Member-only story
Where the line falls between an assistant, an automation and an agent
Most of what companies call agents today are something else, and the difference decides who is accountable when it goes wrong.
Last week I spent an afternoon with the leadership team of a company that is well ahead on AI. People across the business use it every day, and the team is now planning a workforce of people and AI agents working side by side.
They already run all three kinds of AI system. The whole company opens a chat and search tool every morning. Scripts move files between systems. A newer system reads incoming emails and acts on them. Because they move fast, they had been calling all of it “agents.”
Each of those systems has its own cost, risk and owner. One of the leaders wanted words precise enough for every team to ask for exactly what it needs as they scale.
So we drew the line.
Three words for three different jobs
An assistant works one-on-one with a person. It drafts, summarizes, searches and answers. Most companies already own one. A person reads what it produces and decides what happens next.
