Workspace agents

Workspace agents

Understand, build, and use agents for repeatable work in ChatGPT.

Most ChatGPT users already know how to use AI for one-off tasks—like drafting, summarizing, brainstorming, or answering questions. The next phase of AI use is broader and more embedded in day-to-day work. Instead of helping with isolated moments, AI is increasingly being used to support repeatable workflows that depend on shared systems, standard handoffs, consistent outputs, and real-world constraints like timing, accuracy, and process.

That’s where workspace agents in ChatGPT fit. They’re designed to be used for repeatable workflows—work you’d otherwise do manually, re-explaining the steps each time, and copying information between tools.

Core Concepts of Workspace Agents

If you’re new to agent building, let’s focus on the core concepts first so when you start building, you’ll know how to set up your workspace agent for consistent results.

Generally speaking, an agent is a system that carries out a task with three components:

| Component | Description | Example | | :--- | :--- | :--- | | Trigger | What starts the agent | A schedule (“Every weekday at 9am”) or a manual run (“Run now”). | | Process | The steps the agent follows to complete the task | Reviewing inputs, checking for missing information, drafting an output, and handing it off. | | Tools | Approved tools and integrations the agent can use | Slack, a CRM, internal documentation, a ticketing system, or a shared document. |

When to Use an Agent

Agents are most useful when the work is:

• Repeatable: The same task comes up regularly.
• Structured: There’s a clear format for the output.
• Time-based or event-driven: It runs on a cadence or is triggered by an event.
• Tool-based: It requires reading from or writing to systems your team uses.

For open-ended thinking or brainstorming, regular chat is often a better fit.

Agents vs. Traditional API Workflows

Traditional workflows are often deterministic, meaning each step is explicitly defined. Agents are more probabilistic. They operate within instructions and guardrails, but use a model to interpret context, make bounded decisions, and adjust how they move through the work.

Designing a Workspace Agent

Think about what you would clarify before handing work to a person: responsibility, timing, tools, and rules.

| Feature | Campaign Analysis | Feedback Routing | Pipeline Monitoring | | :--- | :--- | :--- | :--- | | Task | Analyze performance and recommend optimizations | Summarize feedback and route to owners | Monitor pipeline changes and highlight risks | | Trigger | Every Monday at 10 AM | Form submission from Slack | Every weekday at 8 AM | | Process | Pull KPIs, identify trends, draft summary | Review feedback, group items, suggest owners | Review pipeline, flag risks, propose steps | | Tools | Analytics tools, shared document | Slack, ticketing system | CRM, email, pipeline tracker | | Guardrails | Approval required before budget changes | Create draft tickets; escalate high-priority | Draft insights; no direct outreach without approval |

Common Workflow Patterns

• Synthesis: Pull information from multiple places and package it for decisions.
• Triage: Process inbound items and ensure they reach the right next step.
• Analysis & Recommendation: Interpret data and turn it into a first deliverable.
• Content Creation: Create or update content, then tailor it for a specific channel.
• Planning & Execution: Turn goals into scheduled work and system updates.

Building Your Agent

Agent building is an iterative process.

1. Start in plain language: Describe the job, success criteria, and constraints.
2. Choose tools and connectors: Select approved apps for the workflow.
3. Choose a trigger: Decide when the agent should run.
4. Add guardrails: Set boundaries and human-in-the-loop checkpoints.

Always remember that human judgment is essential to oversee the agent's output and ensure it aligns with the broader context.