Skip to main content
RAUM AI Assistant is the Zendesk-facing RAUM product. It helps agents research tickets and draft replies inside Zendesk, and it can also process selected tickets autonomously when you enable Agent Mode. It combines the ticket conversation with your procedures, AI Guidelines, Zendesk knowledge, uploaded documents, websites, and approved external tools before it produces an answer.

Set up AI Assistant

Connect Zendesk, provision automation, install the Marketplace app, and run a safe first test.

Use the Zendesk app

Learn the controls agents use directly inside Zendesk.

​
Two operating modes

​
Copilot Mode

Copilot Mode keeps the Zendesk agent in control. An agent asks RAUM to generate or transform content inside the ticket editor. RAUM can retrieve relevant knowledge and approved tool data, then returns a draft for the agent to review before anything is sent. Use Copilot Mode when:
  • you are validating a new RAUM deployment;
  • the ticket requires human judgment;
  • the reply or external action is sensitive; or
  • you want agents to approve the final wording.
Learn about Copilot Mode →

​
Agent Mode

Agent Mode lets RAUM process webhook-triggered Zendesk tickets without waiting for an agent to click Generate. Administrators decide how confidence levels are handled, whether RAUM posts a public reply or an internal note, and what should happen to the ticket status when no explicit status decision was produced. Learn about Agent Mode →

​
What grounds an Assistant answer

AI Assistant can use several source types:
  • Procedures — factual knowledge and step-by-step runbooks.
  • AI Guidelines — response and translation rules targeted by channel, language, Zendesk metadata, and other conditions.
  • Zendesk Help Center articles — synchronized Guide content.
  • Ticket Fields — field descriptions and option guidance that help RAUM interpret and populate Zendesk fields.
  • Macros — synchronized Zendesk macro context.
  • Documents & Websites — uploaded or crawled knowledge.
  • Connections — live API or hosted MCP data and actions.
  • Gallery and Updates — optional library resources when those features are enabled.
The Assistant workspace deliberately keeps Zendesk-only resources separate from the public AI Chatbot trust boundary.

​
Bring your own AI provider

RAUM uses the AI provider credentials configured for your organization. RAUM’s backend receives the Zendesk context, applies your organization’s knowledge and policies, and sends the required model requests using the provider you selected. RAUM does not resell model capacity or route the request through a RAUM-owned AI account. Configure AI models →

​
Where administrators work

The AI Assistant workspace is organized into:
  • Dashboard — operational metrics.
  • Sources — procedures, guidelines, Zendesk resources, documents, and websites.
  • Library — Gallery and Updates when enabled.
  • AI — model status, models, Brand Voice, and Coaching.
  • Connections — API Connections, MCP Servers, and Authentication.
  • Zendesk — setup, automation, agent experience, enrichment, and Zendesk security.
  • Email & Agent Chat — Assistant behavior for Zendesk email and realtime agent conversations.
  • Activity — ticket activity, system activity, and audit views.
  • Quality & Testing — CSAT & Outcomes, Knowledge Gaps, and Simulation.
Some tabs depend on your role, product entitlement, or enabled feature flags. If a page described here is not visible in your workspace, your organization may not currently have access to that feature.
  1. Connect AI models and Zendesk.
  2. Synchronize the Zendesk sources you want RAUM to use.
  3. Add procedures and response guidelines.
  4. Configure read-only connections before state-changing actions.
  5. Test normal tickets in Copilot Mode.
  6. Use Simulation for full-pipeline testing.
  7. Enable Agent Mode only for the confidence levels and ticket scope you are ready to automate.

How Assistant works

Follow the reasoning, retrieval, tools, and reply flow.

Build your knowledge

Understand every source available to AI Assistant.

Connect external systems

Configure live API and MCP tools.

Quality & testing

Review outcomes, knowledge gaps, and safe simulations.