CASE STUDY · AGENTIC AI
An orchestrated personal concierge for a travel platform.
Anonymised · Sector: Online travel platform · Solution partner: Brandsmashers Tech
- 6 agentsOne concierge coordinating five specialists across planning, booking and disruption
- ProactiveRebooking options prepared before the traveller has to ask
01 · PROJECT OVERVIEW
From catalogue to concierge.
A digital-first travel platform offers flights, stays and activities to leisure and business travellers. Customers value choice, but they value confidence and calm when plans change even more.
As part of a personalisation and service modernisation initiative, the platform set out to move from a search-and-book catalogue to a guided experience that stays with the traveller before, during and after a trip.
Brandsmashers Tech was engaged to design and build an orchestrated concierge.
- Designing the concierge and specialist agent architecture
- Integrating inventory, pricing, booking and notification services
- Building a consent-aware traveller profile and event stream
- Delivering conversational and notification experiences across web and mobile
- Testing complex travel scenarios before release
- 1Helping travellers plan faster with tailored itineraries.
- 2Remembering consented preferences across trips.
- 3Handling disruptions proactively instead of waiting for customers to call.
02 · THE CHALLENGE
Plenty of choice, not enough guidance.
The platform faced a gap between the choice it offered and the guidance travellers needed, especially when plans changed.
- PROBLEM 01Generic search and long decision cycles
Search results were the same for everyone. Travellers spent long sessions comparing options across tabs, and many left without booking.
- PROBLEM 02Preferences that were never remembered
Seat choices, dietary needs and family travel details had to be entered again on every trip, which made the experience feel impersonal.
- PROBLEM 03Disruptions handled reactively
Travellers usually discovered delays or cancellations on their own, then contacted support, creating spikes of anxious contacts at the worst moments.
- PROBLEM 04Dependent decisions across many systems
Dates affect flights, flights affect stays and stays affect activities. A single recommendation model could not hold that chain together or act when one part changed.
- PROBLEM 05Marketing disconnected from real plans
Promotions were not linked to what each traveller was actually planning or had enjoyed before, so relevance was low.
Travel is a chain of dependent decisions. A concierge has to hold the whole chain.
03 · THE AGENTS
One concierge, five specialists.
A conceptual architecture: the concierge talks to the traveller and coordinates specialists that share permissioned memory.
Converses with the traveller, understands goals and coordinates the specialists.
- Preference and memory agent
Maintains consented traveller preferences and trip history.
- Itinerary planning agent
Builds and compares itinerary options against budget, timing and preferences.
- Inventory and pricing agent
Queries live availability and price changes through partner APIs.
- Disruption watch agent
Monitors flights and bookings, and prepares alternatives before the traveller asks.
- Booking and servicing agent
Executes changes within defined rules, with the traveller’s confirmation for any paid action.
CHECKS BEFORE ANYTHING RUNS
04 · THE APPROACH
Personal with consent, proactive by design.
A balanced strategy: personalisation with consent, proactive service, and clear human support for stressful moments.
- AConcierge-led orchestration
- A concierge agent that converses with the traveller and coordinates specialists
- An itinerary planning agent that builds and compares options against budget, timing and preferences
- An inventory and pricing agent working with live partner availability
IMPACTTravellers describe a trip in plain language and receive two or three tailored options with reasons.
- BConsent-aware memory and personalisation
- A preference and memory agent that keeps traveller choices and trip history
- Transparent controls to view, change or delete preferences
- Relevant suggestions based on what the traveller actually enjoyed
IMPACTLess repetition, more relevance, and trust built through control.
- CProactive disruption handling
- A disruption watch agent monitoring flights and bookings
- Rebooking alternatives prepared before the traveller asks
- Proactive notifications with clear next steps
IMPACTMany travellers are informed and offered options before they need to call support.
- DSafe servicing and quality assurance
- Explicit traveller confirmation before any payment or change
- Human support for complex or high-stress situations
- Scenario testing for multi-leg trips, currency changes and conflicting preferences
IMPACTConfident automation that travellers can rely on.
- 1→Traveller request
- 2→Concierge plans
- 3→Specialists compare
- 4→Traveller confirms
- 5→Booking made
- 6Trip watched
05 · RESULTS
What changes.
Travellers plan in a conversation instead of across dozens of tabs, and when plans change, the platform reaches them first with options ready.
| MOMENT | BEFORE | AFTER |
|---|---|---|
| Planning | Dozens of tabs and filters | The traveller describes the trip in plain language and gets two or three tailored itineraries with reasons |
| Booking | Preferences re-entered every time | Preferences applied, with an option to change them |
| Flight delayed | The traveller finds out alone, then calls support | The disruption agent notices first, drafts rebooking options and notifies the traveller |
| After the trip | Generic promotions | Relevant suggestions based on what the traveller actually enjoyed |
Projected targets, not measured results. Specific percentages are intentionally not stated: a responsible estimate needs the platform’s own baseline data, which is established during discovery.
- ShorterTime from first search to confident bookingMeaningful reduction as comparison work moves to the agent.
- Modest upliftSearch-to-booking conversionVaries by business; set against the platform’s own baseline.
- LowerSupport contacts during disruptionsTravellers are informed and offered options early.
- Gradual riseRepeat bookingsFrom remembered preferences.
- ImprovedTraveller sentimentOn planning and disruption handling.
- Faster, calmer planning
Comparison work moves to the concierge; travellers choose between a few reasoned options.
- Trust through control
Preferences are remembered with consent and can be viewed, changed or deleted.
- Fewer anxious calls
Disruptions are spotted early, with alternatives ready before the traveller asks.
- A personal foundation that scales
A scalable, personalised service foundation for future growth.
06 · DELIVERABLES
How Brandsmashers built it.
- Agentic AI developmentThe concierge and specialist agents, with shared, permissioned memory.
- API and system integrationsInventory, pricing, booking and notification services.
- Data engineeringA consent-aware traveller profile and an event stream for disruption signals.
- Web and mobile developmentConversational and notification experiences across app and web.
- Quality assessmentScenario testing for multi-leg trips, currency changes and conflicting preferences.
- AI
- Agent orchestrationPermissioned memoryPlanning
- Integrations
- InventoryPricingBookingNotifications
- Data
- Consent-aware traveller profileDisruption event stream
- Experience
- Conversational UIPush notificationsWebMobile
- Quality
- Multi-leg tripsCurrency changesConflicting preferences
TAKEAWAYS
Lessons for consumer-facing leaders.
- 01Personalisation only works with transparent consent and control.
- 02Proactive service often matters more than faster reactive service.
- 03Keep a human path for stressful moments.
- 04Start with one journey, such as disruption handling, before the full concierge.
- Travel
- Hospitality
- Marketplaces
- Subscription platforms
- Experience platforms
YOUR TURN
Want a concierge your customers can trust?
Brandsmashers Tech designs agentic experiences that plan, remember with consent and act before customers have to ask. Start with one journey and we’ll help you prove it.