Google is integrating Duplex's calling capabilities into Gemini: users can have AI contact businesses on their behalf, gather information, confirm prices, and even book services. AI used to be primarily responsible for answering questions; now it is beginning to complete a real communication process for people.
What This Is
Duplex is Google's voice-conversation system, launched two years ago to initially simulate human voices and make restaurant and hotel reservations. Gemini is Google's next-generation large language model, which excels at understanding complex instructions and calling external tools. The core of this upgrade is to make Gemini do more than offer advice: after receiving authorization, it can place calls, communicate with businesses, and advance a task to completion.
Capabilities like this are generally known as "agents"—AI that can understand goals, break tasks into steps, and invoke tools to execute them. But real-world deployment still depends on cooperation on the other end of the call, the scope of the user's authorization, and whether businesses are willing to accept AI agents.
Industry View
What deserves attention is that competition among AI products is shifting from "how intelligently does it answer?" to "can it get the job done?" For ordinary businesses, this also means phone reception, customer follow-ups, and appointment confirmations may gradually be handed over to AI, reducing operating costs and extending service hours.
But the risks are equally clear: AI calls may expose a user's identity, misunderstand requirements, or make commitments beyond the authorized scope. If a pricing dispute or booking error occurs, it remains unclear whether responsibility lies with the user, Google, or the business. Moreover, if businesses broadly adopt procedures to resist AI agents, the efficiency advantage of AI calling will be diminished. What Google has demonstrated is the boundary of what the technology can do, not a mature commercial closed loop.
Impact on Regular People
For enterprise IT: Businesses can connect AI to customer service, sales, and appointment workflows, but they must first define authorization, auditing, and error-correction mechanisms to prevent agents from overstepping on pricing, refunds, contracts, and other sensitive matters.
For individual professionals: Administrative, customer service, and sales roles will increasingly use AI to handle calls and information confirmations. As repetitive communication declines, the value of judging customer intent and navigating complex conversations will rise.
For the consumer market: Consumers may obtain quotes and book services faster, but they should verify whether AI has confirmed key conditions and retain call records, order confirmations, or access to a human customer service representative.