Live Nov 18 · Cohort 3Nov 18, 19:00 MSK — free hands-on AI agents live workshop · Live Cohort 3 starts Nov 23Join Nov 18 Live →Cohort 3 (Nov 23)
ReleaseAPI / Product · ActionableRB.RU→ original

Yandex Launches "Act!" AI Agent in Alice That Makes Calls and Books Services on Its Own

Yandex has launched the "Act!" agent in Alice AI. It is a mode in which the neural network does not just answer a question but takes a multi-step assignment and carries it through to a result: it places phone calls, books services, browses websites and extracts data from documents. For now, the feature will go to a few thousand active users with an Alice Plus subscription.

AI-processed from RB.RU; edited by Hamidun News
Yandex Launches "Act!" AI Agent in Alice That Makes Calls and Books Services on Its Own
Source: RB.RU. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceRB.RU ↗
Editor-in-Chief
Key metric #1Can make phone calls and book services without human involvement
Key metric #2Works with websites and gathers information from documents
Access at launcha few thousand Alice Plus subscribers

Chatbots learned to answer nicely a long time ago. But when you need to book a service appointment for Wednesday, find a table for four and check the price in a contract at the same time, a person still ends up sitting with ten browser tabs open and a phone in hand. Yandex is trying to remove this routine: Alice AI now has a personal agent called "Act!" that takes on multi-step assignments.

What exactly "Act!" can do

According to Yandex's press office, as relayed by RB.RU, the agent can carry out assignments consisting of several steps, make phone calls, book services, work with websites and gather information from documents. The key word here is "multi-step." An ordinary assistant answers a single request. An agent breaks a task into steps, decides for itself what to do next, and comes back to the user with a result.

A simple example: you ask it to find a barbershop near your home and book a slot for Saturday. A chatbot will give you a list of addresses. An agent in this setup should open the websites, check the schedule, and if there is no online booking, call and arrange it by voice. The call is the most unusual part: many small businesses still take appointments by phone, and without a voice channel the agent hits a wall.

Why this matters to developers and product managers

Three signals stand out here.

First: competition is shifting from the question "whose model is smarter" to "whose model gets the job done." Agent features are ceasing to be conference demos and are becoming part of a mass-market product, with access tied to the paid Alice Plus subscription.

Second: websites and services get a new type of visitor, the AI agent. If your booking flow, order form or pricing page is built so that a human can figure it out but a program cannot, some customers will be lost. Clear forms, up-to-date schedules and understandable contact details become not only a matter of convenience but a condition for an agent being able to handle you at all.

Third: the phone remains a full-fledged interface. If your business answers calls, be prepared for a synthetic voice on the other end. It is worth deciding in advance how you feel about that and how you verify that a booking is actually confirmed.

Limitations and what comes next

The launch is cautious: at the first stage, only a few thousand active chat users with a subscription will get the feature. This is standard practice for agents, because errors in actions cost more than errors in text. A wrong answer can be ignored, but a wrong booking has to be canceled, sometimes at a financial loss.

The published announcement leaves several important questions unanswered. How does the agent ask for confirmation before acting? What happens if the chain of steps breaks in the middle? How does it introduce itself to the other party during a call? What data from documents is sent for processing and how is it stored? The maturity of the product should be judged by these, not by polished demos.

Practical takeaway. If you build a product with appointments or bookings, check it through an agent's eyes: can the flow be completed without a CAPTCHA and hidden steps, are the schedules up to date, is a working phone number listed. If you are building your own agent, take the main lesson from Yandex's launch: start with a narrow group of users, limit the range of tasks, and build in human confirmation before irreversible actions. In a few months it will become clear how well "Act!" copes outside ideal scenarios.

Story

Key Questions Answered

What exactly "Act!" can do?

According to Yandex's press office, as relayed by RB.RU, the agent can carry out assignments consisting of several steps, make phone calls, book services, work with websites and gather information from documents. The key word here is "multi-step." An ordinary assistant answers a single request. An agent breaks a task into steps, decides for itself what to do next, and comes back to the user with a result. A simple example: you ask it to find a barbershop near your home and book a slot for Saturday.

Why this matters to developers and product managers?

Three signals stand out here. First: competition is shifting from the question "whose model is smarter" to "whose model gets the job done." Agent features are ceasing to be conference demos and are becoming part of a mass-market product, with access tied to the paid Alice Plus subscription. Second: websites and services get a new type of visitor, the AI agent. If your booking flow, order form or pricing page is built so that a human can figure it out but a program cannot, some customers will be lost.

What are the key details on Limitations and what comes next?

The launch is cautious: at the first stage, only a few thousand active chat users with a subscription will get the feature. This is standard practice for agents, because errors in actions cost more than errors in text. A wrong answer can be ignored, but a wrong booking has to be canceled, sometimes at a financial loss. The published announcement leaves several important questions unanswered. How does the agent ask for confirmation before acting? What happens if the chain of steps breaks in the middle?

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Private AI Builders Community · Vibe Club

Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews

Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.

Join Vibe Club →Explore Hamidun Academy
What do you think?
Loading comments…