AI Agent
Also known as: Agentic AI, AI workflow agent
Definition
An AI agent is an AI-enabled workflow component that can plan or execute multiple steps, sometimes using tools or business systems. In a business context, safe agents require bounded scope, permissions, logging, evaluation, human approval points, and incident handling.
Not every AI workflow should be called an agent. Many valuable workflows are safer as copilots or automations with clear human review.
Agent autonomy should expand only after quality, permissions, and monitoring are proven.
Related terms
- Human-in-the-Loop — A workflow design where a person reviews, approves, corrects, or escalates AI output before sensitive action is taken.
- Internal Copilot — An employee-facing AI assistant that helps with research, drafting, summarization, retrieval, or coordination while a human remains in control.
Where this gets applied
- Process Documentation — Sales process, customer success playbooks, technical runbooks, financial close calendars, hiring rubrics.
- Technical Debt — Quantification in dollars, not adjectives. Then a remediation plan that runs in parallel with delivery.
- Compliance & Security — SOC 2, CMMC, FedRAMP, security baselines for post-acquisition standardization.