AI Agent

Also called AI Agents and Autonomous Agent

Software in which a language model decides its own next steps, choosing which tools to use, acting, checking the result and carrying on until a goal is met, rather than following a fixed script.

Why it matters

Agents can take on whole tasks, such as researching leads, resolving support tickets or fixing code, which changes what a small team can do and what software can be sold as. They also cost more per task, can go wrong in ways that are hard to predict, and need clear limits on what they are allowed to touch.

For example

A support agent reads a customer's complaint, looks up the order, issues a refund within a set limit, and hands anything above that limit to a person.

Go deeper

Building effective agents Anthropic · guide Anthropic's guide to building with language models, which separates agents, where the model directs its own process, from workflows that follow set steps, and argues for starting with the simplest design that works. Open anthropic.com →

Related terms

Agentic Workflow Large Language Model Model Context Protocol Outcome-Based Pricing

Also in Starting Up

See how founders actually handle this on Building the product, part of the Starting Up hub.

eChai Partner Brands The partnership →