Airline intelligence / XiamenAir

XiamenAir uses AI to review safety reports and parse NOTAMs

XiamenAir and Huawei launched a set of AI applications in September 2025: an agent that pre-processes safety reports, customer service agents and a large model that parses NOTAMs for dispatch. Huawei says the safety report process is more than 40% more efficient. Neither company has published accuracy figures.

The operating picture

Large models in safety, service and dispatch

XiamenAir B737-86N, B-5309, MSN 2254 (05 2007), as MF 842 Singapore (SIN) - Hangzhou (HGH), Flight time 4 45 (28987179114). Photograph: Charles from Port Chester, New York / CC BY 2.0 (cropped to fit). Illustrates the airline, not an AI deployment.

Airlines receive large volumes of text that staff must read before they can act: crew safety reports, passenger enquiries and NOTAMs. At Huawei Connect in September 2025, XiamenAir and Huawei released a smart aviation solution built on the airline's operating data and Huawei's cloud and large models, aimed at that work.

XiamenAir presents the work as part of its Smart XiamenAir strategy. Huawei said the two companies would extend AI into more core areas of the airline, without naming them.

Selected implementations

The work behind the evidence

01 / Deployed

Safety reports read first by an AI agent

The operational problem

Safety teams must read every report to find those that need action, and dispatchers work through a stream of NOTAMs around the clock.

What the evidence reports

Huawei said on 18 September 2025 that XiamenAir's safety report agent moves review from manual processing to AI pre-processing followed by human review, raising overall process efficiency by more than 40%. Customer service agents generate replies within seconds, and a large model parses the hundreds of thousands of NOTAMs the airline receives each year to reduce the round-the-clock workload in dispatch.

Operational significance

Human review stays in the safety process, with the AI sorting and summarising first. That is the cautious design a safety system needs, and the 40% gain describes throughput, not judgement.

Evidence still needed

The figure is Huawei's, without baseline or method, and no accuracy figure is given for NOTAM parsing. No XiamenAir executive is quoted by name in the sources reviewed.

Sources and verification (3)

163.com (civil aviation trade coverage) · Evidence grade B

Huawei (official news) · Evidence grade B

Huawei: XiamenAir and Huawei release Smart Aviation solution · Evidence grade B

A source can verify an announcement without verifying its projected outcome.

The wider register

Other tracked activity

Other records are shown separately, including non-AI technology and programmes at different stages of deployment. A register entry does not establish a measured AI benefit.

No other publicly listed records.

Editorial assessment

XiamenAir's AI targets the reading load in safety, customer service and dispatch, with people still making the decisions.

The next evidence worth having is accuracy: how often the safety agent misjudges a report, and how reliably the NOTAM model extracts the restrictions that matter to dispatch.

Live data from the Aerostack register. Status reflects published evidence only. See Our approach for how records are graded.

Aerostack tracks AI deployments across airports and airlines, with the published evidence behind each record. Photography is real aviation imagery. Record thumbnails may show the named airport or airline; other photographs do not.

Search Aerostack

Rydoo Accounts Payable AI invoice automation at Norwegian

GoodMaps indoor navigation at MidAmerica St. Louis Airport

Stan robotic valet parking at Lyon Saint-Exupéry

Facial-recognition eGates at Heathrow

Q-Gate biometric passport-control terminals at Almaty

Arrow Analytics EZ-Board computer-vision carry-on analytics test at PIT Gate A7

FAA SMART airspace and trajectory management platform

Dubai Police AI systems for passenger screening, unattended baggage and camera analysis at Dubai airports