AI Service Desk Agent with Remote Diagnostics
Go · LLM tool calling · ScreenConnect · ImmyBot

An AI layer on top of a managed service provider's ticketing platform. Every incoming ticket is read and classified: simple requests get an immediate, useful reply to the end user, and urgent ones are escalated to the right Tier 1, 2 or 3 technician. For device problems, the agent identifies the user's actual machine through ScreenConnect and ImmyBot, runs a diagnostic scan on it, and hands the technician the probable cause and the steps to fix it.
Technologies
My role
Built the AI triage and diagnostics flow inside a team platform: ticket classification and auto-replies, escalation rules, the device-identification step and the remote-scan tooling the model is allowed to call.
Highlights
- Classifies every ticket, auto-replies to ~30% of routine tickets and escalates high-priority ones by tier
- Resolves the requester to their real device before acting, so the scan runs on the right machine
- Remote diagnostic scan whose findings become step-by-step fix instructions for the technician
- Device diagnosis in minutes instead of ~30 minutes per machine
- Model access through a gateway with an allow-list of models and a usage budget
- Destructive actions require an explicit human acknowledgement; everything is scoped per client tenant
Screenshots and diagrams
