If your team handles bookings, vendor follow-ups, or service coordination, the hard part often isn’t finding information. It’s getting the last mile done when the only path is a phone call to a real person.
That’s where many AI workflows still stall. They can draft, summarize, and suggest next steps, but they stop short when the work leaves the screen and turns into a live conversation.
Google brings Duplex calling into Gemini Tasks
Google Research says Duplex is now integrated into Tasks in the Gemini app, letting the assistant place calls to businesses on a user’s behalf for real-world errands such as reservations and appointments. The launch matters because it extends Gemini from on-screen help into live task completion where no API or booking form exists.
Google also says the calling stack now runs on a Gemini-based architecture and can handle broader, free-text requests rather than only tightly defined verticals. In the same workflow, Google says Tasks in Gemini can coordinate over email and RCS and support secure transaction processing through its Agent Payments Protocol.
- Phone-based task execution. Google says Tasks in the Gemini app can now call businesses for users and handle requests such as reservations and appointments.
- Broader task coverage. Google says the system is no longer limited to narrow domains and can work from free-text task descriptions across more business categories.
- Unified Gemini stack. Google says Duplex’s calling system has moved from older task-specific pipelines to a unified architecture built with Gemini models.
- Multi-step coordination. Google says Tasks in Gemini can combine calling with other actions such as email, RCS coordination, utility bill analysis, and secure transaction handling through its Agent Payments Protocol.
The bigger shift is from assistance to completion
For business owners, this is a useful reminder that a lot of work still sits outside clean software systems. Suppliers don’t always have portals, local partners don’t expose APIs, and routine coordination often depends on whoever has time to call, wait, clarify, and follow up.
That means the winning AI workflows won’t just generate answers. They’ll need to carry context across channels, keep a record of what happened, and fit into a process a team can review and improve.
Even when another vendor handles the actual phone call, the business problem around it is operational. Someone still needs the request captured cleanly, the constraints attached, the follow-up task assigned, and the outcome visible to the rest of the team.
That’s the gap many companies should focus on now: not whether an AI can sound human on a call, but whether the surrounding workflow is organized enough for AI to do useful work without creating confusion.
How Metomorph helps teams operationalize this pattern
Metomorph is useful on the workflow side of that shift. It gives teams a place to hold the task, the context, and the follow-through around messy real-world work, instead of leaving each request buried in chat threads or someone’s memory.
Imagine your team manages office operations for multiple locations. A manager needs updated pricing from three local repair vendors, plus the earliest available service window that won’t conflict with site access rules.
In Metomorph, that request could live inside a project with the relevant notes, documents, and constraints already attached through project context. The task board can track each vendor outreach step, who owns it, and what is blocked or waiting.
Once replies or call notes come back, the same project can hold the running record of options so the next person doesn’t have to reconstruct the situation. That is especially helpful when work moves across shifts or departments.
If your team wants a consistent intake process for these requests, Metomorph can also standardize it with custom forms and route the result into the same shared workspace. The practical advantage is simple: fewer dropped details, clearer ownership, and a better chance that AI-supported work actually fits how your team operates.
What to do next
- List the tasks in your business that still require calls, texts, or manual follow-up. Those are the best candidates for your next AI workflow review.
- Separate the conversation layer from the process layer. Even if another tool handles outreach, make sure your team has one place for constraints, status, and outcomes.
- Standardize request intake before you automate anything. A short form or structured task template will do more for reliability than a clever prompt alone.
- Pick one recurring coordination job and map it into a shared project with task tracking and context attached. That’s a concrete first step toward making real-world AI work manageable.
Explore Project context, Task board, Custom forms in Metomorph.
Source: Google Research, How Google Duplex evolved to also call businesses for you in Gemini (November 20, 2026).