Appearance
AI Agent
Junto's AI agent is the core of the platform. It is an AI-powered assistant that triages incoming signals, gathers context from your connected integrations, and helps your team resolve issues through a conversational interface.
How It Works
When a new signal arrives (typically synced from a ConnectWise PSA ticket), the agent can begin working on it:
- Bootstrap -- The agent loads context about the signal: ticket details, customer information, device data, prior notes, and any relevant runbooks.
- Tool access -- Based on the integrations your organization has connected, the agent can query ConnectWise, NinjaOne, ITGlue, Sophos, Pax8, and Microsoft 365 for additional information.
- Conversational triage -- The agent presents its findings and recommendations in a chat interface. Technicians can ask follow-up questions, request specific actions, or provide guidance.
- Action execution -- When the agent needs to take an action (updating a ticket, adding a time entry, running a script), it requests approval from the technician before proceeding.
Each agent session runs in its own isolated environment with strict multi-tenant scoping. The agent can only access data and tools that belong to your organization, and tool access is further restricted to the specific company associated with the ticket being triaged.
Available Tools
The agent has access to tools from each connected integration, organized into three categories:
- Read tools retrieve information without making changes. Examples include looking up a ticket, listing devices, or searching documentation.
- Write tools modify data in your external systems. Examples include updating a ticket status, adding a note, or creating a time entry.
- Execute tools perform operations beyond simple data changes, such as running a script on a remote device through NinjaOne.
The specific tools available depend on which integrations your organization has connected.
Tool Approval System
Junto uses a tiered approval system to keep your team in control.
Approval Policies
Each tool has an approval policy that determines when human approval is required:
- Auto-approved -- Read-only tools like looking up a ticket or searching documentation run without asking.
- Once per session -- After a technician approves a tool once during a triage session, subsequent uses in that session are auto-approved.
- Always require approval -- High-impact tools always prompt for approval, even if they have been used before in the same session.
How Approval Works
When the agent wants to use a tool that requires approval:
- The agent pauses and sends an approval request to the triage interface.
- The technician sees the tool name, a description of what it does, and the specific parameters.
- The technician can approve, deny, or modify the parameters before approving.
- If denied, the agent receives the denial reason and adjusts its approach.
Role-Based Permissions
Junto uses a five-level role hierarchy that determines what each team member can approve:
| Role | What They Can Approve |
|---|---|
| Owner | Everything, including billing and organization settings |
| Admin | Full operational access to all tools and settings |
| Manager | High-risk tools, team management |
| Operator | Medium and low-risk tools |
| Technician | Low-risk tools and standard workflows |
Organization-Level Overrides
Administrators can customize approval policies at the organization level. For example, you might decide that creating internal notes should be auto-approved while keeping time entry creation as an always-approve action.
Runbook Execution
The agent can also operate in runbook mode, where it follows a documented procedure step by step:
- The agent reads the runbook content and executes each step using the available tools.
- Tools mentioned in the runbook are auto-approved to allow unattended execution.
- The agent creates a ConnectWise ticket, logs time, and adds notes documenting what it did.
This is useful for repetitive tasks like new user onboarding, password resets, or scheduled maintenance procedures.
Next Steps
- Connect your integrations to give the agent tools to work with.
- Configure processors to automate signal classification.
- Create runbooks for common procedures the agent can follow.