OpenAI is expanding the writing capabilities of ChatGPT Work, allowing the assistant to draw on connected workplace information to produce writing that more closely reflects an individual’s established communication style.
Rather than simply generating polished text in a generic AI voice, the idea is to help ChatGPT understand how a person typically communicates—including the tone, wording, formatting, and stylistic preferences that appear across their existing work.
ChatGPT Work Can Learn From Your Existing Work Context
When supported workplace apps are connected, ChatGPT Work can access relevant information from sources such as Gmail, Google Drive, Slack, SharePoint, and other workplace systems, subject to the permissions and settings configured for the workspace. OpenAI describes Work as an agent designed to work across connected apps and files to complete longer, more involved tasks.
That context can be particularly useful for writing.
For example, your previous communications may reveal that you:
- Prefer short or highly detailed sentences.
- Regularly use certain phrases or expressions.
- Follow a particular structure when writing emails.
- Have a distinctive way of opening or closing messages.
- Use specific capitalization, punctuation, or formatting habits.
- Prefer a formal, conversational, technical, or concise tone.
- Follow particular terminology or language used within your organization.
Instead of asking ChatGPT to “make this sound professional” every time, connected work context can give it more information about how you actually communicate.
OpenAI’s own examples for connected apps include asking ChatGPT to describe and suggest improvements to an individual’s email style in Gmail, as well as drafting communications using information from connected workplace sources.
From Generic AI Writing to a More Personal Voice
One of the biggest challenges with AI-generated writing is that it can sometimes sound noticeably different from the person who is supposed to have written it.
A message may be grammatically perfect but still feel too formal, too enthusiastic, too repetitive, or simply unlike the user’s normal voice.
The new approach aims to reduce that gap.
Imagine that someone normally writes emails using short paragraphs, direct language and a particular sign-off. If ChatGPT has access to appropriate examples of that person’s work, those patterns can help guide future drafts.
The result isn’t necessarily a perfect imitation of every sentence. Instead, the goal is to provide writing that feels more consistent with the user’s established style and workplace context.
That could be especially useful for recurring tasks such as client emails, internal updates, project summaries, reports, proposals and other professional communications.
Connected Apps Remain Permission-Based
This capability depends on which applications are connected and what access the user or organization has granted.
For example, ChatGPT’s Google integrations can access information needed for supported requests based on the user’s Google permissions and workspace settings. OpenAI says connecting a Google app does not automatically create a personal synchronized index, and organizations can also impose additional administrative controls.
Similarly, SharePoint access depends on the organization’s Microsoft account permissions and workspace configuration.
Slack can also be connected so ChatGPT can work with messages and files that the user already has permission to access.
That means ChatGPT isn’t simply given unrestricted access to every piece of company information. What it can retrieve depends on the connected service, account permissions, workspace configuration and available features.
Privacy and Data Controls Matter
The ability to use personal writing patterns from workplace information also makes data controls particularly important.
OpenAI says that information accessed directly through connected Google apps is not used to train its generalized models, except in situations such as when the information is manually copied or uploaded into a ChatGPT conversation, included in a response, or submitted as feedback.
Organizations can also control which applications and actions are available to employees. Workspace administrators can manage connected applications and their permissions, while Microsoft and Google administrators may need to approve certain access scopes.
Users should therefore review their organization’s policies before connecting sensitive email, documents or communication systems.
Setting Up Your Writing Preferences
The information provided describes setup through ChatGPT’s Writing style controls, while OpenAI’s current documentation also emphasizes that Work can use connected apps, files and project context to create and refine workplace content.
The broader Work experience is designed to let users bring together relevant context, describe the desired outcome and have ChatGPT create or refine the resulting document, presentation, spreadsheet or report.
That makes personalized writing just one part of a much larger shift toward AI assistants that can work with a person’s existing digital workspace rather than operating only from the text typed into a single chat.
A Step Toward More Personalized AI Assistants
The significance of this development goes beyond making emails sound slightly more natural.
If AI assistants can understand how an individual communicates, what information they routinely work with, and which tools they use, they can potentially become much more useful for everyday professional tasks.
Instead of repeatedly explaining:
“Write this in my usual tone.”
the assistant can increasingly use established context to produce a first draft that requires less editing.
That could save time across thousands of routine communications while allowing users to concentrate more on decisions, ideas and higher-value work.
At the same time, the more personal context an AI system can access, the more important permissions, transparency and privacy controls become. The long-term value of personalized AI will therefore depend not only on how accurately it can imitate someone’s writing, but also on how responsibly that personalization is implemented.
