Tech • IA • Crypto
Build Hour: Workspace agents in ChatGPT
8/10TL;DR
OpenAI has introduced Workspace Agents in ChatGPT, enabling teams to automate complex, multi-step workflows across tools like Slack, Google Workspace, and Microsoft apps.
KEY POINTS
What Workspace Agents Are
Workspace Agents are Codex-powered AI agents designed for team use, capable of handling long-running, multi-system tasks. They can access files, tools, and code, operate in the background, and retain memory to improve over time. Unlike single-prompt assistants, they execute workflows autonomously and can be shared across teams.
Availability and Pricing
The feature is currently in research preview for ChatGPT Business, Enterprise, Edu, and teacher plans. It is free until May 6, after which pricing will shift to a credit-based model depending on task complexity and compute usage.
Meeting Preparation Automation
One showcased use case is a meeting prep agent that connects to Google Calendar, Drive, and Gmail. It automatically reviews upcoming meetings, researches participants, and generates structured briefs with summaries, objectives, and insights. The system can save users hours of daily preparation time and distribute outputs via email.
Slack-Based Software Review Agent
Another example is a software request agent integrated into Slack, designed to handle employee tool requests. It evaluates requests against internal policies, compares with approved tools, checks license availability, and can escalate decisions to Jira. This reduces manual IT workload and limits redundant software spending.
No-Code Agent Creation
Agents can be built using natural language prompts, with ChatGPT guiding setup, tool integration, and workflow design. Users can start from scratch, use templates, or import existing workflows. This lowers the barrier for non-engineers to create advanced automation.
Skills, Memory, and Customization
Agents rely on skills—codified best practices or workflows—and can combine multiple skills in a single task. They also include persistent memory, allowing them to store context and improve performance over time. Memory is scoped per user or per shared environment, such as a Slack channel.
Enterprise Controls and Governance
Organizations retain strict control over agent behavior through permissions, role-based access controls, and compliance APIs. Builders can limit tool access (e.g., read-only calendar access), while admins can monitor usage and define who can create or share agents.
Integration Ecosystem
Workspace Agents support Google Workspace, Slack, and Microsoft tools like SharePoint, Outlook, and Teams. While Slack integration is currently more advanced, broader support across enterprise tools is actively expanding.
Positioning vs Other AI Tools
Workspace Agents are positioned as team-oriented automation tools, distinct from Codex for individual use and the Agents SDK for custom product integrations. They represent an evolution of custom GPTs, adding multi-step execution, scheduling, and cross-system capabilities.
CONCLUSION
Workspace Agents mark a shift toward collaborative, autonomous AI systems that can execute real business workflows, potentially reshaping how teams manage routine operations and internal processes.