AI Agents
AI Agents
An Outrun agent is a specialist with a defined goal, instructions, connected context, tools, and an approval posture. Agents can participate in conversations, run recurring work, and coordinate with other specialists through agent groups.
Current Templates
The template library evolves with the product. The current customer and operational templates include:
| Area | Templates |
|---|---|
| Customer lifecycle | Pre-Sales Agent, Lead Qualification Agent, Sales Closer, Customer Success Agent, Customer Service Agent |
| Prospecting and follow-up | Outbound Prospecting, Deal Follow-up, Lifecycle Follow-up |
| Account operations | Activity Digest, Daily Account Review, Proactive Nudge, Proactive Nudge Scan |
| Assistants and recurring work | Executive Assistant, Workspace Setup Assistant, Daily Agent Job |
| Data operations | Data Entry Agent |
Choose the role closest to your desired outcome. Templates provide a starting configuration; your connected knowledge, tools, rules, and approval choices determine what the deployed agent can do.
The Parts of an Agent
Goal and instructions
Define the job, success condition, boundaries, tone, and escalation rules. Keep the role narrow enough that a reviewer can tell whether it behaved correctly.
Knowledge and context
Attach approved business knowledge and supported sources that the role needs. A connection may expose retrieval, actions, synchronization, or a subset of those capabilities. Check the relevant integration documentation before relying on a specific operation.
Tools and actions
Tools let an agent retrieve data or propose and perform supported actions. Available tools depend on the integrations configured in the workspace. Give each role the minimum tool access required for its job.
Human-in-the-loop controls
Use human review for customer-facing messages, external writes, or other sensitive actions while validating a workflow. A reviewer can inspect and respond to proposed work according to the available approval controls.
See Human in the loop for setup and operating guidance.
Agent Groups and Handoffs
Use an agent group when a customer or operational workflow needs more than one specialist. A group can coordinate roles and preserve relevant context across a handoff—for example:
- a Pre-Sales Agent answers the initial product question;
- a Lead Qualification Agent applies your criteria;
- a Sales Closer handles the next sales step; and
- a Customer Success Agent continues the relationship after the deal.
Start with one reliable role. Add a group after the individual inputs, actions, and escalation rules are understood.
Correction Memory
When a person corrects agent work, Outrun can store that example and retrieve relevant corrections in similar situations. Corrections are useful for exceptions and nuance; repeated systemic errors should be fixed in the source knowledge, instructions, tools, or approval policy.
Corrections are guidance, not a guarantee. Continue reviewing outcomes as your product, policies, and customer situations change.
Recommended First Deployment
- Pick one template and one measurable outcome.
- Connect only the context and tools required for that job.
- Document when the agent must stop or ask for a person.
- Keep approval enabled for sensitive actions.
- Test a representative case and an edge case.
- Inspect the activity and correct the underlying configuration.
- Expand its scope only after the role earns trust.
