MMOOZZGG.com

AI assistant

In short

tutu, a Russian travel service, uses an AI voice agent on its incoming line that closes 20–25% of calls with no operators, according to CNews on 23 September 2026. That is 10,000 to 12,000 requests a month from about 50,000 calls. The article explains the pilot stages and why results depend on answers already written in the knowledge base.

Key ideas

  • The tutu contact center gets about 50,000 incoming calls a month.
  • After the first stage, the AI voice agent closes 20–25% of all calls with no operators, CNews reported.
  • The pilot ran from January to June 2026 and covered telephony integration, speech recognition tuning and load balancing.
  • The agent works with confidence where answers are already written in the knowledge base.
  • The result depends on how complete the knowledge base is for each type of call.

Prepared by AI from the article

AI voice travel agent: tutu hands 20–25% of calls to AI

NewsHotels and tourism

Easy2 min read

An AI voice travel agent now answers the incoming line of tutu, a Russian travel service. CNews reported on 23 September 2026 that after the first stage the agent closes 20–25% of all calls with no operators. That is 10,000 to 12,000 requests a month. This article explains what it does and how a hotel or a small travel agency can use the experience.

On this page · 4 sections

What did tutu do?

The tutu contact center gets about 50,000 incoming calls a month. After a purchase, travelers ask about flight status, baggage rules, and refund and exchange rules.

A voice agent is an AI agent that holds a phone conversation. In the first seconds it finds the topic of the call. If the question belongs to set scenarios, such as a certificate, a timetable or a flight status, the agent advises on its own from the company knowledge base. In unusual cases it moves the call to the right team with the history of the dialog.

What are the stages and numbers?

The pilot ran from January to June 2026: integration with telephony, tuning of speech recognition and load balancing. Speech recognition is a program that turns spoken words into text. Now the agent works in production.

After the first stage, the agent fully closes 40% of requests on topics covered by the knowledge base. Across the whole line, this is 20–25% of calls. It freed a resource of 5 FTE, that is five full-time employees, and they moved to hard cases.

The next step is to connect the agent to accounting and booking systems, so that exchanges, reissues and confirmations of changes are done during the call. The voice line is part of a wider AI strategy. In June 2026, tutu opened search and booking of flights, hotels, trains and buses to AI agents through MCP (Model Context Protocol, an open standard that links AI apps to tools and data), Kommersant wrote.

What does it mean for tourism?

The gap between 40% and 20–25% shows what the result depends on: the agent works with confidence where the answers are already written in the knowledge base. For a hotel or a travel agency the conclusion is direct. First write the rules for check-in, transfer, payment and refund, then launch the agent.

The second conclusion is about order. tutu first taught the agent to answer and pass on hard cases, and left operations with an order for a later stage. A small company can go the same way: start with information and taking requests, not with canceling a booking.

What should a hotel or agency check now?

  1. List 20 frequent questions of guests or tourists, and check that each has a written answer: check-in time, early check-in, transfer, refund on cancellation.
  2. Count how many calls arrive after 7 p.m. and at weekends.
  3. Decide what the agent never does: it does not change a price, does not promise a refund and does not cancel a booking.
Contact us