Skip to content

Journey Recovery Agent · Flights & Aviation

Your best next step when travel changes.

A supervised AI journey recovery agent for travelers affected by flight disruptions. It analyzes the impact, ranks recovery options around your priorities, and waits for your approval before any action.

AI-powered travel recovery • Atlas Travel APIs • Human approval by design

Demo scenario

CGK

Jakarta

SIN

Singapore

NRT

Tokyo Narita

Flight disruption detected

Your connection to NRT may be missed.

Connection at risk

Recovery options found

  • Fastest +Rp1.500.000
  • Cheapest +Rp650.000
  • Most comfortable +Rp2.100.000

Approval required before rebooking.

Agent activity

  1. Disruption detected
  2. Impact analysis completed
  3. 3 recovery options found
  4. Awaiting your approval

The problem

When your itinerary breaks, every decision matters.

Missed connections

A delay upstream can silently break a connection downstream — before anyone tells you.

Scattered information

Alternatives live across airline apps, OTAs, and airport screens. Gathering them costs precious time.

Decisions under pressure

Rebooking windows are short, and comparing price, time, and comfort in minutes is hard.

Hidden cost and downstream impact

A cheap fix can cost more later: hotels, missed transfers, and lost deposits.

The value

From disruption management to proactive journey orchestration.

Understand the impact

See what the disruption changes across your complete itinerary — connections, timing, and downstream plans.

Compare the right alternatives

Find options ranked around your time, budget, comfort, and priorities — each with a clear reason.

Stay in control

Review the trade-offs and approve any high-impact action yourself. The agent prepares; you decide.

Experience principles

  • Proactive. Alerts you when meaningful disruption occurs.

  • Context-aware. Uses the complete itinerary and preferences.

  • Transparent. Explains facts, assumptions, and recommendations.

  • Actionable. Always provides a clear next step.

  • Safe. Requires approval for high-impact actions.

  • Resilient. Falls back to Demo Mode when providers are unavailable.

  • Calm. Uses clear language during stressful events.

How it differs from what exists today

Traveler searches manually

Agent searches based on complete context

Separate sources and channels

Unified journey recovery workflow

Generic recommendations

Preference- and constraint-aware options

Passive chatbot response

Multi-agent planning and tool execution

Opaque recommendation

Transparent reason and trade-off

Automatic or manual action without context

Approval-controlled supervised action

Focus on one flight

Impact across the complete itinerary

Existing experience

MYTripMate AI

Agent workflow

How it works

Nine supervised agents carry your journey from disruption to recovery — and pause whenever your approval is required.

Phase 1 · Understand

The full journey context is loaded and the disruption is detected and analyzed.

  1. 1

    Context Agent

    Load context

    Trip, segments, preferences, budget, and priorities are loaded into one journey context.

    Result: Journey context ready

  2. 2

    Disruption Detection Agent

    Detect disruption

    Delays, cancellations, and schedule changes are identified on the active itinerary.

    Result: Disruption event recorded

  3. 3

    Impact Analysis Agent

    Analyze impact

    Connection risk, arrival shift, and downstream plans are measured across the itinerary.

    Result: Impact analysis ready

Phase 2 · Decide

Alternatives are searched, budget-checked, and ranked around your priorities.

  1. 4

    Recovery Search Agent

    Search alternatives

    Candidate recoveries are gathered through Atlas Travel APIs or a labelled Demo Mode adapter.

    Result: Candidate options gathered

  2. 5

    Budget Agent

    Check budget

    Every candidate is checked against your budget envelope and its price impact.

    Result: Price deltas computed

  3. 6

    Recommendation Agent

    Rank & recommend

    Options are ranked around price, speed, comfort, and your priorities — with reasons.

    Result: Transparent recommendation

Phase 3 · Recover

You approve, the agent executes or simulates, and monitoring stays active.

  1. 7

    Approval Agent

    Request approval

    High-impact actions pause here: you see the cost and trade-offs before anything runs.

    Result: Awaiting your decision

  2. 8

    Recovery Execution Agent

    Execute recovery

    Approved actions are executed or simulated, and every result is recorded.

    Result: Recovery applied

  3. 9

    Monitoring Agent

    Monitor & alert

    The recovered journey stays monitored, and new risks surface early.

    Result: Journey monitored

The orchestrator can pause at any stage — for clarification, missing data, or your approval. Nothing high-impact runs without you.

Agentic AI

More than a chatbot.

MYTripMate AI is not a passive chatbot. Nine cooperating agents plan, call travel tools, compare alternatives, explain their reasoning, and wait for your approval before acting.

Disruption Detection Agent

Identifies delays, cancellations, and schedule changes on your active itinerary.

Impact Analysis Agent

Measures connection risk, arrival impact, and downstream consequences.

Recovery Search Agent

Searches alternatives through Atlas Travel APIs or a labelled Demo Mode adapter.

Recommendation Agent

Ranks options around price, speed, comfort, and your priorities — with reasons.

Supporting agents in the workflow

  • Context Agent

    Loads and keeps the trip, segments, and traveler context up to date.

  • Budget Agent

    Checks price deltas against your budget envelope.

  • Approval Agent

    Pauses every high-impact action until you explicitly approve it.

  • Recovery Execution Agent

    Executes or simulates the approved recovery action and records the result.

  • Monitoring Agent

    Follows the recovered journey and raises new risks early.

Demo scenario · Simulated options

Choose the recovery that fits your priorities.

Cheapest

Rp3.450.000

+Rp650.000

Duration
14h 40m
Stops
2 stops
Arrival
Next day, 09:35
Risk
Low connection risk
Budget fit
High budget fit

Why this option

Best when protecting the budget matters more than arrival time.

Trade-off: Longer journey with an overnight layover.

Approval required before any rebooking.

Fastest Recommended in this demo scenario

Rp4.300.000

+Rp1.500.000

Duration
8h 25m
Stops
1 stop
Arrival
Today, 21:10 — 2 hours earlier than disrupted plan
Risk
Tight but viable connection
Budget fit
Above baseline budget

Why this option

Best when arriving on schedule protects downstream plans.

Trade-off: Higher price to recover the most time.

Approval required before any rebooking.

Most Comfortable

Rp4.900.000

+Rp2.100.000

Duration
10h 05m
Stops
1 stop · long layover lounge access
Arrival
Today, 23:40
Risk
Low connection risk
Budget fit
Premium segment

Why this option

Best when comfort and a low-stress connection come first.

Trade-off: Higher cost for cabin comfort and a calmer layover.

Approval required before any rebooking.

Safety & trust

Autonomous where it helps. Human when it matters.

MYTripMate AI can search, analyze, and recommend. It does not rebook, cancel, or charge you without explicit approval — ten safety rules govern every agent.

  • No invented availability

    The AI never fabricates flight availability, prices, schedules, or booking status.

  • Facts vs analysis

    Provider facts are always distinguished from AI-generated analysis.

  • Source-referenced recommendations

    Every recommendation references its source metadata.

  • No mutation without approval

    Booking, rebooking, cancellation, and payment wait for explicit approval.

  • Approval cannot be bypassed

    No natural-language interpretation can skip authorization or approval.

  • Validated structured output

    AI structured output is validated before it is used.

  • Untrusted content isolated

    Prompt injection and external content are treated as untrusted input.

  • Sensitive data protected

    Sensitive information is never unnecessarily included in prompts or logs.

  • Clarify before assuming

    When data is insufficient, the system asks or shows a warning — it never guesses.

  • Honest degradation

    Stale or unavailable provider data triggers an appropriate warning, not silence.

Recovery action

Approval preview

Rebook to selected option

Price change
+Rp1.500.000
Arrival change
2 hours earlier
Status
Awaiting your approval

Try the real flow in the interactive demo below — fully simulated.

Demo Mode — Simulated recovery action

See the journey recovery agent in action.

Run the real recovery pipeline — from disruption detection to a ranked recommendation waiting for your approval.

Build your own simulated trip — choose a route, date, and disruption — then walk through impact analysis, recovery options, your approval, and a monitored recovery.

Agent Chat

Or ask the recovery agent directly.

MYTripMate AI is conversational by design: it explains the disruption, compares recovery options, and asks for your approval — never acting on its own.

  • Context Agent: Loading your itinerary and preferences...

  • Impact Analysis Agent: Checking connection risk and arrival impact...

  • Recovery Search Agent: Searching simulated recovery options...

  • Budget Agent: Comparing price impact against your budget...

  • Recommendation Agent: Ranking cheapest, fastest, and most comfortable options...

  • Approval Agent: Waiting for your confirmation before any action...

Demo Mode · Simulated travel recovery assistant

Try it with the “Ask the recovery agent” button in the bottom-right corner of the page.

Technology

Built for the Alibaba Cloud x Atlas Agentic AI Hackathon

Laravel 12

Orchestration API and business workflow.

Atlas Travel APIs

Travel data and itinerary operations.

Alibaba Cloud / Qwen

AI reasoning and agent intelligence.

Qoder

Development, testing, and documentation workflow.

Redis Queues

Asynchronous agent jobs and monitoring.

MySQL

Traveler, trip, recommendation, approval, and audit data.

Human Approval Workflow

Supervised autonomy for every high-impact action.

Impact & growth

Turn disruption into a trusted service moment.

  • Higher recovery conversion

    Relevant, immediately actionable recovery recommendations convert stressful moments into resolutions.

  • Increased customer lifetime value

    Helping travelers during high-stress moments builds lasting loyalty.

  • Lower support workload

    Automated first-line disruption handling absorbs repetitive support questions.

  • Contextual ancillary revenue

    Recovery moments surface genuinely useful add-ons — hotels, transfers, insurance, lounges, and replacement activities.

  • Greater trust and retention

    A negative disruption becomes a positive service moment.

  • Better operational intelligence

    Aggregated disruption, recovery, and preference data enriches every future journey.

Target outcome — directional goals, not measured statistics.

FAQ

Frequently asked questions

MYTripMate AI is a supervised journey recovery agent for flight disruptions. It detects disruptions, analyzes the impact on your itinerary, finds recovery options, explains the trade-offs, and waits for your approval before any high-impact action.

No. A search tool lists flights for a new trip. MYTripMate AI starts from your existing, disrupted itinerary and reasons about the impact — connections, timing, budget, and downstream plans — before proposing recoveries.

No. Booking, rebooking, cancellation, and payment always require your explicit approval. The agent prepares and explains; you decide.

The disruption enters your journey context, the agent analyzes connection risk and arrival impact, searches alternatives, ranks them around your priorities, and presents a recommendation with the price impact — then waits for you.

The design integrates Atlas Travel APIs for travel data and itinerary operations. In this hackathon build, some flows use simulated data, which is always labelled as Demo Mode.

The page and demo run fully on simulated data. Options shown in Demo Mode are illustrative and never presented as live availability or booking status.

Yes. Recovery options are ranked around your budget envelope, time sensitivity, comfort preferences, and other priorities you set.

This is a hackathon demo. Interactions on this page are simulated and clearly labelled Demo Mode. No real booking or payment happens.

High-impact actions (rebooking, cancellation, payment) create an approval request showing the price and arrival impact. The action proceeds only if you approve it, and approvals can expire if conditions change.

This landing page uses no personal data. In the product design, traveler data is used to run recovery journeys, and every agent step is recorded in an audit timeline.

Be ready for what happens next.

Experience a calmer, clearer way to recover your journey when plans change.