Two travelers need to be at a wedding by 15:00. They checked a bag, paid to sit together and bought both tickets on the same booking. At the airport, their first flight is canceled. An assistant finds two seats on the next departure. It leaves them on different connections and gets one traveler in after the ceremony.

The proposed seats were real. The replacement trip was still wrong. An airline agent has to recover the purpose of the journey, the passengers and the services they paid for. It also needs to know which changes the airline can actually make from that channel.

That makes disruption a more revealing test than routine booking. To see why, start with the choices passengers make before they ever reach the gate.

Airline demand, capacity and digital preferencesIATA measures traffic and capacity separately from surveyed passenger preferences.

What passengers want and what airlines can deliver

Demand and capacity. IATA reported that 2025 global passenger demand, measured in revenue passenger kilometers (RPKs), rose 5.3% from 2024. Available seat kilometers rose 5.2%, and passenger load factor was 83.6%. Load factor is an aggregate ratio of passenger traffic to seat capacity. It cannot tell an agent whether a specific alternative has inventory at the time of disruption.

Passenger preference. In IATA's 2025 Global Passenger Survey, 78% wanted to use a smartphone integrating a digital wallet, passport and loyalty cards for booking, payment and airport processes. Reported digital-wallet use rose from 20% in 2024 to 28% in 2025, an eight percentage-point change. The survey indicates appetite for a connected digital journey. It does not measure willingness to let an agent commit funds or rebook without approval.

Industry infrastructure. IATA's Modern Airline Retailing program uses NDC and ONE Order to support a transition to Offers and Orders. NDC covers shopping, order creation, payment and servicing data exchange. Capabilities still vary by airline, partner and channel. An implementation should maintain a capability matrix for involuntary changes, ancillary servicing, refunds and interline actions. McKinsey's B2C goods scenario explicitly excludes services, so it cannot be used to size airline agentic commerce.

Task reliability. Dell'Acqua and colleagues studied 758 knowledge workers in a preregistered field experiment with BCG consultants. AI assistance improved performance on some tasks. On one selected task outside the model's capabilities, the study reported an average correctness decline of 19 percentage points across the AI groups. The experiment did not test travel. It supports an evaluation design that separates routine shopping from unusual servicing and checks correctness before increasing authority.

The $180 fare that costs $245

Two travelers consult an agent at an airport rebooking desk
Rebooking has to preserve the traveler’s constraints, including onward segments and the services already paid for. Editorial illustration

Two passengers need to arrive before 15:00, travel together and check one bag each. In an illustrative quote, a $180 fare plus $45 for a bag and $20 for a seat costs $245 per passenger. A $230 fare that includes both costs $15 less per traveler, or $30 less for the pair, assuming otherwise equivalent terms. The agent should compare the bundle against the passenger's constraints and show any differences in change terms.

When the first segment is canceled, price comparison gives way to a recovery decision. The relevant unit is the entire order: passengers, onward segments, services and current policy. The agent can propose viable alternatives, but only the airline's approved workflow can establish availability and commit a change. Where it cannot confirm authority or retain a requested service, the agent should identify the unresolved point for an operator.

StageOperational testEvidence of success
Search and quoteCompare fares with bags, seating and different termsPrice and inclusions match the current bookable offer
Confirm and paySimulate offer expiry or repricing before paymentNew total is approved before commitment
RecoverCancel the first segment with onward flights affectedProposal preserves the itinerary's constraints or explains the trade-off
ServiceReplay a partial change or failed refund notificationOrder and financial states agree before status is communicated

The boundary runs through servicing

By 2030, offer comparison could be widely available while order servicing remains uneven across carriers and channels. One partner may show an itinerary correctly yet lack permission to reissue it after a change. Another may handle a voluntary change but be unable to preserve a paid service in a disruption.

Track the share of orders that can be serviced in the channel where they were booked, with the right passengers and ancillaries retained. A rising share supports giving the agent more responsibility. Where it stays low, a feasible proposal handed to an airline operator is a sensible outcome.

Test the itinerary, then replay the disruption

  1. Baseline. Choose one route family, one carrier and one distribution channel. Build cases from anonymous shopping and servicing records. Record quote accuracy, time to feasible offer, repeat contact and avoidable reissue cost.
  2. Shadow test. Evaluate routine cases and disruption cases independently. Include families on one booking, paid bags, accessibility requests and partial cancellations. A ticketing specialist should label each proposal's feasibility and policy compliance at the time it was made.
  3. Assisted launch. Permit agents to prepare offers and proposals for an operator. Compare against the current workflow by case severity and service retention. Measure completed outcomes, including cases the agent abstains from or hands off.
  4. Decision. Give write authority only to a narrowly defined action with confirmed order state and a recovery path. Stop on an unauthorized change or a confirmed loss of a required service. Review interline cases separately before expanding.

Evidence limits. IATA's traffic and survey figures describe the market. The Harvard–BCG result comes from consulting tasks. The offer comparison and pilot are GAEO.ai illustrations, so local baselines and recorded outcomes should decide where to invest.

Sources and related research