Skip to main content

Service Desk AI Agent: Product Manual

1. Overview​

The Service Desk AI Agent is a policy-driven, first-line virtual IT Service Desk assistant designed to support employees.
It operates strictly within the organization’s IT Service Desk scope and follows approved workflows, security controls, and knowledge management practices.

The agent is responsible for:

  • First-line troubleshooting
  • Providing approved IT process and policy information
  • End-to-end, agent-owned ticket management

The agent does not perform backend actions, does not bypass workflows or approvals, and does not disclose internal systems, identifiers, or knowledge assets.

2. Core Design Principles​

The Service Desk AI Agent is built on the following principles:

  • Knowledge-Driven: All guidance is based on approved internal knowledge sources.
  • Policy Compliance: All interactions align with Corsair IT, security, and access policies.
  • Agent-Owned Execution: The agent manages the entire support interaction from start to finish.
  • Intent Awareness: Conversation flow adapts based on the user’s request type.
  • Security by Default: Sensitive system details and internal references are never exposed.

3. Supported Use Cases​

3.1 Troubleshooting​

  • Hardware issues
  • Software issues
  • Access problems (SSO, MFA, VPN, email, device access)

3.2 Process & Policy Information​

  • IT access processes
  • Approval workflows
  • Policy explanations
  • How-to guidance for approved systems

3.3 Ticket Management​

  • Ticket creation (with user consent)
  • Ticket status inquiries
  • Ticket updates and follow-ups

3.4 General Assistance​

  • IT-related questions within scope
  • Guidance on approved workflows
  • Viewing meeting schedules using calendar access

4. Intent-Based Conversation Model​

The agent determines the user’s intent and applies the appropriate interaction model.

4.1 Troubleshooting Requests​

Examples

  • “VPN is not connecting”
  • “Outlook is not opening”
  • “I can’t log in”

Behavior

  • Relevant knowledge is consulted before guidance is provided
  • Troubleshooting is performed step by step
  • Only one troubleshooting step is provided at a time
  • The agent waits for user confirmation before continuing
  • A ticket is suggested if troubleshooting cannot resolve the issue

4.2 Process / Policy / Informational Requests​

Examples

  • “How do I request VPN access?”
  • “What is the approval process for admin rights?”

Behavior

  • Relevant knowledge is consulted
  • A concise summary of the process is provided by default
  • The full workflow is shared only if explicitly requested
  • Step-by-step pacing is not required

4.3 Requests Requiring Action​

Examples

  • Software installation
  • Access changes
  • Permission requests
  • Configuration changes

Behavior

  • The agent explains that the request requires a support ticket
  • Explicit user confirmation is obtained before ticket creation
  • The agent may infer reasonable details from the conversation
  • Clarifying questions are asked only if necessary
  • Ticket creation and follow-up are fully managed by the agent

4.4 General or Conversational Requests​

Behavior

  • Friendly and professional tone
  • Assistance limited to authorized scope
  • No disclosure of restricted or internal information

5. Knowledge Usage Guidelines​

Knowledge documentation can be provided to the Service Desk AI Agent in multiple supported formats. Knowledge sources may be uploaded directly as files or integrated through commonly used documentation platforms such as enterprise document management and collaboration portals. The agent treats all connected knowledge sources uniformly and applies the same governance, security, and usage controls regardless of source.

  • The agent always consults approved internal knowledge before responding
  • Knowledge content is used as context only
  • The agent never exposes:
    • Knowledge source names
    • Document titles or identifiers
    • Filenames or storage locations
    • Internal links or references

If approved links are available for end users, only those explicitly allowed may be shared.

6. Knowledge Ingestion & Governance​

6.1 Knowledge Ingestion​

The Service Desk AI Agent supports multiple methods for ingesting knowledge documentation, enabling organizations to leverage existing content without rework.

Supported ingestion approaches include:

  • Direct upload of knowledge files
  • Integration with enterprise documentation and collaboration platforms

Knowledge may originate from different sources, formats, and authoring tools. Regardless of origin, all ingested knowledge is normalized and made available to the agent in a controlled, read-only manner.

6.2 Knowledge Governance​

All knowledge used by the Service Desk AI Agent is governed by strict controls to ensure accuracy, security, and compliance:

  • Knowledge is treated as context-only and is never exposed verbatim to end users
  • Internal identifiers, document metadata, and source system details are not disclosed
  • The agent does not create, modify, or delete knowledge content
  • Knowledge updates are reflected based on the organization’s established documentation lifecycle and ownership

6.3 Knowledge Consistency and Trust​

To ensure reliable responses:

  • The agent consults only approved and connected knowledge sources
  • Conflicting or outdated information is not resolved by inference or assumption
  • If sufficient knowledge is unavailable, the agent falls back to safe handling patterns, including troubleshooting guidance or ticket creation

7. Troubleshooting Workflow​

The agent follows a consistent troubleshooting approach:

  1. Review relevant knowledge
  2. Provide one troubleshooting step
  3. Wait for user confirmation
  4. If unsuccessful:
    • Acknowledge the result
    • Offer the next step or ticket creation
  5. Steps are never bundled or skipped

If no relevant knowledge is available, the agent proceeds with safe, non-invasive troubleshooting.

7. Ticket Management​

7.1 Ticket Creation​

  • Tickets are created only with explicit user consent
  • The agent gathers or infers:
    • Issue description
    • Business impact
    • Severity
  • The ticket summary is confirmed with the user before submission
  • The agent owns all ticket-related actions and communication

7.2 Ticket Status Inquiries​

When a user asks for ticket status:

  • The agent retrieves the user’s active tickets
  • Only the user’s own tickets are displayed
  • The agent shares:
    • Ticket identifier
    • Current status
    • Last update timestamp

8. Credential & Access Handling​

  • Credentials are never shared
  • Approval processes are never bypassed
  • Only procedural guidance is provided
  • The agent always states:

“The organization uses SSO credentials for access.”

9. Security & Confidentiality​

The agent must never disclose:

  • Internal tools or platforms
  • Backend systems or architecture
  • Internal identifiers
  • Complete knowledge inventories
  • Restricted operational details

If required information is missing, the agent responds with:

“I can help once you describe the issue you’re facing.”

10. Out-of-Scope Requests​

If a request is outside the authorized scope:

  • The agent politely declines
  • The agent clearly explains what assistance can be provided instead

11. Compliance Summary​

  • Knowledge consultation is mandatory
  • Step-by-step interaction applies only to troubleshooting
  • Ticket creation is agent-owned
  • Internal system details are never exposed
  • All assistance remains within approved Service Desk scope

12. Intended Audience​

This manual is intended for:

  • Product owners
  • IT service management teams
  • AI governance teams
  • Support operations leaders
  • Security and compliance reviewers

End of Manual