Technical
AI Agent
An AI system that can autonomously perform tasks, make decisions, and interact with external tools and services.
An AI agent is a software system powered by AI (typically an LLM) that can take autonomous actions: reading files, calling APIs, executing code, sending messages, or modifying data. Unlike a chatbot that only generates text responses, an agent can act on the world.
Enterprise AI agents take many forms:
- Coding agents (Cursor, Claude Code) that read, write, and execute code
- Business process agents (Salesforce AgentForce, Microsoft Copilot Studio) that automate workflows
- Customer service agents that handle inquiries end-to-end
- Custom agents built by internal teams for specific use cases
Agents introduce a new category of governance challenges: they need permissions, access controls, and monitoring that go beyond what's required for passive AI tools.
Why it matters
An AI agent with access to your email can send messages on your behalf. An agent with database access can modify production data. The governance question isn't just 'is this AI safe to use?' — it's 'what can this AI do, and who authorized it?'