Glossary
AI agent
In one sentence
An AI agent is an autonomous software system that uses a large language model as its reasoning engine, integrates with external tools and APIs to take actions, and maintains memory across interactions — distinguishing it from a stateless chatbot.
1. The three components of an AI agent
Every AI agent combines (1) a reasoning core — typically an LLM that decides what to do next, (2) a tool layer — APIs, databases, or systems the agent can read from and write to, and (3) memory — persistent context that lets the agent learn from prior interactions instead of starting fresh each time. Remove any of the three and you no longer have an agent; you have a chatbot, a script, or a search tool.
2. Agent vs chatbot
A chatbot answers questions inside a single conversation. An AI agent takes actions on the world: it can update a CRM, create a ticket, send an email, query a database, or trigger a workflow. The distinction is action, not just intelligence. A chatbot that can answer questions about Salesforce is a chatbot. One that can update a Salesforce record is an agent.
3. Agent vs workflow automation
Traditional workflow automation (Zapier, n8n, IFTTT) follows fixed rules — if this happens, do that. AI agents reason about the right action given context. The agent decides what to do, what tool to use, and whether to ask the human for confirmation. Workflow automation is deterministic; agents are decisional.
4. Single-agent vs multi-agent systems
A single-agent system has one reasoning core acting across tools. A multi-agent system has multiple specialized agents that coordinate — one might handle research, another writing, a third action. Multi-agent designs are useful when tasks split cleanly along domain lines but add coordination overhead and failure modes.
5. Why permissions matter for agents
An agent that acts on real systems needs the right access — too little and it cannot help, too much and it becomes a security risk. The mature approach mirrors each user's native permissions in each source system. The agent then operates with the exact authority the user has, no more and no less.
How this works in Cogito
Cogito is a shared AI agent for business teams. It reasons over your full stack (Notion, Linear, Slack, Intercom, and more), takes actions across those tools with human-in-the-loop approval, and inherits each user's native permissions live. The "shared" part is the differentiator — most AI agents are per-user; Cogito is one agent the whole team talks to.
Related terms
See the concept in action.
Try Cogito free with $20€20CHF 20 credit. No card required.
Get Started FreeLast reviewed