I think AI can help with. Data is the new oil — how do you get it powerful enough and still keep control? If my competitor knows what I paid for that airplane, I'm done. You still have to have control of it. But knowing what I paid for an airplane versus what the market value is versus what somebody shopping for it would pay — that's really valuable. It's that fine line of protecting yet integrating data. And this is not an aviation-only problem.

    Alex Brooker: That tension between a multi-million dollar data platform program and a point solution that perhaps causes more problems than it solves if you just keep doing the shortcut each time. And I think AI will help that too — how do I get a data lake faster?

    Kevin Hightower: So you'll shrink that time, probably shrink that cost. But let's say you cut it down to a year and a million dollars — there's still a year and a million dollars you've got to spend to get to where you need to be. And then during that year, something else has changed. In five years, you have to do it again because you've got all these other things that have come out, or new software-as-a-service somewhere else that you don't have access to the data from. And the proliferation of MCPs and how to connect to everything — that problem has to be solved sometime in the next year, two years, five years, or ten years, depending on how you look at it.

    And then the last thing — true craziness — is if fusion power, quantum computing, and AI continue at this pace and all three come together at the same time, there are a lot of interesting things ahead. When power's not a problem anymore —

    Alex Brooker: Power's not a problem, yeah.

    Kevin Hightower: Quantum computing can explore all these different paths of what an LLM could decide. Simplistic terms, but if those three come together — I mean, looking at history, only one happens at a time. AI has gotten better over the past five years. Fusion looks better, but it always looks better until it actually is a real thing. And quantum is making progress, but we still can't do a lot with it. But if all three of those happen at the same time — innovation will be unreal.

    Alex Brooker: I think AI is helping unlock the other two in terms of software systems and the way of designing things. Back to that formal testing and stuff — doing it faster and faster and faster.

    Kevin Hightower: Yeah, yeah.

    Alex Brooker: Listen, Kevin, thank you so much for recording this conversation and coming on the podcast. It's been an absolute pleasure.

    Kevin Hightower: Thank you so much for this discussion. Thanks for having me.

    Alex Brooker: Welcome back anytime.

    Previous Episode

    Ep 36: Code Is Twenty Percent of the Thing

    Next Episode

    Ep 38: Who Can Still Design a Plane?

    Related episodes

    Building AI for aviation or travel?

    Airside Labs builds working AI prototypes on your data and tests them against the edge cases of your operation. Our agent tools give Claude, ChatGPT or your own agent a catalogue of 6,700+ aviation AI use cases and aviation lookups it can trust.

    About the Travel Tech Podcast

    Hosted by Alex Brooker, founder of Airside Labs, the Travel Tech Podcast features authentic long-form conversations with leaders, innovators, and startup founders shaping the future of aviation and travel technology. No scripts, no sponsors, no sales pitches — just real conversations about building technology that matters.

    Have a story from travel or aviation tech? Come on as a guest.