How to choose AI for airlines: a buyer's guide to agentic AI solutions

Where does AI for airlines pay off first, and why?

An airline runs on documents and checks as much as on aircraft: dispatch releases and NOTAMs, airworthiness directives and service bulletins, maintenance records and deferral logs, leases, crew licenses and training files, fare filings, refund requests, bag claims, air waybills, safety reports and revenue coupons. Much of the cost sits in skilled people reading those documents, comparing them with a rule or a record, and keying what they find. That is the work AI does well, because the answer is on a page and can be checked.

Margins leave little room for that cost. In June 2026, IATA cut its forecast for 2026 industry net profit to $23 billion on revenue of $1.17 trillion, a 2.0% net margin, with fuel forecast at $351 billion, or 31.4% of operating costs. When fuel and labor are hard to move, time recovered in operations, maintenance and the back office is one of the few levers left.

Agentic AI pays off first where three things hold: the volume is high, the inputs are documents or messages, and a qualified person already reviews the result. In those places agents read, check and draft, and the dispatcher, engineer, crew controller, refund analyst or revenue accountant decides faster with the evidence in front of them.

Which airline processes should we start with?

Start with work that is document-heavy, measured in minutes, dollars or deadlines, and already reviewed, so the human check stays in place while the reading and keying go away. Leave the release of a flight, the return of an aircraft to service, the crew legality call and every regulatory filing with the certificated or accountable people who sign them today.

• Flight operations and dispatch: releases assembled and checked before the dispatcher and captain sign, NOTAMs ranked for one flight with the original text, and recovery plans drafted for the operations center when weather or IT breaks the schedule.
• Maintenance and engineering: each new airworthiness directive matched to the tails it applies to, every deferred-defect clock tracked across the fleet, and technicians' fault questions answered from manuals and fleet history, with the page cited.
• Fleet and leasing: lease obligations and return conditions turned into tracked dates, and aircraft records searched or audited for gaps long before a redelivery.
• Crew: every license, medical and training record kept current, and pairings and rosters checked against FAA Part 117 and EASA flight time limits before they publish.
• Revenue, distribution and NDC: fare filings checked before they go out, agency debit memos evidenced and defended, and NDC changes and refunds serviced for agencies.
• Customer service and disruption: disrupted passengers rebooked by policy, required refunds decided on time with the reason, and complaints answered inside the DOT clocks.
• Loyalty, airports and ground: missing-miles claims validated, ground handling invoices matched to turn records, and mishandled bag claims settled within liability limits.
• Cargo: emailed quote requests turned into priced drafts, air waybill data checked before acceptance, and Shipper's Declarations checked against the IATA Dangerous Goods Regulations.
• Safety, regulatory, finance and inflight: safety reports classified and routed, rule changes from the FAA, DOT, TSA, EASA and ICAO turned into owners and procedure changes, flown coupons matched to sales, and fuel and catering invoices matched to uplift and loads.

What should we demand from a vendor of AI for airlines on citations, sign-off and audit?

Demand that every extracted value, every match and every drafted sentence links to the exact passage it came from: the paragraph of the directive, the line of the maintenance record, the clause of the crew contract, the fare rule or the refund request. An answer without a source should be treated as unverified, especially where an inspector or auditor will later ask how a conclusion was reached.

Demand that accuracy is measured on your own documents before go-live and monitored after, field by field and decision by decision, against what your experienced staff concluded. Ask for error rates and for the cases the system sends to a person, not a single headline number.

Demand named human sign-off wherever responsibility sits with a person: the dispatcher and captain on a release, the certifying technician or inspector on maintenance work, the records engineer on a back-to-birth trace, the crew controller on a legality call, the analyst who issues a refund and the manager who files a report with a regulator. The person who signs should see the evidence, the rule applied and the draft, and the system should record who signed and when.

In aviation, the record is often the product. A compliance package that cites the directive, the effectivity and the task card is what a lessor, an auditor or a regulator relies on; a summary without that trail creates more review work, not less.

What about security, data residency and the systems airlines already run?

Airline data spans passenger records and payment data, crew personal and medical information, security programs, safety reports given under confidentiality, and commercial terms with lessors, partners and agencies. Ask where the data lives, who can see it and what happens to it afterwards. Botminds runs as multi-tenant SaaS, in a dedicated cloud environment, or fully on-premises, with the same controls in each, so data can stay in the region your obligations require. Your inputs and outputs are not used to train models.

Look for encryption in transit and at rest (Botminds uses TLS and AES-256), tenant isolation, single sign-on with MFA, role-based access to each dataset and workflow, and an audit log that records every human and agent action and cannot be altered. Botminds holds SOC 2 Type II and ISO 27001 and supports GDPR, with audit reports available on request. Safety and security material needs access drawn tighter than the rest, visible only to the roles entitled to it.

Ask how the vendor works with the systems you already run: reservations and departure control, flight planning, maintenance and engineering, crew management, revenue accounting, cargo, ERP and the shared mailboxes where much of the work still arrives. A good answer reads from those systems and writes back through approved routes, so staff are not asked to keep a second copy of the truth. Expect your own security and third-party risk review of the AI vendor, and plan for it from the start.

How do we run a pilot of agentic AI in airline operations?

Pick one process, one team and a fixed window, such as 6 to 10 weeks. Agree before you start what you will measure, and take a baseline from the current process so the comparison is real.

• Use your own recent documents, including the hard ones: scanned records from a mid-life aircraft, handwritten task cards, irregular-operations days, multi-leg refund cases and messy air waybills.
• Measure cycle time end to end, such as hours to assess a new directive, days to close a records gap, or days to issue a required refund.
• Measure touch time for the dispatcher, engineer, analyst or agent, and count the corrections they make.
• Count money found or protected: invoice overcharges caught, debit memos defended, refunds paid on time instead of late, delay minutes avoided.
• Check that every output a person signs can be traced to its source page, and that the audit log shows who did what.
• Decide in advance what result means expand, adjust or stop.

Should we build, buy point tools, or use a platform?

Building in-house gives control but means owning document reading, citations, review screens, access control, audit and model changes for every process, usually with an IT team already stretched by operational systems. Point tools solve one step well, but an airline with maintenance, crew, refunds, cargo and revenue accounting work ends up with many vendors, many security reviews and evidence spread across many places.

Adoption is already broad and scaling is the harder part. In Oliver Wyman's 2025 MRO survey, 64% of respondents reported using AI, up from 58% a year earlier, and 58% said their AI investments met or beat expectations, up from 20%; the firm also reports that many maintenance organizations remain at the pilot stage, held back by data quality and legacy systems.

A platform with a library of ready solutions sits between the two: each solution fits a specific airline process, and all share one security model, one audit log and one way of citing and signing. Botminds is the Agentic Solution Engine: it builds governed, agent-driven solutions from a specification and delivers them live. Its airline library has 116 solutions in 14 areas, from flight operations and dispatch to maintenance, fleet and leasing, crew, revenue and NDC, customer service, loyalty, airports, cargo, safety, regulatory compliance, revenue accounting and inflight supply.

Whatever you choose, check that you can export your documents, extracted data and decisions, so changing course never means losing your records.

What changes for dispatchers, technicians, crew schedulers and agents?

The job moves from searching and typing to judging. Dispatchers review a release that is already assembled and checked; engineers decide effectivity with the directive and the fleet records side by side; crew controllers see the legal options first; refund and complaint teams work the cases the rules do not settle. Corrections should become rules for the next case, so the same fix is not made twice.

Roles that sign stay with people: the captain and dispatcher who release a flight, the technician who returns an aircraft to service, the director of maintenance, the accountable manager, the safety and security leads and the analyst who issues money. Plan training on reading the evidence trail, and update manuals and procedures to say where AI drafts and where people decide.

Demand on these teams is rising. IATA reports that air cargo volume reached a record in 2025, up 3.4% on 2024. Oliver Wyman's 2026 MRO survey puts maintenance spend above $136 billion in 2025 and close to $140 billion in 2026, with labor and material shortages named as top concerns.

What does it cost to wait?

Every month without change is another month of paying skilled people to search and re-key, and of delays and payouts found late. Airlines for America puts the average cost of aircraft block time for US passenger airlines at $100.76 per minute in 2024, from DOT Form 41 data. SITA's 2026 Baggage IT Insights reported 4.9 mishandled bags per 1,000 passengers in 2025, down 23% on the year, still costing the industry an estimated $6.3 billion.

Rules with fixed clocks are also in force or arriving, and each one asks for evidence that is hard to assemble by hand:

• DOT refunds: under the April 2024 final rule, airlines must refund automatically when a flight is canceled or significantly changed, such as a domestic delay of 3 hours or more or an international delay of 6 hours or more, within 7 business days for card purchases and 20 calendar days for other payments, and must refund the bag fee when a checked bag misses the delivery window (12 hours after a domestic arrival, 15 to 30 hours internationally). Most requirements apply from 28 October 2024. In July 2026 DOT extended a pause on enforcement for renumbered flights to 7 July 2027 while it reconsiders the definition of a cancellation.
• Accessible travel: DOT's wheelchair rule, published 17 December 2024 and effective 16 January 2025, presumes an airline mishandled a wheelchair or scooter returned in a different condition and adds training requirements for staff who handle devices or assist passengers.
• Safety management: the FAA's April 2024 rule extends safety management system requirements to Part 135 operators, Part 91.147 air tour operators and certain Part 21 certificate holders, with operators due to declare compliance by 28 May 2027.
• Emissions: for the first CORSIA phase (2024 to 2026), airlines must cancel eligible emissions units by 31 January 2028, and an EDF analysis in July 2026 counted only about 41 million eligible units available.

Solutions to look at first

• Airline Disruption ManagementA crew of agents builds recovery plans for the OCC when weather or IT breaks the schedule
• Flight Dispatch ReleaseAgents assemble and check each release, ETOPS included, so dispatcher and captain sign it
• Airworthiness Directive Compliance TrackingReads each new AD, decides which tails it applies to and proves compliance, for engineers
• Deferred Defect ManagementTracks every MEL deferral clock and plans each fix before it expires, for maintenance control
• Aircraft Records ManagementAsk anything across an aircraft's scanned records and get the answer with its source page
• Aircraft Redelivery ManagementAgents run a leased aircraft's return to the lessor, from 24 months out to signed acceptance
• Crew Qualification ManagementA crew records repository keeping every license, medical, passport and training record current
• Flight and Duty Time ComplianceChecks pairings and rosters against FAA Part 117, 121.467 and EASA FTL, for crew planners
• Airline Refund AutomationDecides and issues DOT-required automatic refunds on time, with the reason, for refund teams
• IROPS Passenger ReaccommodationRebooks and notifies every disrupted passenger by policy and value, for disruption teams
• Mishandled Baggage ClaimsSettles delayed, damaged and pilfered bag claims within liability limits, for baggage teams
• Dangerous Goods Acceptance CheckChecks Shipper's Declarations against the IATA DGR and flags issues, for DG acceptance staff
• Aviation Safety ReportingReads, classifies and routes every employee safety report for airline safety analysts
• Passenger Revenue AccountingAgents match each flown coupon to its sale and book earned revenue, for revenue accountants

Every airlines solution →

A checklist for any vendor

1. Does every extracted value, match and drafted sentence link to the exact page it came from?A good answer is a live demo on your document: click an effectivity call, a refund decision or a crew qualification date and land on the line it came from.
2. How will you measure accuracy on our documents before go-live, and keep measuring after?Look for field-level and decision-level results on your own sample, compared with answers your staff confirmed, and ongoing monitoring that catches drift.
3. Where does a person sign, and what does the system record when they do?Flight releases, return to service, records traces, crew legality, refunds and regulatory reports should each need a named approver, with the evidence shown and the signature logged.
4. What happens to the cases the system is unsure about?They should go to a person with the reason stated and the evidence attached, never be passed through silently or dropped.
5. How does it work with the systems we run today?It should read from reservations, departure control, maintenance, crew, revenue accounting and cargo systems and the shared mailboxes, and write back through approved routes, with no second copy of the truth.
6. Where can it be deployed, and where does our data live?Options should include multi-tenant SaaS, a dedicated cloud environment and on-premises, with the same controls in each and data kept in the region you need.
7. Are our inputs or outputs used to train your models or anyone else's?The answer should be a plain no, written into the contract.
8. Which security certifications and audit reports can you share?Expect a current SOC 2 Type II report and ISO 27001 certificate, plus encryption, SSO with MFA, role-based access and an unalterable audit log.
9. How do you keep safety reports, security material and crew personal data apart from everything else?Access should be set per dataset and role, with retention rules and verifiable deletion, so confidential material is seen only by those entitled to it.
10. Can we export our documents, extracted data and decisions at any time?Yes, in standard formats, so you keep your records whatever you decide later.

Questions