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 · Actionable3DNews AI→ original

"Alisa AI" Learns to Make Calls on Behalf of Users: Yandex Tests a Voice Agent

Yandex has begun testing outgoing calls for "Alisa AI": the user sets a task, and the assistant dials an organization or a private individual and talks in the user's place. The recipient sees the label "Alisa, request from a user" on the screen, so there is no impersonation. This is a step from a chatbot that answers to an agent that sees a task through to a result in the real world, where there is no API.

AI-processed from 3DNews AI; edited by Hamidun News
"Alisa AI" Learns to Make Calls on Behalf of Users: Yandex Tests a Voice Agent
Source: 3DNews AI. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary source3DNews AI ↗
Editor-in-Chief
Key metric #1Calls can be made to both companies and private individuals on the user's behalf
Key metric #2The recipient sees the label "Alisa, request from a user"
For now this is a testlaunch dates and terms have not been disclosed

Most AI assistants write texts very well and search for information reasonably well, but they hit a wall wherever a live person is needed on the other end. Booking a doctor's appointment, checking whether a table is available, finding out from a car service how much an oil change costs: half of such tasks are still solved by a phone call, not a button in an app. Yandex decided to break through this wall head-on: "Alisa AI" is now testing a feature in which it calls organizations and private individuals on its own on behalf of the user.

What Is Actually Known

There are few facts so far. According to Vedomosti, citing a company representative, Yandex has launched testing of the new feature. The assistant makes calls to organizations and to ordinary people. The recipient's phone screen shows the label "Alisa, request from a user". Everything else was not disclosed in the publication: there is no information on who will get access first, how much it will cost, which scenarios are allowed, and how the assistant reports on the outcome of the conversation. Any details beyond this remain guesswork and should not be built into plans.

Why This Matters More Than It Seems

The approach here is a fundamental one. Usually the "agency" of AI is built through integrations: a service has an API, the assistant calls it and gets a response. But most hair salons, clinics, small shops and simply acquaintances have no API and never will. Yet everyone has a phone. The call becomes a universal interface to a world that has not been digitized: the agent does not need to negotiate an integration with every business, a phone number is enough.

The idea itself is not new. Voice assistants that call on a person's behalf have been shown before, and worldwide this is a long-discussed class of products. What is new is something else: this is a mass-market Russian assistant with a large audience, and Yandex is bringing such a feature into a product, not into a conference demo.

The label on the recipient's screen deserves separate mention. It is a product decision and an ethical one at the same time. If the AI called silently, without disclosing itself, it would quickly turn into a tool for spam calling and fraud, and people's distrust of any unfamiliar call would grow. A label worded "request from a user" shows that a specific person stands behind the call, not a mass mailing.

What This Means for Developers and Product Managers

First: voice as a channel for agents is returning to the agenda. If you run a business service with phone booking, prepare for assistants to start calling you. This means that scripts, IVR systems and operators must be able to hold a dialogue with a machine too: answering clearly, without long menus, with confirmation of data.

Second: if you are building your own agents, look at this case as an example of interface choice. When there is no integration, the cheapest path sometimes runs through an existing human channel rather than through endlessly writing connectors.

Third: limitations and risks must be built in from the start. A call is an irreversible action. If the assistant misunderstands the task and books you for the wrong date, you will be the one fixing it. Hence the questions worth asking of any voice agent: how the task is confirmed before the call, what happens when the other party responds unexpectedly, who is responsible for an error, and how call recordings are stored.

What Is Still Missing

There is no quality data: how many calls end successfully, how the assistant behaves when the other party interrupts, goes off topic or simply hangs up. There is no information on the legal framework: consent to processing the voice and personal data of the other party, storage of recordings. And it is unclear whether recipients will be able to refuse to talk to an AI.

For now this is a test, and it should be taken as a signal of direction, not as a finished product. But the direction is clear: assistants stop only answering and start acting in a world where most tasks still have no button, but do have a phone number.

Story

Key Questions Answered

What Is Actually Known?

There are few facts so far. According to Vedomosti, citing a company representative, Yandex has launched testing of the new feature. The assistant makes calls to organizations and to ordinary people. The recipient's phone screen shows the label "Alisa, request from a user". Everything else was not disclosed in the publication: there is no information on who will get access first, how much it will cost, which scenarios are allowed, and how the assistant reports on the outcome of the conversation.

Why This Matters More Than It Seems?

The approach here is a fundamental one. Usually the "agency" of AI is built through integrations: a service has an API, the assistant calls it and gets a response. But most hair salons, clinics, small shops and simply acquaintances have no API and never will. Yet everyone has a phone. The call becomes a universal interface to a world that has not been digitized: the agent does not need to negotiate an integration with every business, a phone number is enough. The idea itself is not new.

What This Means for Developers and Product Managers?

First: voice as a channel for agents is returning to the agenda. If you run a business service with phone booking, prepare for assistants to start calling you. This means that scripts, IVR systems and operators must be able to hold a dialogue with a machine too: answering clearly, without long menus, with confirmation of data. Second: if you are building your own agents, look at this case as an example of interface choice.

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…