What Are Runbooks?

Runbooks in Junto are structured procedures that standardize how common IT tasks are performed. They serve as documented playbooks that guide both the AI agent and human technicians through repeatable workflows, ensuring consistency and reducing errors.

Purpose

• Standardize procedures across your team so every technician follows the same steps.
• Guide the AI agent by providing it with the exact workflow to follow when handling tickets.
• Reduce resolution time by eliminating guesswork with clear, step-by-step instructions.
• Maintain quality by encoding best practices into reusable, version-controlled documents.

Scope Types

Every runbook has a scope that determines its visibility:

• Global -- Available across all companies in your organization. Use for universal procedures like password resets or new user onboarding.
• Company-specific -- Scoped to a single client company. Use when a client has unique systems or policies that differ from your standard procedures.

You can fork a global runbook into a company-specific variant when a client needs a customized version.

Audience Modes

The audience setting controls who the runbook is written for:

• AI Agent (Agentic) -- The AI agent executes the runbook autonomously, calling integration tools and completing steps without human intervention.
• Hybrid -- The AI agent and a human technician collaborate. The agent handles automated steps while prompting the technician for decisions requiring human judgment.
• Human -- The runbook serves as a reference guide for technicians. The AI can surface it as context but does not execute it.

How Runbooks Work with the AI Agent

When a ticket arrives, the AI agent automatically matches it against your published runbooks. If a relevant runbook is found, the agent follows the documented steps using integration tools.

Key integration points:

• Tool mentions -- Runbooks reference specific tools (ConnectWise, NinjaOne, etc.) using @mentions. The agent calls the referenced tools based on these mentions.
• Required inputs -- Runbooks can define inputs that must be gathered before execution, such as a user email or license type.
• Required integrations -- The system auto-detects which integrations a runbook depends on based on the tools it references.

Lifecycle

A runbook moves through three statuses:

• Draft -- Being written or revised. Not available to the AI agent.
• Published -- Active and available for the AI agent to use when triaging tickets.
• Archived -- Retired. No longer matched to tickets, but version history is preserved.