According to McKinsey, fewer than a third of travellers have used generative AI for travel-related tasks, and 84% of those who have report that it improved their experience. South African figures follow the same shape. Under a quarter have used AI to plan a trip, and almost all of them were satisfied with the result.

What travellers use it for suggests where it goes next. General research tops the list at 54%, followed by inspiration and local food recommendations at 43% each, transport planning at 41%, itinerary building at 37% and budgeting at 31%.

Those habits form part of a broader change in how buying decisions get made, described as agentic commerce. Traditional booking journeys run through defined channels, with the traveller moving from a search engine to a comparison site to a booking page. In an AI-led journey, the questions, the narrowing down and the choice all happen in conversation.

"Agentic commerce is fundamentally changing how people discover and choose brands. As journeys become more conversational and intent-driven, companies have to rethink how they show up, not just to be found, but to be chosen, and to deliver value in that moment," says Ndidi Oteh, CEO of Accenture Song.

Search data shows the shift underway. Phocuswright found that general search engines fell from 51% to 36% as travellers' most-used research resource inside eighteen months, while use of platforms such as ChatGPT for trip planning nearly tripled. Sixty-one percent of travel executives are experimenting with or scaling agentic AI.

Most hotels reach those conversations indirectly, through an online travel agency's listing or a subscription service that places their inventory in front of the AI assistant. A smaller number have connected their own systems. Radisson Hotel Group took that route, building an app in ChatGPT with Accenture that queries its reservation system across more than 1,000 hotels in over 100 countries.

In use, it works like any other ChatGPT conversation. The traveller opens a chat, tags the hotel group's app at the start of the prompt and describes the trip in ordinary language rather than as filter settings. A family weekend in Umhlanga. A hotel in Cape Town with a gym and a pool within walking distance of the V&A Waterfront. What comes back is a set of properties matching the description, with current rates and availability, amenities, location context and results plotted on a map. Follow-up questions narrow the list without starting again. Choosing a property hands the traveller to the hotel group's own booking page to finish the reservation.

Getting that far takes more than a chatbot. A hotel's rates, availability and property information sit in systems designed to serve its own website and its distribution partners. An external AI system cannot read them without help, so it falls back on scraped summaries and whatever was in its training data, which is how a traveller ends up with a recommendation for a property that has no rooms left. Model Context Protocol is the open specification written to close that gap, letting AI applications query live systems directly.

"With travel, planning is becoming easier for consumers, but more competitive for brands," shares Emily Weiss, senior managing director and global travel lead at Accenture. "Travellers want to compare options quickly, narrow down choices and book with confidence, often in a single journey. That puts pressure on brands to show up with accurate, up-to-date, and easy-to-navigate information at every step."

Conversational search also changes what happens when a traveller compares prices. Travellers have always carried the work of comparing rates across booking sites, and best-rate guarantees added a further task, since claiming a lower price meant completing a form and attaching a screenshot. Automated matching removes both steps. The system detects lower publicly available rates for a group's properties on third-party booking platforms and matches them on the hotel's own site without any action from the traveller. At Radisson, the feature is live across all properties worldwide, covering rates from Booking.com, Expedia, Hotels.com, Agoda, Priceline, Trip.com and MakeMyTrip, among others.

"Today's travellers expect speed, simplicity and confidence when they book," says Gianni Di Fede, global chief commercial officer at Radisson Hotel Group. "Price matching is our commitment to ensuring that guests who choose to book directly with us never have to wonder whether they could have found a better deal elsewhere. By automating the process in real time, we are making direct booking simpler and more transparent while delivering greater value to both guests and hotel partners."

"AI is radically transforming how people search for and book hotel stays, and we are committed to being at the forefront of this shift," he adds. "With Accenture, we are reimagining hotel discovery for the next generation of travellers, meeting guests in the planning moment with a branded experience that makes it easier to find, compare, and book Radisson Hotels properties."

Discovery moves to the traveller, and the booking itself completes in the hotel group's own environment. That division suits how people say they want to buy. Seventy-seven percent of South African consumers would not yet hand a purchase to an AI agent, and travellers who plan a stay in conversation can finish it somewhere they already know, with the rate the assistant quoted them. Reservation management, trip modification, loyalty recognition and concierge functions are being added to the conversational side as they mature.