OpenAI has introduced ChatGPT Work, a new agentic mode capable of gathering information from multiple applications, documents and connected services before combining it into a finished business deliverable.
Instead of requiring users to manually copy information into a conversation, ChatGPT Work can search connected sources, identify relevant context and use the combined information to create documents, spreadsheets and presentations.
The new mode is powered by GPT-5.6 and is available through ChatGPT’s desktop and web applications, although access varies according to device and subscription.
What is ChatGPT Work?
ChatGPT Work is designed for tasks that require more than a single prompt and response. Users can provide multiple documents and connect external services through plugins, allowing the AI agent to search across those sources and assemble the information required to complete a task.
For example, a user could ask ChatGPT Work to:
- Combine sales data from a spreadsheet with customer information from a connected business application
- Review several research reports and produce an executive presentation
- Compare project documents and create a consolidated status report
- Analyse information from different departments and prepare a management briefing
- Turn source materials into a formatted spreadsheet, document or slide deck
The important change is that ChatGPT is no longer limited to generating an answer inside the chat window. It can use information from several sources to produce a more complete end product.
ChatGPT can plan before it acts
For complex requests, ChatGPT Work can ask follow-up questions and develop a proposed sequence of steps before beginning the work. This gives users an opportunity to clarify the objective, confirm assumptions and correct the approach before the agent processes large volumes of information.
A typical workflow could involve:
- Understanding the requested business outcome
- Identifying the documents and connected applications required
- Asking questions about missing information or ambiguous instructions
- Preparing a plan for completing the task
- Collecting and analysing the relevant data
- Producing the requested document, spreadsheet or presentation
This planning capability is especially useful for business assignments where the quality of the outcome depends on choosing the right sources and applying the correct process—not merely writing convincing text.
GPT-5.6 provides the intelligence behind Work
ChatGPT Work runs on GPT-5.6, OpenAI’s latest model family. GPT-5.6 is available in three principal variants—Sol, Terra and Luna—which are designed for different balances of capability, speed and token consumption.
This gives OpenAI a way to match the model to the task. A demanding assignment involving complex reasoning and multiple sources may justify a more capable model, while routine document processing could potentially use a faster and more economical option.
For organisations, this distinction matters. The most powerful model is not necessarily the best choice for every workflow. Cost, turnaround time, accuracy and the sensitivity of the task should all influence model selection.
Why combining sources matters
Most workplace knowledge is fragmented. Customer information may be stored in a CRM, financial figures in spreadsheets, policies in documents and project discussions in collaboration platforms. A conventional chatbot usually depends on the user finding and supplying this information manually.
ChatGPT Work attempts to remove that bottleneck. By retrieving context from connected sources, the AI can help users work across organisational boundaries without repeatedly switching applications or recreating information.
This positions ChatGPT Work as a potential orchestration layer above existing business systems. It does not necessarily replace those systems; instead, it helps users find, combine and transform the information stored within them.
Potential business applications
Sales and account management
A sales team could combine CRM records, meeting notes, proposals and market research to produce an account plan or customer briefing. The AI could identify gaps in the available information and ask the user for clarification before drafting the final document.
Management reporting
Managers could bring together departmental updates, financial data and project reports to create a consolidated executive summary. This may reduce the time spent manually collecting updates and reformatting them for leadership meetings.
Research and analysis
Analysts could supply reports from several sources and ask ChatGPT Work to compare findings, identify inconsistencies and prepare a structured presentation. Human verification would still be necessary, particularly when the result informs a significant decision.
Project delivery
Project teams could use the agent to review plans, risks, meeting notes and status reports before producing a current project overview. This could make it easier to identify conflicting dates, unassigned actions or emerging delivery risks.
Accuracy remains more important than presentation quality
Generating a polished presentation is not the same as producing an accurate one. When an AI system combines multiple sources, it may encounter outdated information, conflicting figures, incomplete documents or ambiguous terminology.
Businesses should therefore establish a review process that checks:
- Whether every important claim can be traced to a source
- Whether figures have been copied and calculated correctly
- Whether outdated documents were excluded
- Whether conflicting information was identified rather than silently resolved
- Whether the final output follows organisational templates and policies
- Whether a qualified person has approved the result
The greatest productivity gains are likely to come from using ChatGPT Work to accelerate research, consolidation and initial production while retaining human accountability for the final output.
Connected data introduces governance questions
ChatGPT Work becomes more capable as it receives access to additional documents and services. That same access also increases the importance of permissions, data classification and auditability.
Before connecting business systems, organisations should determine:
- Which data sources each user and agent may access
- Whether source permissions are preserved when information is combined
- How confidential information is protected in generated outputs
- Whether prompts, retrieved data and completed files are retained
- How agent activity can be reviewed and audited
- Which actions require explicit human approval
An employee should not be able to retrieve restricted information through an AI agent simply because the agent has been connected to a wider organisational data source.
ChatGPT Work availability
According to the initial announcement, GPT-5.6 and ChatGPT Work are broadly available to users of the Windows and macOS desktop applications, including free users.
Access through the web application and mobile devices is initially available to Pro, Enterprise and Edu subscribers. OpenAI plans to extend access to Plus and Business customers at a later stage.
Availability, model limits and connector support may continue to change as OpenAI expands the service.
From chatbot to digital worker
ChatGPT Work reflects a wider shift in enterprise AI. The first generation of generative AI tools helped people write and summarise content. The next generation is expected to locate information, coordinate multi-step workflows and produce deliverables across several applications.
If ChatGPT Work can reliably preserve source context, respect access controls and generate accurate outputs, it could reduce a significant amount of administrative work. Its real value will not be measured by how convincingly it writes, but by whether it can transform scattered organisational data into trustworthy, usable results.