The Cockpit Is Aviation's Most Under-Served AI Frontier — and the Industry Knows It
There’s a pattern worth naming directly: the aviation industry consistently talks about AI in flight operations as though the cockpit is the last place it belongs. Scheduling optimization, disruption recovery, fuel planning, crew tracking — these are the domains where AI investment is flowing. The flightdeck, by contrast, tends to get EFB feature updates and better chart rendering. That gap isn’t accidental, and it isn’t purely regulatory. It’s a choice the industry keeps making, largely because the cockpit is the one place where the stakes are obvious enough to make everyone cautious.
I think that caution is starting to cost the industry more than it saves.
Two Pilots, One Law, and What It Actually Signals
In early 2026, Congress passed legislation requiring at least two qualified pilots on all U.S. commercial airline flights — a move widely read as a direct response to single-pilot operations research and the creeping advance of autonomy programs. The bill is a statement of regulatory philosophy: human oversight stays in the loop, full stop. That’s defensible policy. But a significant number of people in the industry seem to have internalized it as something broader — as permission to stop investing meaningfully in AI tools that help the pilots who are undeniably going to be there anyway.
When I think about why that conflation persists, I don’t think it’s purely one thing. Some of it is genuine confusion — the autonomy conversation and the pilot-assistance conversation have been tangled together long enough that people have stopped separating them carefully. But some of it is also convenient. Pointing to the two-pilot mandate as a reason to slow-walk cockpit AI investment is easier than confronting the harder question of what it would actually take to build something worth deploying. Honestly, I think it’s roughly equal parts both.
The mandate doesn’t say AI can’t help those pilots plan better, brief smarter, process NOTAM loads faster, or surface deteriorating weather patterns before ATC does. It says two humans need to be in the seat. That’s a different constraint than “don’t build AI for the cockpit,” and conflating them is a strategic error.
Airbus’s DragonFly program, which tested a suite of pilot-assistance technologies on an A350-1000 beginning in 2023, is precisely the kind of work that shows what serious cockpit AI investment looks like. The fact that it hasn’t moved to commercial deployment yet is something people in this space point to as evidence that even Airbus is hedging — but I’d read it more charitably than that. Aviation certification timelines are long by design, and a program of that complexity moving from flight-test to commercial product in three years would actually be fast by historical standards. The pace reads as normal to me, even if the broader institutional caution around cockpit AI is real.
Where the AI Is Actually Going
The Aviation Today piece on next-generation avionics notes that AI-assisted flight management is “optimizing routes dynamically, taking into account weather, air traffic congestion, and fuel efficiency targets” — but if you look closely at where those systems actually live, they’re predominantly in the flight planning and dispatch chain, not in tools the pilots themselves are interacting with in real time. The data gets cleaner, the route gets optimized, and then it gets loaded into the FMS and the pilot flies what the ground system decided. That’s not cockpit AI. That’s ground AI with a cockpit endpoint.
There’s nothing wrong with that model — but it’s not a substitute for pilot-facing decision support. Optimizing a departure route before pushback is genuinely useful. Giving a crew situational intelligence during the flight — adaptive weather routing when the forecast breaks wrong, NOTAM context surfaced at the moment it becomes relevant, real-time turbulence modeling feeding directly into the EFB rather than being relayed through a datalink message — that’s a different and, I’d argue, higher-value problem.
A lot of current development energy is going into NOTAM parsing and presentation, which strikes me as the right place to start. The NOTAM briefing problem is real, it’s well-understood, and it’s one where AI can add immediate, legible value without requiring the kind of real-time autonomy that makes certification teams nervous. But I’d like to see the industry move more directly toward turbulence avoidance next. That’s a workflow where the pilot-facing gap is operationally consequential in a way that’s hard to route around — better ground-based forecasting helps, but it doesn’t fully substitute for intelligence that’s adaptive and crew-accessible during flight.
The EFB platforms that have pushed hardest into this space — phase-aware information surfacing, AI-assisted NOTAM parsing, contextual weather overlays — are still mostly challengers, not the dominant players. The dominant players compete on chart coverage, reliability, and device management programs. Those are real and important, but they’re infrastructure competition, not intelligence competition.
The Question the Industry Keeps Avoiding
The honest version of this debate isn’t “should AI be in the cockpit?” — it’s already there in rudimentary form. The honest version is: “why haven’t we committed to making it genuinely useful at the point where a pilot actually needs it, rather than stopping at the point where it becomes politically complicated?”
Having spent time working on commercial flight deck products, I think the buyer-user misalignment piece of this is real and worth taking seriously. The airlines procuring EFB platforms are not the same people flying them, and that gap in perspective tends to pull procurement conversations toward compliance requirements and device cost rather than toward the quality of in-flight intelligence. But I don’t think you can put all of that on the buyer side. Vendors haven’t yet built something so compellingly useful that it forces the conversation — the barrier is genuinely on both sides. The procurement culture shapes what vendors prioritize, and what vendors have built so far hasn’t been disruptive enough to change that procurement culture. Both things are true at once.
The cockpit is where the operation either holds or doesn’t. It seems like the right place to put the investment.