FlyJets Launches JetGPT AI Agent After Beta Phase, Covering About 400 Aircraft

FlyJets, a New York-based charter marketplace, has moved its artificial intelligence agent, JetGPT, out of beta testing. The tool went live on the company's web and mobile apps this month, following a beta phase that began in February.

Users type or speak requests, such as all empty legs out of Van Nuys this week or all shared seats on heavy jets from Los Angeles to Teterboro this month. The agent returns availability and pricing from operators on the platform, and results can be exported to a spreadsheet. It searches only operators listed on FlyJets, covering about 400 aircraft, and the company said pricing and availability data are no more than five minutes old.

FlyJets is a New York benefit corporation that does not own or operate aircraft. It built the agent on a large language model it declined to name, using retrieval-augmented generation with its internal database as the source of authority. In a demonstration, CEO Jessica Fisher pulled a list of flights under $25,000, downloaded the results and listed seats on a charter for sale on the FlyJets exchange.

Users can book through JetGPT, which handles the early part of the process before passing them to the standard checkout pages. The human-staffed advisory desk remains available throughout. The company does not take payment until an operator confirms it can fly the trip or, for a shared flight, until the group is ready to go. Other AI agents can access FlyJets as they would another booking site but cannot use its model.

Fisher said the agent is an efficiency tool rather than a replacement for brokers, who see the same information as the end customer. She said clients are expected to use posted prices as a basis for starting negotiations. The biggest users so far are clients seeking large data sets or very specific searches, and the company said the agent is cutting time and back-and-forth from booking.

Fisher said JetGPT will not be used to tailor prices to individual customers, a practice known as profiling. All clients see the same pricing and aircraft information, and FlyJets does not build automated client profiles. Conversation data is stored in the company's internal database, and the underlying model is not permitted to train on it outside FlyJets' own use.

JetGPT is tied to the FlyJets exchange, which lets users sell individual spaces on chartered flights and resell empty legs. Shared flights are confirmed once they reach half capacity.

The next phase, to be rolled out over the coming year, will add pricing suggestions for people selling spaces. The second version is also planned to connect to flyer and operator dashboards, fleet operating models and the company's sustainability programme, which rewards clients who buy carbon offsets or sustainable aviation fuel. Fisher said FlyJets plans to move from a proprietary agent to a proprietary model.

A natural-language layer over a closed inventory changes where charter shopping starts. Questions that once went to a broker or a phone call, such as every empty leg from one airport this week, become a query against live data, and the export-to-spreadsheet function suits buyers who compare many options at once. The inventory still comes from listed operators, so the tool's usefulness is bounded by how many aircraft and operators sit on the platform, here about 400 aircraft.

The design keeps the commercial chokepoints where they were. Payment is not taken until an operator confirms the trip, and a human advisory desk stays in the loop, so the agent compresses search and early booking rather than operator confirmation. Brokers see the same information as the end customer, which narrows the information gap that part of their value has rested on, while posted prices are described as the starting point for negotiation rather than a final fare.

The next dated markers are the pricing suggestions for people selling spaces and the connection to operator dashboards and fleet operating models, both slated for the coming year. The planned shift from a proprietary agent to a proprietary model is the step that would show how far FlyJets intends to differentiate on the technology itself.