Spring Airlines builds its own AI to read NOTAMs and plan flight recovery
Spring Airlines has developed its own AI to read notices to airmen and pass the critical ones to the right people, and a flight recovery system that combines machine learning with optimisation to replan the schedule when typhoons or faults strike. The airline puts the recovery system's retiming accuracy at up to 87.6%.
The operating picture
Built in house, aimed at the operation
B-6258 Airbus A320 Spring Airlines (7338877986). Photograph: Aeroprints.com / CC BY-SA 3.0 (cropped to fit). Illustrates the airline, not an AI deployment.
Spring Airlines, the low cost carrier based in Shanghai, develops its own revenue management, network, crew rostering and operations control systems, according to its 2025 environmental, social and governance report. The report says AI reached use at scale in key parts of the business in 2025.
The featured systems were built by the airline's own IT teams. The same team developed a flight group assistant that uses natural language processing to answer key messages, which Spring says raised the reply rate by 90%.
Selected implementations
The work behind the evidence
01 / Deployed
An AI assistant for notices to airmen
The operational problem
Pilots and dispatchers face a large daily volume of NOTAMs, written in coded text, and a critical notice missed in the volume is a safety risk.
What the evidence reports
Spring Airlines' 2025 ESG report, published on 30 April 2026, says it introduced AI to NOTAM processing in 2025 to address persistent bottlenecks and risks in manual handling. The system combines a specialist knowledge base with an AI parsing engine to recognise notices automatically, structure them and send them to the right recipients. The report describes an "AI plus human" way of working, with an AI NOTAM assistant developed by the airline itself that parses several types of critical notice, and says Spring presented the work at the national civil aviation aeronautical information user conference in September 2025.
Operational significance
NOTAM overload is a familiar safety concern in flight operations, and this system keeps a person in the loop while the AI does the sorting.
Evidence still needed
The ESG report gives no accuracy, volume or time saving figures for the NOTAM system.
A source can verify an announcement without verifying its projected outcome.
02 / Deployed
Machine learning to replan a disrupted day
The operational problem
When a typhoon, flow control or a technical fault disrupts the schedule, controllers must retime and reassign flights quickly, often across tens of thousands of flights.
What the evidence reports
Spring Airlines described its flight allocation system in a staff profile on its news site on 6 March 2026. Built by its IT operations control team, the system combines machine learning with operations research optimisation to generate recovery plans, and Spring put the accuracy of its retiming recommendations at up to 87.6%. It can simulate a delay with one click and replay past operations. The flight allocation project won an award in a national civil aviation innovation contest in 2025.
Operational significance
Recovery planning is where minutes of controller time translate into delays and costs, and Spring has put a figure on how often the system's advice is right.
Evidence still needed
The accuracy figures come from a staff profile on the airline's own site. Spring has not defined how accuracy is measured or published the effect on punctuality.
Warning of diversions and delays before they happen
The operational problem
Operations teams usually respond to diversions and delays after they start, when the options are fewer and dearer.
What the evidence reports
Spring Airlines said in the same 6 March 2026 profile that its team built a platform that monitors risks across operations and service and uses AI prediction models to identify potential diversions and delays in advance, moving from handling problems after the event to warning before it. Spring puts the platform's warning accuracy above 95%, and says it applies the regulator's principle of early detection and early warning.
Operational significance
Early warning gives controllers time to act, which links this system to the recovery tool that replans the day.
Evidence still needed
The 95% figure is the airline's own, with no definition of how it is measured or how far ahead the warnings come.
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.
NLP flight-group assistant at Spring Airlines · Deployed
A source can verify an announcement without verifying its projected outcome.
Editorial assessment
Spring Airlines is building operational AI itself, aimed at NOTAMs and disruption recovery, two of the most demanding tasks in an airline's operations centre.
A fuller account of how the accuracy figures are measured, and what the systems have done to punctuality, would strengthen the case.
Live data from the Aerostack register. Status reflects published evidence only. See Our approach for how records are graded.
Aerostack register · 4 tracked deployments
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.
.jpg)