Anatomy of a mass disruption
On 19 July 2024, a faulty software update cascaded through global IT systems and grounded thousands of flights worldwide within hours. Airlines that had planned for mechanical failures, weather delays, and crew shortages found themselves facing something they had not modelled: total system incapacity across every channel simultaneously. The event affected an estimated 8.5 million passengers globally and cost the airline industry approximately USD $1.5 billion in direct losses.
But that was just the headline figure. The downstream costs – customer compensation claims, regulatory fines, reputational damage, loyalty programme attrition, and the operational cost of manual recovery – pushed the true impact significantly higher. Airlines that handled the disruption well retained customer loyalty. Airlines that handled it poorly saw booking declines that persisted months after services returned to normal.
Mass disruption events in aviation are not rare. The Bureau of Infrastructure and Transport Research Economics (BITRE) reports that Australian domestic airlines cancelled approximately 3.8 per cent of scheduled services in the year to June 2025, with cancellation rates spiking above 10 per cent during weather events. International operations are even more vulnerable: volcanic eruptions, geopolitical conflicts, pandemic-related restrictions, and technology failures have each triggered mass cancellations in the past five years.
The pattern is always the same. An event triggers mass cancellations. Thousands of passengers simultaneously need rebooking, refunds, hotel accommodation, and information about their rights. They call the airline. Every single one of them calls the airline. Within minutes, queue times exceed an hour. The airline's app crashes under load. Social media fills with frustrated passengers posting screenshots of hold screens. The disruption becomes a customer experience catastrophe layered on top of an operational one.
Why contact centres buckle
Airline contact centres are staffed for normal operations with a margin for expected peak periods. A typical Australian domestic carrier might staff for 200 concurrent calls during peak hours, with the ability to surge to 300 through overtime and schedule adjustments. During a mass disruption affecting 50 flights carrying 10,000 passengers, the actual demand can exceed 3,000 concurrent call attempts within the first hour. No staffing model can absorb a tenfold increase in demand.
The mathematics of queueing theory make this clear. When arrival rate exceeds service rate by even a small margin, queues grow without bound. When arrival rate exceeds service rate by a factor of ten, the system is not congested – it is broken. The average wait time does not increase linearly; it increases exponentially. At ten times capacity, a centre designed for a 2-minute average wait time would theoretically produce wait times exceeding 20 hours.
In practice, most callers abandon after 30 to 45 minutes, call back, and are counted again – creating a feedback loop of demand that keeps the system saturated long after it would otherwise have recovered. Airlines typically take 48 to 72 hours to clear the backlog from a major disruption event, during which every other customer service interaction – routine bookings, loyalty enquiries, special assistance requests – is also delayed.
| Disruption Phase | Without AI | With AI Surge Response |
|---|---|---|
| 0–1 hour | Queues overflow, 60+ min waits | AI absorbs surge, <3 min wait |
| 1–6 hours | System saturated, mass abandonment | Rebooking underway at scale |
| 6–24 hours | Backlog growing, social media crisis | 70% of affected passengers rebooked |
| 24–72 hours | Still clearing backlog | Normal operations resumed |
AI rebooking at scale
AI-powered rebooking during mass disruption is not simply answering calls faster – it is resolving the underlying problem at a pace that matches the scale of the event. When a flight is cancelled, each affected passenger needs the same core service: find the next available flight to their destination, rebook them on it, and handle any ancillary requirements such as hotel accommodation, meal vouchers, or connecting flight adjustments.
A multi-agent AI system can process these rebookings simultaneously rather than sequentially. While a human agent handles one rebooking at a time – typically taking 8 to 12 minutes per passenger for a disrupted itinerary – an AI system can conduct hundreds of rebooking conversations concurrently. Each conversation is connected to the airline's reservation system in real time, seeing actual seat availability and fare rules rather than working from cached data.
The AI agent identifies the passenger, retrieves their itinerary, determines which flights are affected, searches for alternatives that match the passenger's preferences (timing, class, routing), presents the options, confirms the selection, updates the booking, and issues a new confirmation – all within a natural voice conversation that typically takes 3 to 5 minutes.
For complex rebookings – passengers with multi-segment itineraries, unaccompanied minors, passengers requiring special assistance, or those holding tickets with restrictive fare rules – the AI agent can either handle the complexity directly (using SurgeGuard's elastic capacity) or escalate to a human agent with full context. The critical difference is that simple rebookings, which typically represent 60 to 70 per cent of disruption calls, are cleared immediately rather than occupying human agents who could be handling the complex cases.
Multilingual crisis comms
International airline disruptions present a challenge that domestic operations do not: the affected passengers speak dozens of different languages. When a typhoon grounds flights across Southeast Asian hubs, the passengers stranded in Sydney, Melbourne, and Brisbane include native speakers of Mandarin, Cantonese, Japanese, Korean, Thai, Vietnamese, Hindi, and Indonesian – in addition to English. Staffing a contact centre with agents fluent in all these languages on short notice is functionally impossible.
AI voice agents solve this through real-time multilingual capability. The same agent that conducts a rebooking conversation in English can conduct the next one in Mandarin, then Vietnamese, then Japanese – with native-quality pronunciation, culturally appropriate register, and natural conversational flow in each language. This is not translation overlaid on an English script. It is genuinely multilingual conversation that understands and responds in the passenger's preferred language from the first word.
During a disruption, this capability is not a convenience – it is a duty-of-care requirement. Passengers who cannot communicate in English are the most vulnerable during disruption events. They are less likely to understand airport announcements, less able to navigate rebooking kiosks, and most dependent on telephone-based assistance. An AI system that can serve them in their own language, immediately, without queue time, provides a level of service that most airlines cannot match with human staffing alone.
Proactive vs reactive service
The traditional airline disruption response model is entirely reactive: wait for passengers to call, then try to help them. This model guarantees long queues, frustrated customers, and a recovery period measured in days rather than hours. The alternative is proactive service – reaching out to affected passengers before they call, with rebooking options already prepared.
AI makes proactive disruption management operationally feasible. When a flight is cancelled, the system can immediately identify all affected passengers and initiate outbound contact. Rather than waiting for 180 passengers to individually call and queue, the AI calls them – each passenger receiving a personalised call within minutes of the cancellation, presenting rebooking options tailored to their itinerary and preferences.
The passenger experience shifts from "calling the airline in frustration and waiting an hour" to "receiving a call from the airline within ten minutes with a solution already proposed." The emotional dynamic changes entirely. A passenger who is contacted proactively with a rebooking option feels cared for. A passenger who cannot get through after an hour on hold feels abandoned. Both may receive the same rebooking, but their perception of the airline – and their likelihood of booking with that airline again – will be dramatically different.
Proactive outbound also reduces the total volume of inbound calls. Every passenger rebooked through outbound contact is a passenger who does not call in. If outbound contact reaches 70 per cent of affected passengers before they call, inbound volume drops to 30 per cent of what it would otherwise be – bringing it within the capacity of existing human teams to handle the complex cases and exceptions.
Building disruption resilience
Disruption resilience is not a product – it is an operational capability that requires planning, infrastructure, and tested processes. The airlines that handle disruptions well are those that have invested in the systems and protocols before the disruption occurs, not those that scramble to improvise when it happens.
The first element is elastic capacity. An airline's AI system needs the ability to scale from handling routine call volumes to absorbing a tenfold surge within minutes, without degradation in response quality or system performance. This requires infrastructure that is provisioned for peak, not average – or more efficiently, infrastructure that scales dynamically in response to demand. SurgeGuard is designed specifically for this pattern: dormant during normal operations, instantly active when demand exceeds threshold.
The second element is integration depth. An AI agent that can only provide information is useful during a disruption. An AI agent that can actually rebook passengers, issue hotel vouchers, process refunds, and update loyalty accounts is transformational. This requires deep, real-time integration with the airline's passenger service system (PSS), revenue management system, hotel management platform, and loyalty programme. These integrations need to be built, tested, and maintained before they are needed.
The third element is scenario planning. Not every disruption is the same. A weather event at a single airport is different from a system-wide IT failure. A crew shortage on one route is different from a volcanic ash cloud grounding all flights across a region. Each scenario produces different call volumes, different passenger needs, and different resolution options. The AI system needs pre-configured response playbooks for each scenario type, with the ability to adapt in real time as the situation evolves.
The airlines that build disruption resilience now – before the next mass event – will protect their revenue, their reputation, and their customer relationships. The cost of preparation is a fraction of the cost of failure. And in an industry where the next disruption is not a question of if but when, the investment case is unambiguous.