Notes from the workshop / 2 min read

AI, Automation, and Agents: What Do You Actually Need?

Three illustrated examples show AI drafting a reply, automation creating a follow-up reminder, and an agent finding information and preparing a reply.

If every technology conversation sounds like a new vocabulary test, you're not alone. AI, automation, and agents often get mentioned together. Understanding the distinction helps you ask for the right kind of help.

Let's use one situation: a potential customer asks about your services.

AI helps work with information

An AI tool might help turn your notes into a clear explanation or draft a response using information you provide. You review the wording, correct details, and decide whether to use it.

That's useful assistance. It doesn't automatically mean the tool can access your inbox, know your latest prices, or send anything for you.

Automation follows a repeatable process

An automation might record a new inquiry in a tracking sheet and create a reminder. The trigger and steps are defined in advance: when this happens, do that.

Automation doesn't always need AI. A straightforward reminder may work well with ordinary rules. Before building it, you need to decide what should happen when required information is missing or a step fails.

Agents can work through tasks using tools

“Agent” is used differently across products. In a common technical distinction, an agent can decide which steps or tools to use as it works toward a goal, while a workflow follows predefined paths.

For a customer inquiry, an agent might retrieve approved information and prepare a response. Its actual abilities depend on the connected tools, permissions, and design. The name alone doesn't tell you what it can do.

A useful way to choose

Ask what you need first:

  • Help shaping information? Start with an AI-assisted draft.
  • The same steps repeated consistently? Consider a simple automation.
  • A task that needs different decisions or tools? Explore an agent, with clear boundaries.

These approaches can also work together. But adding more complexity isn't automatically an improvement.

Before you choose anything, write one sentence describing the result: “Every new inquiry should have an owner and a next step.” That gives you a clearer target than “We need an AI agent.”

Save this guide for your next technology conversation—and start with the job you want done.

Editorial source: Anthropic: Building effective agents. The terminology distinction is paraphrased; tool details are not recommendations. Verified October 11, 2026.