Kakao has strengthened the place-agent function of its artificial intelligence (AI) service Kanana within KakaoTalk and significantly expanded the scope of search and booking support.

The industry said on Tuesday that Kakao recently expanded Kanana’s existing restaurant-focused place recommendations and booking functions into travel, culture and everyday convenience categories through an update. That allows users to receive recommendations for tourist attractions, accommodation, exhibitions, performances, cinemas, gas stations, convenience stores and auto repair shops based on KakaoTalk chat context, and complete bookings without leaving the page.

It also upgraded customised information provided in connection with schedules or answers. If a schedule to visit a specific area is registered in Kanana, a daily morning schedule briefing delivers information on restaurants, parking and menus for that area. Kakao also added a feature that generates and suggests follow-up questions that users may be curious about when it provides place recommendations.

Kakao plans to apply to KakaoTalk a feature that recommends bundled places across multiple industries. It is also reviewing a feature that recommends accommodation, tourist attractions and restaurants at once based on chat context and links the trip itinerary with KakaoMap to display time-slot visits on a map.

The update is part of implementing "agentic AI" that identifies a user’s intent and makes its own decisions and takes action. The plan is to support the user journey from additional exploration to final execution, rather than ending with a one-off search. Kakao plans to strengthen connections with its own services including KakaoMap and KakaoTalk Booking, while also expanding the related ecosystem by cooperating with external partners.

A Kakao official said the strengthened "place agent" lets users complete recommendations, exploration, sharing and bookings within a single conversation. The official said Kakao plans to gradually expand agentic AI that goes beyond search that simply answers questions by interpreting user intent and deciding and combining needed functions on its own.