Explore how to harness AI to personalise travel content at scale for each users unique experience and context, enhancing relevance and engagement.
You’re booking your next holiday or travel adventure, and the itinerary is coming together. You’re excited, right? But then you land on a checkout page or get a receipt email.
It’s almost impressive how anyone can make the thrill of travel feel so … mundane. Booking a trip should be wonderful, but these interactions somehow manage to drain away the magic.
Instead of adding value by offering something meaningful, they are added as an afterthought. They're boring, generic and uninspiring. The engagement rates are minuscule.
Why don't they work?
They probably don’t know you, your preferences or what would actually make your trip better. Yet even if they do, and you’re a regular customer, they have few tools to suggest things that are compelling and personal specifically for you.
It just isn’t scalable.
At least, it wasn’t.
The opportunity for travel businesses is to harness AI to craft engaging experiences unique to each user’s particular context, offering a level of creativity and personalisation that would be impossible without LLM’s.
Our Al-powered solution leverages contextual understanding to make suggestions tailored precisely to the user's unique needs and trip context.
It generates relevant and practical material specific to each travelers trip context such as what to pack, weather insights or important trip information.
It assesses factors like destination climate, cultural norms, and user preferences to provide practical, user-centred guidance for packing, weather, visa requirements or and other important information that the AI identifies as relevant to that specific user.
Experiment with the AI generator by changing the persona parameters.
The contents of this example page were entirely generated by AI from an input of 10 parameters about a persona.
With guidance from a leading AI researcher in Australia, we created a rapid mockup using Wordware.ai, feeding a set of inputs (name, age, and nationality) into LLMs to extrapolate detailed personality traits and preferences and match users with accommodation.
While it technically functioned, it also revealed a hard truth: our solution offered no real advantage over existing recommendation algorithms.
As we investigated existing systems more deeply, we discovered large companies had already built sophisticated algorithms to match users with destinations, accommodations, and activities and our assumption that travel checkout pages are ineffective and fail to leverage the extensive customer data they collect was incorrect.
While existing recommendation systems are smart, they lacked personal context, so our new focus revolved around each users unique context. For example, a solo traveller from (warm) Singapore would prepare for Norway very differently than a family from (cold) Canada; their interests, packing needs, and essential information could vary significantly.
This led to a redefined process:
1. Expand the user persona with context
2. Evaluate how the user’s background affects their travel experience.
3. Match the user with specific options that go beyond generic suggestions.
4. Generate specific content.
Our refined approach encountered challenges specific to LLMs:
Balancing structure and creativity:
Structured outputs (e.g., a standard welcome message) often constrained the AI, preventing it from adding creative elements like suggesting nearby cities to explore. We experimented with prompts to strike a balance, but results remained inconsistent.
Providing Useful Depth:
Context-specific depth was crucial but challenging. For example, a traveller from a warm climate visiting Europe in winter would need detailed packing advice. We considered using conditional queries under each section to dynamically fetch and return this level of detail.
Avoiding Stereotypes:
Relying on LLMs to extrapolate user traits occasionally led to unhelpful assumptions—like associating certain occupations or nationalities with overly stereotypical interests.
