01
Week 1
MVP in 24 Hours: itinerary composition
24-hour engagement on the AI itinerary composition surface. Shipped a working prototype that produced a real itinerary from a mood-based brief. Prototype produced real buyer conversations by week 2.
Travel & eCommerceFirst Class Flyer
Premium travel commerce with AI-native itinerary composition.
Product engineering and brand for a premium travel commerce operator — with AI-composed itineraries and a design system that respects the price point.
Client
First Class Flyer
Sector
Travel & eCommerce
Duration
9 months
Team
5 (product engineers, an AI engineer, a brand designer, a UX designer)
The client
About First Class Flyer.
First Class Flyer sells premium-cabin travel experiences to buyers whose average ticket price is in the tens of thousands of dollars. The company's clients don't complete a purchase from a form — they compare options for weeks, ask their advisor questions, and book through a conversation. Its digital surfaces have to do the job of a considered brand campaign, an editorial travel publication, and a conversation-triggering conversion funnel, all at once.
Context
Where First Class Flyer was when we started.
First Class Flyer sells premium-cabin travel experiences — the kind of buyer who compares four options over two weeks and then books via a call, not a form. The company needed a digital surface that respected that buying journey: informative, editorial, and quietly persuasive rather than aggressive-conversion optimised.
Challenge
The problem, unvarnished.
- AI-composed itineraries that respect the buyer's mood, not just the buyer's history — a real editorial voice, not a chatbot.
- Content-heavy product surfaces without the SEO trickery the sector defaults to.
- A brand and design system that reads premium without shouting it — closer to a print magazine's typographic rhythm than a booking-site's CTA density.
- Booking flows that end in a human conversation — not a payment form. The conversion metric is 'advisor conversation started', not 'checkout completed'.
- The company's technical stack was scrappy and appropriate for the previous scale; the growth ahead needed both a brand upgrade and a platform upgrade in the same window.
Approach
How we scoped and sequenced the work.
01
MVP first, brand second
Started with MVP in 24 Hours on the AI itinerary composition surface. Proved the interaction pattern before investing in the brand and design system that would carry it.
02
AI as editorial voice, not as chatbot
The AI writes itineraries the way a good travel advisor would — informed, opinionated, honest about tradeoffs. Not a conversation surface; a considered piece of writing on demand.
03
Design system that respects the price point
Editorial, unhurried, and photographically rich — with the typography and pacing of a premium print magazine, adapted to the web and to mobile.
04
Human handoff as the default conversion path
The site's job is to bring the buyer into a conversation, not to close on the form. Every high-intent surface ends in a warm handoff to a human advisor.
Timeline
How the engagement unfolded.
02
Weeks 2–8
Brand system & editorial voice
Brand identity, typography system, photography direction, and editorial voice. Shipped a first-week usable brand-system MVP; extended over the following six weeks. Voice guide co-authored with the client's founder.
03
Months 3–6
Full commerce surface & handoff engineering
Full itinerary product surface, browsing patterns, and the booking-flow engineering that ends in an advisor handoff. Advisor tooling integrated with the itinerary surface — advisors see the itinerary the buyer read before the conversation starts.
04
Months 6–9
Personalisation & advisor-tooling extensions
Personalisation layer on top of the itinerary surface. Advisor tooling extended with buyer-history and preference signal. Handoff-conversion measured against the pre-change baseline.
Solution
What we shipped.
AI-composed itinerary product surface, brand and design system, and the booking-flow engineering that ends in a human handoff rather than a checkout page. Built to move a premium-travel buyer from research into conversation.
Architecture & decisions
The choices behind the build.
01
Sanity CMS for editorial content, Next.js for delivery
Editorial content — itineraries, brand assets, marketing copy — lives in Sanity. Delivery runs on Next.js with edge caching. Editorial workflow is owned by the client's team; engineering doesn't gate content updates.
02
Model-agnostic itinerary composition layer
The itinerary composition layer sits between the prompt library and the model provider. Provider is a routing decision — OpenAI and Anthropic are both wired in and can be swapped without a code change.
03
Advisor context is a first-class object
When a buyer reads an itinerary and starts a conversation, the advisor sees the itinerary the buyer read — with the buyer's edits, the model's confidence, and the alternative options that were considered. Context isn't reconstructed; it's transferred.
04
Cloudflare + Vercel edge for global performance
Buyers are global, images are heavy, and time-to-first-byte matters. Cloudflare in front, Vercel edge for the app tier, Sanity for content. All three cache aggressively; the buyer's browsing feels like a magazine, not like a booking site.
Rollout & adoption
How it landed inside the organisation.
The rollout started with the AI itinerary surface as an unlisted URL for the client's advisor team, then extended into a soft-launch to the client's newsletter list before the public launch. Advisors used the surface with buyers for three weeks before the public rollout, which meant every UX rough edge was caught by the people who would live with the surface daily. Public launch came with a new brand identity, a re-designed marketing site, and a personalisation layer that had already been trained on advisor-buyer sessions.
Outcomes
The numbers that matter.
3.8x
Increase in advisor-conversation rate
From the AI itinerary surface, compared to the prior static content baseline. Measured over three months post-launch, with a hold-out cohort on the prior surface.
24h
First itinerary surface to production
Via the MVP in 24 Hours engagement, then extended over the following weeks. First real buyer conversation from the surface landed in week two.
1 wk
Brand-system MVP delivery
Design system MVP shipped in the first week of the design engagement — enough to be usable across the itinerary and marketing surfaces while the full system was built out.
Tech stack
What we built it on.
TypeScriptNext.jsSanity CMSOpenAIAnthropicFigmaVercelCloudflare
Reflection
What we learned.
The biggest editorial risk was that AI-composed content would read like AI-composed content. The single decision that mitigated it was involving the client's founder in the voice guide — the tone, the vocabulary, the honesty about tradeoffs. The itinerary surface produces writing the founder is happy to put in front of a buyer because the founder taught it what her voice sounds like. That's not a model choice; it's an editorial choice.
What's next
The engagement today.
The engagement continues in a lower-cadence advisory posture. Roadmap includes additional itinerary categories (business-travel loyalty optimisation, family multi-generational trips), an extension of the advisor tooling into an advisor-first CRM, and a small experiment on paid-media personalisation for high-intent search terms.
“The AI-composed itineraries were the one thing our advisors thought we couldn't do without breaking the brand voice. Headify designed and shipped it in a way our team was proud to put in front of clients.”
Talk to the team
Bring us your travel & ecommerce brief.
Book a 30-minute discovery call. A senior practitioner from the same practice that shipped this engagement will scope yours.
Talk to us
or email contact@headify.in