Guozhen AIGlobal AI field notes and model intelligence

Realtime AI News

Google explains how Duplex moved into Gemini to call businesses for you

A Google Research blog post traces how Duplex's calling technology evolved into a capability inside Gemini that can phone businesses on a user's behalf. The piece frames the shift as voice agents moving from standalone assistant features into a general model-driven assistant.

Published
Google 详解 Duplex 如何进入 Gemini:让助手代你给商家打电话
Image source: research.google

Google Research has published a post tracing how Duplex's calling capability evolved and eventually landed inside Gemini, where the assistant can phone businesses on a user's behalf. The headline points directly at that path: the feature now also calls businesses for you in Gemini.

Duplex first reached public attention through narrow tasks such as booking a restaurant table by phone, which required speech synthesis, speech recognition and dialogue decisions to work together within seconds. The hard part was never sounding human; it was completing a task over a noisy line with an unprepared person on the other end.

The central message of the post is that this capability has changed position: from a standalone voice feature to something Gemini can invoke. That matches the broader direction of assistant products, where voice, calling and scheduling are consolidated behind a single entry point and a model decides which tool to use when.

Calling a business is a useful test case for agents. The goal is concrete, such as checking opening hours, confirming a booking or cancelling an order, the cost of failure is bounded, and the result can be verified immediately. At the same time it demands open-ended conversation, tolerance for accents and phrasing, and follow-up questions when information is missing.

The hardest moments tend to sit on the exception paths: a business that will not cooperate, details that do not line up, a request to speak to a human. Templates cannot cover those cases, and a model's maturity shows more in knowing when to stop and hand the decision back to the user than in completing the standard script.

The post is a retrospective product narrative rather than an announcement of new availability or timelines, so the signal lies in direction: voice agents are moving from being able to make calls toward being naturally invoked inside an assistant. For users, that means handing over a request instead of opening a separate feature.

Watch three things next: how widely the calling capability is available inside Gemini and under what limits, how Google handles disclosure and transparency during automated calls, and how the experience is managed on the business side. Those will decide whether this stays a demo-grade convenience or a durable part of daily use.

Why it matters

Turning business calls from a standalone trick into an assistant capability shows the voice-agent race moving from demo quality to usability and transparency.

GoogleGeminiAI Agent
Back to realtime news

Nearby Updates

All