Plan multi-step work
Agents can break a task into steps and decide what to do next.
AI Agents
Give workflows agents that can plan and act
Use agents for tasks that need reasoning, tool use, and iteration. Agents can plan steps, call connected tools, observe results, and continue until the task is complete.
Core Capabilities
What this feature helps your team do
Use connected systems
Give agents access to approved integrations, APIs, and workflow actions.
Set limits and guardrails
Constrain steps, tool access, runtime, and budget so agents remain operationally safe.
Agentic workflows
Use agents where fixed automation is not enough
Agents are useful when the task requires evaluation, research, tool selection, or iteration instead of a fixed sequence of actions.
- Multi-step reasoning
- Tool calling through integrations
- Structured output
- Execution and cost limits
How It Works
A practical path from setup to production
Define the mission
Write the objective and operating instructions.
Approve tools
Choose which integrations and actions the agent can use.
Set limits
Control budget, steps, runtime, and output format.
Run in workflow
Use the agent result in downstream automation steps.
Common Use Cases
Where teams use AI Agents
Research and enrichment
Support triage
Operations investigation
Document review
AI Agents questions
Clear answers about how this capability fits into real workflow automation.
Normal AI steps usually produce one output from one prompt. Agents can plan, call tools, observe results, and continue through a task.
Yes, but only through the tools and connections you make available to them.
Yes. You can set operational limits such as step count, runtime, tool access, and budget.
Yes. Agent results can be passed into downstream actions, branches, approvals, and notifications.
Automate Every Job