Travel platform with AI itinerary builder.
A consumer travel platform with an AI itinerary builder that produces booking-ready plans — wired into real inventory through ERPNext, not a chatbot bolted onto a search bar.
What the founder came to us with.
AI-generated travel plans are easy to demo and hard to actually book. Most consumer experiences in the space stop at a chat reply: a list of suggestions with vague timings, missing prices, and no path from text to a confirmed booking.
The founder needed an itinerary builder that crossed that gap — proposals had to map onto real inventory, real prices, and real availability, with the operations side handled by a system the team could already run a travel business on.
How we built it.
Next.js for the consumer surface, optimised for the few flows that actually matter: prompt the AI, refine the plan, see prices update against live inventory, and confirm with one booking action.
ERPNext on the back end runs the parts a travel business already understands: inventory, vendor management, customer records, refunds, and the long tail of settlements and operational issues. Nothing gets reinvented.
Frappe Framework hosts a custom app that bridges the two: trip itineraries, supplier mapping, and the orchestration layer that turns an AI plan into a sequence of bookable items.
The AI itinerary builder is layered on top with two non-negotiables. Every proposed itinerary is validated against real inventory before it's shown to the user — no hallucinated hotels — and there's a deterministic fallback when the model can't ground a recommendation, so the user is never stuck on a chat reply that won't book.
What shipped, and what it means.
The platform launched with a working AI experience that produces itineraries the user can actually buy. Customers move from prompt to booked plan in fewer steps than a traditional search-and-build flow, and the operations team works in ERPNext rather than a custom admin nobody else can run.
The architecture lets the founder add new supplier integrations, new destinations, and new AI capabilities without touching the consumer experience — separation of concerns that pays back the engineering cost within the first few releases.
Engagement facts.
The shape of this engagement, summarised. Items marked TBC are still being confirmed and will be filled in once the founder has approved disclosure.
What it was built on.
Boring choices for the load-bearing parts; the inventive choices live where they earn their keep.
Pillars and engagements that pair with this one.
If this engagement resonates, these are the next places to look — the underlying service pillars and another engagement that ran a similar play.
Got a similar problem? Tell us where you are.
A 30-minute call. We listen, we ask, we tell you whether the play we ran here would work for you — or whether you need something else first.
Book a discovery call