"Jazz Hands": The AI That Is Learning to Fly Airplanes — and the Industry That Is Starting to Believe It

At Quonset State Airport in North Kingstown, Rhode Island, a Cessna Caravan accelerated down the runway and climbed into the sky on Saturday morning while the pilot sitting in the left seat kept his hands folded in his lap. From a back seat, Tim Burns, the chief technology officer of the Boston-based startup Merlin Labs, watched over the airplane's intercom and offered a one-line commentary on what was happening. "Let's see those jazz hands," he said. The joke — casual, confident, deliberately understated — captured something precise about the moment Merlin Labs and the broader aviation industry are in: a technology that was theoretical five years ago and science fiction ten years before that is now demonstrating itself in real aircraft over real airspace, and the people building it have arrived at the stage where they can make jokes about it.

Test pilot Matt Diamond, sitting beside CNN's Pete Muntean in the left seat, was not flying the airplane. He was watching it fly itself, hands visible and open, ready to intervene if necessary but not intervening. The Merlin Pilot system — the artificial intelligence at the controls — handled the takeoff roll, the climb, the cruise, the radio communications with a mock air traffic controller, the approach, and the crosswind landing on runway 34, managing a slight crosswind all the way to touchdown with the kind of smooth, measured control that Diamond described with a technician's understatement: "It's a challenging problem for the automation. But once you crack it, it makes things much easier on the pilot." The airplane CNN rode in during the demonstration is labelled experimental, and Muntean was legally a test subject rather than a passenger. But what it demonstrated was not experimental in the sense of speculative. It was experimental in the sense of working.

The timing of CNN's reporting on Merlin Labs, published on May 24, 2026, is not coincidental. It reflects a convergence of forces that have been building in the aviation industry for several years and that in 2026 have arrived simultaneously in ways that make the question of artificial intelligence in the cockpit not just a technology story but an economic, regulatory, and safety story of the first order.

The most urgent pressure driving the AI aviation conversation is a pilot shortage of a scale that the aviation industry has not experienced in the modern era and that its current training pipelines cannot resolve within any timeframe that alleviates near-term operational constraints. Boeing's most recent projections estimate that the global aviation industry will need more than 600,000 new pilots over the next two decades. The number of pilots being trained and certified annually falls significantly short of what that trajectory demands. Regional airlines in the United States have been most acutely affected, with carriers struggling to fill first-officer seats at rates that have led some to cancel routes, retire aircraft early, and delay planned expansion. The pilot pipeline problem is structural: it takes years of training and flight hours to develop a commercial airline pilot, the cost of that training is substantial and falls primarily on the individual, and the shortage of instructors needed to train new pilots is itself a constraint on how quickly the pipeline can expand. Adding more training programmes does not produce more pilots on a timescale that solves the immediate and medium-term supply gap.

It is in this context that the argument for AI assistance in the cockpit — specifically, AI systems capable of reducing the crew complement required to fly certain types of operations — has moved from the margins of industry discussion to a position where it is being taken seriously by regulators, carriers, and investors simultaneously. Merlin Labs is the most visible of the companies making this argument in 2026, but it is not alone, and its strategy reflects a considered reading of where the regulatory and commercial pathway to AI aviation actually runs.

The Merlin Pilot system is designed, in its current form, not to replace pilots but to reduce how many pilots a specific operation requires. It is what engineers call an AI copilot — a system capable of handling all pilot functions during flight, enabling a single human pilot in the cockpit to monitor and supervise operations that would otherwise require two. This single-pilot or reduced-crew concept is the near-term commercial proposition for Merlin's technology, and the aircraft categories they are initially targeting reflect a careful reading of where regulatory and economic conditions are most favourable for that proposition's success.

The Cessna Caravan — the aircraft used in Saturday's Rhode Island demonstration — is a turboprop aircraft widely used in civilian cargo operations, medical transport, firefighting support, agricultural operations, and regional passenger service in small communities. It is the backbone of companies like FedEx Feeder operations and of bush flying operations in Alaska, where Merlin has previously conducted FAA-supported flight trials in conjunction with Everts Air Cargo and the University of Alaska Fairbanks UAS Test Site. These operations are already largely single-pilot, they serve missions where the cost of a second pilot is economically significant relative to the revenue of the flight, and they operate in conditions — remote locations, challenging weather environments, high-workload approaches — where an AI copilot providing decision support and workload relief to a single human pilot offers a safety argument, not just an economic one. Merlin's CTO Burns identified these mission categories with precision in prior interviews: cargo flights, firefighting, medical missions, agricultural operations on small turboprops. "These mission cases are typically flown single-pilot, are potentially high-risk, and put real pressure on cost and crew," he said. "This is where having an AI Pilot in the cockpit with you really could make a difference."

The Merlin Pilot system's technical architecture reflects the ambition of that proposition. It is not a sophisticated autopilot in the traditional sense — the autopilot systems that have been standard equipment on commercial aircraft since the 1970s are fundamentally rule-based, handling predefined manoeuvres under predefined conditions, requiring human pilot intervention whenever conditions fall outside those parameters. The Merlin Pilot is designed to handle a broader range of unexpected situations than traditional rule-based automation, using what the company describes as an AI-powered autonomy stack fed by sensors and cameras. That external sensor data is fed into flight control computers that tell the aircraft exactly where it is and where it needs to go, processing information in the way that a human pilot processes visual, auditory, and instrument data — dynamically, in context, with the capacity to respond to conditions that were not pre-programmed.

The natural language processing dimension of the system is among its most technically ambitious and most practically significant features. Air traffic control communication is a safety-critical function that requires precise, standardised phraseology and real-time response to sometimes complex and rapidly changing instructions. Traditional autopilot systems cannot listen to and respond to ATC. The Merlin Pilot can. It uses a natural language processing model to listen to instructions from air traffic control — in the CNN demonstration, from a mock controller simulating realistic ATC communications — and responds over the radio using a computerised voice that is capable of understanding and speaking in various accents and voice types. A safety pilot can intervene if anything is lost in translation, maintaining human oversight of the communication loop while the AI handles the routine interaction that comprises the overwhelming majority of normal flight operations.

The military contract that CNN's reporting highlighted is the most consequential external validation of the Merlin Pilot's technical credibility and is where the system is most likely to achieve its first operational deployment at scale. Merlin recently secured a contract worth more than $100 million with the United States Air Force to bring its technology to the C-130 cargo plane — the four-engine turboprop transport that has been the workhorse of American military airlift operations since the 1950s and whose current variant, the C-130J Super Hercules, remains in active production and operational service with the Air Force, Marine Corps, and numerous allied air forces globally. A separate contract with the US Special Operations Command, valued at $105 million, covers integration of the Merlin Pilot system and paths the way to flight demonstrations of the technology on the C-130J. The US Air Force had previously also engaged Merlin under a 19-month contract to enable extended-crew and fully uncrewed tanker operations with the KC-135. The scope of these military engagements — covering cargo, tanker, and special operations aircraft across multiple branches and commands — is what Merlin's leadership describes as the system's first major proving ground. Military aviation, which operates under different regulatory constraints than commercial aviation, provides an environment where an AI flight control system can be validated at operational scale in real mission conditions before the more demanding process of civilian FAA certification is completed.

That FAA certification pathway is where the timeline for commercial application of AI aviation technology becomes genuinely uncertain, and where the most honest conversations about what is and is not achievable within specific timeframes must happen. The FAA's regulatory framework for aircraft certification evolved over decades in response to specific accidents and failures, building a system of safety standards whose foundational logic is the demonstration that a new technology is at least as safe as the existing technology it replaces or augments. Certifying an AI flight control system for commercial passenger operations involves demonstrating that the system performs at or above the safety level of an experienced human pilot across the full range of conditions the aircraft might encounter — not just the normal conditions that comprise 99.9 percent of flights, but the rare and extreme conditions that constitute the certification edge cases where human judgment has historically been the last line of defence. That demonstration requires extensive flight testing, data accumulation, simulation validation, and regulatory review on a timeline that historically runs to years rather than months for novel avionics systems.

Merlin is pursuing FAA certification for the Cessna Caravan Merlin Pilot application as a first step, working in parallel with regulatory engagement in New Zealand, where the company has also been conducting flight trials with the Cessna Grand Caravan 208B. The Caravan certification, if achieved, would be a landmark: the first AI flight control system certified for commercial operations in the US National Airspace System. Burns has described the commercial target market for this first certification as regional cargo flights, firefighting and medical missions, and agricultural operations — civilian analogues of the same mission categories the military applications address, where the economics and the safety argument for single-pilot plus AI operations are most clearly aligned.

The political dimension of AI aviation has become newly relevant in 2026. Transportation Secretary Sean Duffy has publicly promoted artificial intelligence tools as part of the Trump administration's broader push to modernise the nation's air traffic management infrastructure — a priority that has taken on additional urgency following a series of high-profile near-collisions and the deadly collision involving an American Airlines regional jet and a military helicopter at Reagan National Airport in January 2025, which killed 67 people and produced the most intensive scrutiny of air traffic control staffing and technology in decades. Duffy's support for AI tools in aviation encompasses both the cockpit automation angle that Merlin represents and the broader question of AI augmentation of air traffic control, where systems that help controllers manage increasingly complex airspace with inadequate staffing have been identified as both a near-term safety tool and a longer-term structural solution to the controller shortage that parallels the pilot shortage.

Mykel Kochenderfer, a Stanford University professor who specialises in aviation safety and autonomous systems, offered the academic framing that puts the technology in its appropriate context. "Newer AI-assisted systems are designed to handle a broader range of unexpected situations than traditional rule-based automation," he said. This is the technically precise statement of what distinguishes systems like the Merlin Pilot from their autopilot predecessors, and it is simultaneously the statement of what makes them interesting and what makes their certification challenging. The same capacity for generalisation that allows an AI system to handle unexpected situations more effectively than rule-based automation is also the property that makes it harder to prove, through traditional regulatory processes designed around deterministic systems, that the AI system will behave safely in every possible condition it might encounter. The aviation safety community's ongoing engagement with this challenge is not resistance to AI but rather the application of the same rigorous standard to a new technology that has always governed what goes into the cockpit of a commercial aircraft.

While Merlin Labs is the company whose Rhode Island demonstration generated Saturday's CNN story, the broader landscape of AI integration in aviation extends far beyond autonomous flight control into every dimension of how airlines operate, how aircraft are maintained, how passengers are served, and how the airspace itself is managed.

Delta Air Lines has been among the most systematic American carriers in deploying AI tools across its operations. The airline has used AI-powered systems for predictive maintenance — identifying components likely to fail before they fail, reducing unscheduled maintenance events and the delays and cancellations they cause — and has been applying machine learning to maintenance data analysis and technical information retrieval at its Delta TechOps maintenance division. At MRO Americas, the major maintenance, repair, and overhaul industry conference held in Orlando in April, Delta TechOps representatives described a new application involving AI tools for expendable warranty processing — an administrative but commercially significant function where AI's ability to process large volumes of data faster and more accurately than manual review produces measurable operational and financial value.

At the same conference, Aviation Week's moderator noted the industry's caution about overreliance on AI technology even as examples of its successful deployment multiplied. The aviation industry's culture of safety, which has been built over the course of more than a century through a process of learning from accidents and building redundancy against failure at every level of the system, creates institutional caution around any new technology that is directly structural. AI tools applied to maintenance data analysis, customer service, pricing optimisation, and baggage handling can be adopted and iterated without the same safety certification demands as tools that directly control aircraft systems. But the directional logic of AI in aviation — toward greater autonomy, toward reduced crew requirements, toward AI systems that handle more of the functions currently performed by human pilots and controllers — eventually leads toward the most sensitive and most consequential questions the industry can confront.

Delta expects 20 percent of its flights to be priced by AI by the end of 2025, reflecting the extent to which dynamic pricing systems powered by machine learning have become standard revenue management tools. Coforge's Aeronova.AI framework, launched in 2026 specifically for airline retailing applications, aims to automate complex retailing workflows and aid real-time execution, helping airlines manage data flows, optimise pricing models, and deliver consistent customer experiences across booking channels. Singapore Airlines deploys Kris, an AI-powered customer service platform. United Airlines uses AI for scheduling optimisation and operational recovery. Air traffic management researchers including the UK's Project Bluebird are evaluating how AI can work alongside human controllers to make airspace management more efficient, reducing fuel consumption through better routing while improving safety through faster and more accurate conflict detection.

Riyadh Air, the new Saudi carrier that launched initial commercial service in early 2026, has positioned itself publicly as the world's first AI-native airline — a carrier whose operational DNA is built around AI systems from the ground up, in partnership with IBM Consulting, rather than retrofitting AI tools onto legacy infrastructure. The AI-native airline concept, in which artificial intelligence is not added to operations as a productivity tool but is foundational to how the airline thinks about every operational decision, represents the most ambitious near-term expression of where the industry's AI trajectory is leading. Whether Riyadh Air's execution matches the ambition of its positioning will become clearer over the coming year as its commercial operations expand.

The question that the Merlin Labs demonstration over Rhode Island raises most directly — and that Saturday's CNN story brought to millions of people who had not previously been tracking the technical progress of AI aviation — is the one that aviation safety experts, regulators, carriers, pilots' unions, and ordinary travellers are all in different stages of forming an answer to: when, and under what conditions, are they prepared to trust an AI system with the responsibility that human pilots currently bear? For military cargo on a C-130, the answer is forming in the affirmative, with a $105 million contract as evidence. For civilian cargo on a Cessna Caravan in Alaska, the FAA's prior engagement with Merlin's Everts Air Cargo trials suggests the regulatory process is actively underway. For regional passenger operations, the timeline extends further and the regulatory and public acceptance challenges are correspondingly greater. And for the commercial narrowbody and widebody passenger aircraft that carry the overwhelming majority of the world's air travellers every day, the honest answer from every serious participant in the conversation is the same: not yet, and the pathway requires much more demonstrated performance across a much wider range of conditions before the regulatory and social licence for that step can be earned.

Matt Diamond, taxiing the Cessna Caravan back to Merlin's hangar at Quonset State Airport after the AI landed the airplane in a crosswind, offered the practitioner's summary of where things stand. "It's a challenging problem for the automation. But once you crack it, it makes things much easier on the pilot." The AI cracked the crosswind landing on Saturday. The broader question of what it means to crack the full complexity of commercial aviation, and how long that takes, and what it costs in terms of testing and certification and trust-building along the way, is the question that the jazz hands over Rhode Island have raised for the entire industry to answer.

Sources

1. CNN — AI Is Learning to Fly Airplanes — and Aviation Is Starting to Embrace It — https://www.cnn.com/2026/05/24/us/ai-flying-airplanes
2. KTVZ — AI Is Learning to Fly Airplanes — and Aviation Is Starting to Embrace It — https://ktvz.com/news/national-world/cnn-national/2026/05/24/ai-is-learning-to-fly-airplanes-and-aviation-is-starting-to-embrace-it/
3. ePlaneAI — Massachusetts Startup Developing AI Copilots for Military Aircraft to Go Public — https://www.eplaneai.com/news/massachusetts-startup-developing-ai-copilots-for-military-aircraft-to-go-public
4. Runway Girl Network — Merlin Labs Sees Incremental Path to Pilot-Free Passenger Flights — https://runwaygirlnetwork.com/2026/03/merlin-labs-sees-incremental-path-to-pilot-free-passenger-flights/
5. Flying Magazine — Military Moves One Step Closer to Self-Flying C-130J — https://www.flyingmag.com/military-design-review-autonomous-c-130j/
6. New Atlas — Merlin Pilot Introduces AI System for Autonomous Large Cargo Aircraft — https://newatlas.com/aircraft/merlin-pilot-autonomous-large-cargo-aircraft
7. BusinessWire — Merlin Awarded FAA Contract to Deploy Nation's First Air Cargo Network Flown by Non-Human Pilot — https://www.businesswire.com/news/home/20230411006007/en
8. Travel and Tour World — United States to Navigate AI-Driven Aviation: How Artificial Intelligence Is Reshaping the Future of Flight — https://www.travelandtourworld.com/news/article/9us1srtp34zx/
9. Travel and Tour World — USA Travel Tech Alert: Coforge's Aeronova.AI Revolutionizes Airline Retailing and Transforms Passenger Travel Experience — https://www.travelandtourworld.com/news/article/7d6w011u5p9s/
10. aiOla — The Future of AI in Aviation 2026: A Flight Path to Success — https://aiola.ai/blog/future-of-ai-in-aviation/
11. Aviation Week — Airlines Mull AI's Potential, Limits for MRO — https://aviationweek.com/mro/emerging-technologies/airlines-mull-ais-potential-limits-mro
12. AirGuide Business — The Future of Airline Travel: How Technology, Premium Cabins, and Sustainability Are Reshaping Flying — https://airguide.info/future-of-airline-travel-2026/
13. Seeking Alpha — Riyadh Air and IBM Partner to Launch World's First AI-Native Airline — https://seekingalpha.com/news/4529175-riyadh-air-and-ibm-partner-to-launch-worlds-first-ai-native-airline
14. The Points Guy — How AI Is Transforming the Future of Travel — https://thepointsguy.com/travel/tpg-trends-ai-transforming-future-of-travel/