An AI agent is software that uses a large language model to choose its own next steps toward a goal. Given a task and tools, such as CRM search or an email drafter, it picks a tool, checks the result and repeats until the job is done. A workflow follows fixed steps; an agent picks them as it goes.
Also called: LLM agent, Autonomous AI agent, Agentic AI
Anthropic draws the line clearly. Workflows are "systems where LLMs and tools are orchestrated through predefined code paths". Agents are "systems where LLMs dynamically direct their own processes and tool usage". The same article advises finding the simplest solution possible and adding complexity only when needed (Anthropic, Building effective agents, 19 Dec 2024, checked 7 Nov 2026).
A new inbound lead arrives. A research agent reads the company's website, checks its pricing page, searches the CRM for past deals and writes a five-line brief for the rep with a suggested priority. Compare that with a fixed workflow that calls an enrichment service, then asks a language model to summarise the result. For many teams the workflow gives almost the same brief, costs less per run and behaves the same way every time.
Agents suit jobs that need judgement: account research, sorting replies, drafting first answers to support or sales questions. They suit rule-based jobs badly. They cost more per run, and the same input can give a different output. Buyers use agents too, to research vendors, which is one more reason your LLM visibility matters.
Platform guides: Zapier AI agents, AI workflows with n8n and Make AI agents.
Tell us which tools you use and where leads or data get stuck. We will suggest the first three automations worth building.
We reply within one business day.