Autonomous Travel Planning and Disruption Management
Designing intent architecture and real-time re-delegation for complex multi-leg journeys
The Challenge
A travel platform designed an AI agent capable of planning, booking, and managing complex multi-leg journeys autonomously - flights, hotels, ground transport, dining, and experiences. The unique AXD challenge in travel is temporal complexity: a journey is not a single transaction but a chain of interdependent bookings where disruption to any link cascades through the entire itinerary. When a flight is cancelled at 2am, the agent must rebook the flight, adjust the hotel, reschedule the transfer, and notify the restaurant - all while the traveller sleeps. This is absent-state design at its most consequential.
AXD Approach
- ■Designed an Intent Architecture that captured travel preferences as outcome specifications rather than booking instructions: 'arrive in Tokyo by Tuesday evening, prefer direct flights, hotel within walking distance of the conference venue, budget £3,000 total' - giving the agent a mandate rather than a task list
- ■Implemented cascading delegation: the primary travel agent delegated to specialist sub-agents (flights, accommodation, ground transport) with explicit dependency mapping - if the flight agent changed a departure time, the accommodation agent automatically received updated constraints
- ■Built real-time disruption management with autonomous re-planning: when disruptions occurred, the agent assessed the full cascade impact, generated alternative itineraries ranked by proximity to original intent, and executed the best option within delegation boundaries - escalating to the traveller only when no acceptable alternative existed
- ■Created preference memory that evolved across trips: the agent learned not just stated preferences but revealed ones - the traveller who always chose the later flight despite requesting the earliest, the hotel style they actually enjoyed versus what they thought they wanted
- ■Designed a trust-proportional communication model: routine changes (seat assignment, minor timing adjustments) were logged silently, moderate changes (hotel swap within same category) generated a notification, and significant changes (route alteration, budget impact) triggered an interrupt requiring acknowledgement
AXD Principles Applied
- ◆Founding Principle 5: Outcomes Replace Outputs - intent architecture captured what the traveller wanted to achieve, not the specific bookings they wanted made
- ◆Founding Principle 3: Absence is the Primary Use State - disruption management was designed for the traveller who is asleep, in transit, or otherwise unreachable
- ◆Founding Principle 4: Relationships Have Temporality - preference memory accumulated across trips, improving the agent's understanding of the traveller over time
Design Outcomes
- →Intent Architecture enabled the agent to make creative rebooking decisions that honoured the traveller's goals even when specific bookings became unavailable
- →Cascading delegation prevented disruption to one booking from creating orphaned or conflicting bookings elsewhere in the itinerary
- →Autonomous disruption management resolved the majority of travel disruptions without waking the traveller, with full explanation available on return
- →Trust-proportional communication prevented notification fatigue while ensuring significant changes received appropriate human attention
Key AXD Insight
Travel is the domain where AXD's concept of absence is most literal. The traveller who is mid-flight when their connection is cancelled cannot be consulted. The agent must act with full authority, within designed boundaries, and produce an outcome the traveller would have chosen themselves. This is the ultimate test of intent architecture: can the agent honour the spirit of the mandate when the letter becomes impossible?
Related Reading
Frequently Asked Questions
How does AXD handle travel disruptions?
AXD handles travel disruptions through intent architecture (capturing goals not bookings), cascading delegation (automatic constraint propagation across sub-agents), and autonomous re-planning that honours the original mandate while adapting to changed circumstances.
What is intent architecture?
Intent architecture captures travel preferences as outcome specifications ('arrive in Tokyo by Tuesday evening') rather than booking instructions ('book flight BA123'), giving the agent a mandate that enables creative problem-solving when specific bookings become unavailable.
Apply These Principles
This case study illustrates AXD principles in context. To apply them to your own organisation, start with the AXD Readiness Assessment, explore the 12 frameworks in The Practice, or consult the AXD Playbook for a structured implementation guide.