# OpenAI brings ChatGPT Voice into Work, with access to connected apps

> Source: <https://runtimewire.com/article/openai-chatgpt-voice-work-connected-apps>
> Published: 2026-09-23 21:09:38+00:00

# OpenAI brings ChatGPT Voice into Work, with access to connected apps

**The September 23rd update pairs voice with GPT-6 models and workplace tools, extending ChatGPT Work to web and mobile.**

        By [Ryan Merket](https://runtimewire.com/author/ryan-merket)
        · Published 

Primary source: [X](https://x.com/OpenAI/status/2102808325742322002)

## Why it matters

Voice can now serve as an entry point to ChatGPT Work's connected apps and task workflows, putting more weight on model reliability, account permissions and user approval.

OpenAI is bringing ChatGPT Voice into ChatGPT Work on web and mobile, where users can speak to the assistant while it uses connected email, calendar and Slack tools. In a [September 23rd post on X](https://x.com/OpenAI/status/2102808325742322002), OpenAI said Voice can use those plugins, run on [GPT-6 Astra](https://runtimewire.com/models/openai/gpt-6-astra:batch), Sol and Luna, and handle Work tasks such as creating documents, presentations, websites and spreadsheets.

[https://x.com/OpenAI/status/2102808325742322002](https://x.com/OpenAI/status/2102808325742322002)

The move joins three parts of ChatGPT that had been available in separate combinations: spoken conversation, connected work data and Work's multistep task environment. Users can now give a spoken instruction in Work and have ChatGPT draw on tools such as email, calendars and Slack while producing a work output. OpenAI's [ChatGPT Work guidance](https://help.openai.com/en/articles/20001275-chatgpt-work-and-codex) describes Work as an agent for longer assignments and finished deliverables, with access governed by the tools and permissions available in the selected experience.

The model lineup is also recent. OpenAI's [release notes](https://help.openai.com/en/articles/6825453-chatgpt-release-notes) introduced [GPT-6 Sol](https://runtimewire.com/models/openai/gpt-6-sol) and Luna in Work and Codex on September 22nd. GPT-6 Astra, which OpenAI announced on September 3rd, is positioned for complex work involving computer use, coding and multistep tasks. OpenAI's post puts all three models inside the Voice experience, though the models users can access depend on their plan and workspace settings.

OpenAI is pushing Voice past the familiar use of dictation and spoken answers. A spoken request to summarize a Slack discussion or turn calendar and email context into a document could start a task that previously required a typed prompt and a switch into Work. That reduces friction at the point of giving instructions; it does not remove the need to check what the assistant produces or which connected data it uses.

Access will vary by plan. TechCrunch reported on the rollout that Plus and Pro users can use Work on mobile for tasks such as creating documents, drafting email and summarizing Slack messages, while Free and Go users receive access to plugins and connected apps. The same report described voice-driven access to Work as the new part of the release. OpenAI's post itself does not specify the plan-by-plan rollout.

The permission layer remains important because a voice command can lead to actions in accounts that contain private or company information. OpenAI's Work instructions say tasks can pause for user input, sign-in or confirmation. Workspace administrators also control which apps, data and actions are available. Connecting Voice to those systems makes hands-free task initiation more practical, while leaving access rules and approvals as constraints on what the assistant can do.

OpenAI has not described in its post how the experience handles each app's individual permissions, which tasks require confirmation, or whether all three models are available across every supported plan. The announcement establishes the product direction: Voice can now initiate work that draws on connected services, not just answer questions aloud. The practical test will be how reliably the assistant moves from spoken instruction to a usable document or completed task without exceeding the user's intent or workplace permissions.
