AI Agent (CAT-MIP-000000031)
Accepted • 2025-09-19 • by nicole
Definition
An AI Agent is a software-based autonomous entity that interprets natural language prompts and executes multistep reasoning and actions using defined tools, APIs, and structured knowledge. AI Agents interact with the platform, devices, users, and other systems to perform tasks such as triage, reporting, automation, or decision support. AI Agents can chain tools, call scripts, and escalate issues, often acting on behalf of a technician or service provider.
Prompt Examples
- AI Agent, why is device "ENG-LT-22" running slowly this week?
-
Summarize patch compliance across all tenants and suggest remediation for non-compliant devices.
-
What backup risks exist for customers using Microsoft 365, and how should we address them?
Agent Execution
When a prompt refers to an AI Agent, it will:
- Call tools (e.g., APIs, scripts, workflows) to gather data or take action
- Interpret the natural language request and identify intent
- Optionally escalate, log, or interact with other AI Agents
- Reason over context (e.g., device health, policy violations, previous incidents)
- Return human-readable results or initiate further actions
- 'Use structured terminology (from the MSP MCP Terminology Standard) to disambiguate terms like "asset" vs. "device"'
Synonyms
- Agentic AI
- Assistant Agent
- Automation Agent
- GenAI Agent (in LLM-driven systems)
- Intelligent Agent
- Orchestration Agent
- Proactive Assistant
Relationships
- AI Agent calls Action or Tool or Skill
- AI Agent interactsWith User, Device, or Platform
- AI Agent interprets Prompt
- AI Agent isEnabledBy Model Context Protocol (via MCP)
- AI Agent produces Result or SuggestedResolution
- AI Agent queries DataSource or KnowledgeBase
History
| Date | Author | Reason |
|---|---|---|
| 2025-08-07 | nicole | Draft – initial term proposal |
| 2025-09-19 | nicole | Accepted – added to CAT-MIP registry |
| 2026-01-14 | ||
| 2026-03-02 | jimmypuckett | Fix malformed YAML |