Serval Separates AI Help-Desk Reasoning from Controlled Jamf Automations

· Original article ↗

Summary

Serval’s approach separates an AI help-desk agent from a second agent that builds explicit, versioned workflows. People review and test workflows before use; runs are deterministic and auditable, without giving the LLM Jamf credentials.

Key points

  • The help-desk agent interprets employee requests and gathers context but does not act on production systems.
  • A separate automation agent builds explicit, versioned workflows that call Jamf’s API; a person reviews and tests each workflow before publication.
  • Workflows follow consistent steps and are auditable, and the LLM does not hold Jamf credentials or improvise API calls at runtime.
  • Demonstrated workflows cover FileVault key recovery with identity and device checks, hardware refresh approvals against organizational policy, and ticket triage using device history.
  • The session’s evaluation questions include what an agent can do versus propose, whether actions can be reviewed and reconstructed, who approved automations, and how changes are tracked.

Article Details

Topic
Using reviewed, deterministic Serval workflows to let AI agents resolve IT help desk requests through Jamf automations without giving the agents production credentials.

People

Vendors

Products