Why the vocabulary matters commercially
These three words describe systems with different capabilities, risks, and prices. When a proposal uses them loosely you cannot tell what is being sold, and the gap usually appears after signing.
Chatbot: answers questions
A conversational interface over information. It retrieves, it explains, it does not change anything. The risk is a wrong answer, which is why it belongs where information is documented and where escalation is available.
Copilot: assists a person doing work
Embedded in a tool, drafting and suggesting while a human stays in control and accountable. This is the highest-value, lowest-risk category for most businesses, and the reason draft-and-edit is among the first jobs worth handing over.
A copilot suggests and a human commits. An agent commits. That single difference sets the entire risk profile.
Agent: takes actions on its own
It decides what to do and does it — updating records, sending messages, calling systems. Powerful, and it needs permissions, guardrails, and audit logging that the other two don’t, which is why it is usually the wrong default.
Questions that cut through the label
Ask what it can change, what happens when it is wrong, what it is allowed to read, and who reviews its output. The answers tell you which of the three you are actually buying regardless of the word on the slide.
Most businesses want a copilot
The demand is usually for speed on work people already do, not autonomy. Copilots deliver that with a fraction of the governance burden, and they generate the correction data that would justify anything more autonomous later.