'ChatGPT'... From Answering Questions to Executing Tasks
At the end of August, OpenAI pushed ChatGPT a step further toward transforming from a question-answering assistant into a tool capable of executing tasks and interacting with user-related websites and applications, as part of a series of rapid updates that expanded its capabilities in automation, work, and content creation.
The latest of these updates arrived on August 31, when ChatGPT Workspace and Codex gained the ability to use tools provided by certain supported websites directly from the browser embedded in the desktop app. This allows the system to discover available tools on a page and use them to complete tasks, while maintaining access controls and confirmation prompts for sensitive actions.
On the same day, OpenAI enabled the creation of custom sticker packs from an idea or image, which can be shared via supported messaging apps. It also expanded browser extension support to include Microsoft Edge, Brave, Opera, and Vivaldi.
On August 28, the company added the ability to link multiple Google accounts to ChatGPT, including Gmail, Google Calendar, and Google Contacts. This allows users to search across multiple inboxes or review schedules from different accounts within the same conversation. The feature is available globally for supported plans on the web, desktop, and mobile apps.
Tasks Move When Changes Occur
The scheduled tasks system saw a significant update on August 25, as tasks in ChatGPT Workspace gained the ability to automatically respond to changes in supported apps, rather than relying solely on user-defined time triggers.
These events include the arrival of a new message in Gmail, a post in a Slack channel, or activity on a GitHub pull request. This allows users to instruct ChatGPT, for example, to draft a summary upon receiving a specific message or to formulate next steps after a new comment arrives.
Actions requiring user approval remain pending until reviewed. The company also enabled the sharing of scheduled tasks, allowing another user to review instructions, modify them, link their own apps, and create an independent version.
Tasks are not limited to new events; they can also be used for reminders, repetitive work, periodic updates, and monitoring specific changes, with capabilities and usage limits varying by plan and task type.
ChatGPT Handles Apple Messages
On August 20, OpenAI added the ability to use the Apple Messages app through ChatGPT Workspace and Codex on Mac devices equipped with Apple Silicon.
The Apple Messages extension can read, search, compose, and send iMessage, SMS, and RCS messages via the Messages app. By default, ChatGPT requests user approval for the message content and recipients before sending.
Images with Transparent Backgrounds
On the same day, OpenAI announced experimental support for transparent backgrounds in the GPT-4o image model via the API, enabling the creation of reusable graphical elements over various backgrounds.
The feature targets use cases such as product images, graphic designs, website mockups, and marketing campaigns, eliminating the need to remove image backgrounds in a separate step after generation.
Together, these updates reveal a clear direction for OpenAI to expand ChatGPT’s role from answering questions to executing, following up on, and interacting with applications and websites, within the permissions granted by the user.
Thus, the next phase of the AI competition appears to be less about the sheer volume of information an assistant can provide, and more about its ability to execute and follow through on tasks on the user’s behalf.
