On August 28 and 29, 2026, OpenAI shipped two updates that individually look like feature additions but together signal a structural shift: ChatGPT can now operate inside Apple Messages on Mac and across multiple Gmail, Google Calendar, and Google Contacts accounts in a single conversation. For enterprise teams, these updates matter less as convenience upgrades and more as evidence that AI is leaving the chat window and becoming part of the operating layer where knowledge work actually happens.
Here is what changed, how it works, and what it means for your team.
What OpenAI Shipped This Week
Multiple Google accounts (August 28, 2026). Until now, ChatGPT users who connected Gmail, Google Calendar, or Google Contacts could only link one Google account per connector. A seller managing a personal inbox and a company inbox had to run two separate sessions or pick one. OpenAI confirmed on August 28 that Plus, Pro, Business, and Enterprise users can now link more than one Google account per connector. Adding a second account is a four-step process: open Plugins in the sidebar, find the connected app, click the gear icon, and add the account through Google’s standard authorization flow. The change is live on web, desktop, iOS, and Android. Workspace admins retain the ability to restrict it for managed accounts, according to OpenAI’s release notes.
Apple Messages integration (August 29, 2026). ChatGPT’s Mac desktop app now connects to the Apple Messages application. Once enabled with Full Disk Access permissions on macOS, ChatGPT can search message history, read conversations across iMessage, SMS, and RCS, draft replies based on thread context, and send messages with explicit user approval. The integration works inside ChatGPT Work and Codex, which means it is available in the same environment enterprise teams use for background agent tasks and software development. OpenAI confirmed that the extension processes messages locally on the device and does not index or upload message content to its servers, except where a user explicitly saves a conversation to the cloud. Every send action requires the user to confirm the message content and the recipient list. Continuous approval mode, which removes that confirmation step, is available but marked as a setting that reduces your final review opportunity, per OpenAI’s own guidance.
Scheduled tasks for free users. On August 25, OpenAI extended scheduled task functionality to Free plan users, who can now set up to five active tasks that run once or once per day. Paid plans carry higher task counts and can trigger actions based on connected-app events rather than a fixed clock. This matters for the Apple Messages and Gmail integrations: a scheduled task can monitor a connected Gmail account for a specific type of email and summarize it on a recurring schedule, without the user opening ChatGPT at all.
The Pattern: Ambient Presence, Not Chat Expansion
Taken together, these updates are not incremental improvements to a chat product. They represent ChatGPT following the same trajectory that productivity software has always followed: from a dedicated interface users visit to an ambient layer that attaches to the surfaces where work already happens.
This pattern has a direct precedent. Microsoft embedded Copilot into Outlook, Teams, and Word not because chat was broken, but because the switch cost of leaving an email to open an AI tool was high enough that most people did not bother. When the AI lives inside the inbox, usage compounds differently. OpenAI is building toward the same outcome: ChatGPT as the AI layer inside iMessage, Gmail, and eventually more native OS surfaces.
For enterprise teams already using ChatGPT Work, the practical implication is direct. Agents can now receive context from messages and calendar events, not just from documents and prompts typed into ChatGPT. A background Codex task that checks a Gmail inbox for client replies and updates a project summary no longer requires a human to copy and paste between apps. A sales rep can ask ChatGPT to find the last three iMessages from a prospect, draft a follow-up, and queue it for approval, all without switching windows.
What This Means for GTM and Enterprise Operations Teams
The teams that benefit most from these updates are the ones with the highest messaging volume and the most fragmented context: sales, customer success, executive support, and technical account management.
| Feature | Plans Available | What It Unlocks |
|---|---|---|
| Multi-account Gmail | Plus, Pro, Business, Enterprise | Cross-inbox search, unified scheduling across two calendars, CRM update drafts from email threads |
| Google Contacts connector | Plus, Pro, Business, Enterprise | Contact lookup, auto-fill from org directory |
| Apple Messages integration | ChatGPT Work, Codex | iMessage search, reply drafting, send with approval |
| Scheduled tasks | All plans (limits vary) | Recurring inbox summaries, event-triggered actions, monitoring without manual check-ins |
GTM teams running account-based motions often track prospects across a personal domain and a company domain. The single-account cap on Gmail connectors was a genuine friction point: a rep managing relationships across two email identities had no clean way to pull context into a single thread. That limit is now gone for paid plan users.
Enterprise operations teams benefit from the combination of scheduled tasks and multi-account Gmail. A weekly revenue summary that pulls data from two email accounts and a shared calendar is now achievable without a custom integration or a dedicated automation platform. The same workflow that previously required a Zapier setup or a developer building a script can run directly inside ChatGPT Work.
Privacy and Security Considerations
Enterprise security teams will have two questions about the Apple Messages integration: where does the data go, and who controls what the AI can send.
On data residency: OpenAI’s stated design keeps message processing on-device. The extension reads from the local Messages database on macOS. It does not build a stored index of message history in OpenAI’s infrastructure. This is a different model from cloud-based integrations like the Gmail connector, where content is fetched at query time from Google’s servers. The practical difference: Apple Messages content stays under macOS access controls, not under OpenAI’s API security model.
On send authorization: the default configuration requires explicit user confirmation before ChatGPT sends any message. The two-step confirmation (content and recipients) means a rogue agent or accidental approval does not result in a message being sent. Workspace admins in managed ChatGPT deployments retain the ability to restrict the integration entirely before users can enable it.
The multi-account Gmail connector raises its own question: can a user connected to both a personal and a company Gmail account accidentally expose data from one context in the other? OpenAI has not published a detailed isolation specification for multi-account connectors. Enterprise admins in regulated industries should treat multi-account Gmail with the same caution as any tool that touches personal and corporate data in the same session, until OpenAI publishes explicit account-isolation documentation.
What to Watch Next
The Apple Messages integration is Mac-only at launch. A future iOS or iPadOS version would extend the same capability to mobile, which is where a significant portion of enterprise messaging traffic actually originates, particularly in customer-facing roles where reps stay off a laptop for most of the day.
The scheduled tasks expansion to free users matters beyond the consumer market. Many enterprise organizations have employees on personal ChatGPT accounts who are experimenting with automations before IT approves a formal enterprise rollout. As those experiments work, the business case for an enterprise contract compounds. OpenAI’s strategy of giving free users a functional but limited version of scheduling is a customer-development motion as much as it is a product decision.
Finally, the combination of ChatGPT Work and native OS integrations brings ChatGPT closer to the function that enterprise AI workspace products like Slack Code and Claude Cowork are trying to occupy. The difference is surface area: ChatGPT is embedding into the communication layer at the OS level, while workspace AI tools are embedding into the collaboration layer at the application level. The two approaches are converging, and 2026 is likely the year that distinction starts to matter to IT buyers choosing between platforms.
For teams building AI-native GTM and operations systems, the immediate action is the same regardless of platform: map where your real workflow context lives, and connect AI to those surfaces first. A ChatGPT that can read your Gmail and draft your iMessage reply is useful. An AI system designed around your actual revenue and operations data, connected to your communication channels, and governed by your security policies, is a competitive advantage. If you want to think through what that looks like for your organization, book a call with the Enera team.
Sources: OpenAI release notes August 28, 2026, Startup Fortune: OpenAI multi-Gmail accounts, India TV News: Apple Messages integration, Daily Beirut: OpenAI Apple Messages