OpenAI has introduced a feature named company knowledge for its ChatGPT Business, Enterprise, and Edu plans. The function connects the artificial intelligence platform to a business’s internal tools to provide answers informed by internal company context.
In many professional environments, critical information is distributed across a wide array of unconnected internal applications. This includes documents, spreadsheets, internal messages, emails, support tickets, and project management trackers. The separation of these tools often means that finding a complete and accurate answer to a business question requires manual searches across multiple platforms, creating inefficiencies and potential information gaps. The most relevant data for a specific task may exist in several locations simultaneously, making a unified understanding difficult to achieve.
On October 21, 2025, OpenAI released company knowledge to address this challenge. The feature is designed to aggregate the context from these disparate applications directly within the ChatGPT interface. By connecting with a company’s internal information ecosystem, it enables the platform to generate responses that are specific to that organization’s operations, data, and internal discussions. The stated goal is to assist users in making decisions, executing tasks, and advancing their work with greater access to relevant internal information.
When activated, company knowledge makes the information held within connected applications such as Slack, SharePoint, Google Drive, and GitHub more useful and accessible to users within the organization. The capability is driven by a variant of the GPT-5 model, which has been specifically trained to conduct searches across multiple data sources. This allows it to synthesize information from various locations to provide more comprehensive and precise answers. To ensure transparency and allow for verification, every response generated using this feature includes explicit citations, indicating the exact source of each piece of information used in the answer.
A practical application of this function can be seen in preparing for a client meeting. An employee could ask ChatGPT to generate a briefing for an upcoming call. The system would then access and process information from multiple connected sources relevant to that client. It could pull recent messages from a dedicated client channel in Slack, extract key details from email exchanges, review notes from the last call stored in a Google Docs file, and identify any support escalations logged in Intercom tickets since the previous meeting. The final output would be a synthesized briefing document compiled from these varied sources.
The company knowledge feature is available to all subscribers of the ChatGPT Business, Enterprise, and Edu plans starting from its release date. A core aspect of its design is its adherence to pre-existing data access permissions within the client organization. This security protocol ensures that ChatGPT, and by extension the user, can only access information and documents that the user is already authorized to view. The system does not grant any new or elevated access privileges, instead operating strictly within the boundaries of each individual’s established permissions.
To use the feature, a user begins by selecting the “Company knowledge” option located beneath the message composition area in the ChatGPT interface. For first-time use, the system will prompt the user to establish connections to their work-related applications. This is a one-time setup process. Subsequently, for any new conversation where internal context is required, the user must enable the feature. Once activated for a given chat, ChatGPT will automatically search across all connected applications to incorporate relevant context into its responses.
As the system processes a query, users can observe its activity in a sidebar panel. This interface provides a transparent view of the sources ChatGPT is examining and illustrates how it is utilizing the retrieved information to construct its answer. Upon completion, the final response is presented alongside a detailed list of the sources it consulted. This includes the specific snippets of text or data that were drawn from each source. Users can click on any of these citations to navigate directly to the original file or message for further detail or context.
The system is engineered to manage complex and sometimes conflicting information found within an organization. Company knowledge can execute multiple, parallel searches to resolve discrepancies and provide clarity on ambiguous questions where a single, clear answer may not exist. This capability is particularly useful in situations where different teams may have differing perspectives or when a final decision on a topic has not yet been reached. For instance, a query such as, “Where did we land on company goals for next year?” could prompt ChatGPT to summarize various discussions, highlight points of agreement, and identify differing viewpoints, thereby equipping the user to facilitate the next steps with their team.
This multi-source analysis also enables the generation of more balanced and comprehensive responses. When asked to “Report on customer feedback from the mobile launch,” the system is designed to go beyond a simple sentiment summary. It can also incorporate related support tickets, pull direct customer quotes from communications, and even suggest potential next steps based on the synthesized feedback. This provides a more complete and nuanced picture of the situation, intended to support better-informed decision-making by leadership and product teams.
The feature incorporates date filters and an ability to reason during its search process, making it proficient at retrieving time-based information. A user asking for a “quick update on our company performance” could receive a response compiled from multiple Slack channels, Google Docs, and Google Sheets. The system would rank these sources by recency and quality to deliver the most current possible view. It is also capable of retrieving historical data from a specific period. A user could ask it to “Look back at our company performance in Q1,” and the system would locate the relevant past files and messages without requiring the user to perform manual archival searches.
OpenAI has implemented several layers of control and security for organizations using this feature. Administrators on Enterprise and Edu plans have the ability to manage access to the connected applications that serve as the data sources for company knowledge across their entire workspace. Within this administrative dashboard, they can create custom roles and configure group-level permissions. This granular control ensures that users are only able to access the information that is necessary for their specific roles and responsibilities.
Regarding data privacy, the system is built to respect all existing user permissions within a company’s IT infrastructure. OpenAI has stated that it does not train its models on customer data from these plans by default. For security, all data transmitted and at rest is protected with industry-standard encryption. Access management at scale is facilitated through support for Single Sign-On (SSO) and System for Cross-domain Identity Management (SCIM). Furthermore, organizations can implement IP allowlisting, a security measure that restricts access to ChatGPT exclusively to traffic originating from company-approved IP addresses.
For compliance and regulatory requirements, administrators are provided with access to conversation logs. This is accomplished through the Enterprise Compliance API, which allows for auditing, reporting, and fulfillment of regulatory obligations. These security and privacy measures are part of a broader enterprise-grade framework designed to give organizations control over their data and its usage.
Company knowledge is positioned to assist with more than simple informational queries like “How do I file an expense report?” Its ability to understand work context allows it to be a more effective partner in complex tasks such as drafting plans, compiling reports, or getting up to speed on a new project. One specific use case involves transforming customer insights into strategic documents. ChatGPT can synthesize recent customer feedback from Slack, analyze survey results presented in Google Slides, and identify key themes from support tickets to help inform product roadmap planning.
Another application is the creation of timely reports. Following a marketing campaign, a user can instruct ChatGPT to pull relevant contacts or deals associated with the campaign from a CRM like HubSpot. It can also incorporate information from creative briefs and post-mortem notes stored in Google Docs, and gather highlights shared in email threads to generate a comprehensive performance summary report.
In the domain of software development, the feature can be used to help build release plans. ChatGPT is able to scan a company’s GitHub repository for open “TODO” items, check a project management tool like Linear for related tickets, and search through engineering channels on Slack for any unresolved bug reports. It can then summarize what tasks are outstanding, which items are already being tracked, and what issues still need to be formally logged, streamlining the planning process for a new feature release.
There are current limitations to the feature. As of its launch, users must manually select company knowledge each time they start a new conversation. If it is not selected, ChatGPT might still utilize connected applications to answer questions, but the responses will not have the same depth or include the detailed citations provided when the feature is explicitly enabled. OpenAI has stated that it is working to merge these two experiences in the coming months.
Another current constraint is that when company knowledge is active, ChatGPT cannot perform live web searches or create data visualizations like charts and images. To use these other capabilities, a user can turn the company knowledge feature off within the same conversation. This allows them to retain the context already established in the chat while accessing the platform’s other functions. OpenAI has announced plans to integrate company knowledge with the full suite of ChatGPT capabilities in the future.
OpenAI is also actively expanding the number of tools that can be connected to the system to make its responses more comprehensive. This week, the company introduced new connectors for a range of additional applications, including Asana, GitLab Issues, and ClickUp, among others. This expansion of supported integrations is an ongoing effort to broaden the scope of internal knowledge that the feature can access.
