Free card

The Chatbot-or-Agent Test Card

The one-page test that classifies any AI task in seconds: does it need to ACT, or just ANSWER? Three questions, a worked cheat-sheet of common tasks, the fake-agent red flags, and the questions to ask before you pay for an 'agent'. So you stop overpaying for chatbots with a bigger price tag.

A 3D render of a dark glossy console with a glowing coral core, two tiles carrying the Claude mark, on warm cream, suggesting a two-sided ACT-versus-ANSWER test
Get it free →

Classify any task in seconds

One question settles it: does the task need the AI to ACT, or just ANSWER? Answer is a chatbot, act is an agent. Run it on any task before you build or buy, and you stop guessing.

Stop overpaying for chatbots

Half the 'agents' being sold are chatbots with a bigger price tag. The red-flags list and the buyer questions on the card tell you which is which, so you do not pay agent prices for chatbot work.

It is the same model underneath

The difference was never intelligence. It is whether the AI is allowed to act and keeps going until the job is done. Understand that and you spec, buy, and reject AI tools correctly.

What's inside

  • The one test: ACT vs ANSWER, the line that classifies any AI task
  • The three questions that run the test (steps, tools, outcome vs answer)
  • The printable test card, ready to pin above your desk
  • A sort-any-task list: common business tasks split into chatbot or agent
  • The fake-agent red flags: how to spot a chatbot wearing an agent price tag
  • Buyer questions: what to ask a vendor before you pay for an 'agent'

Questions

What is the actual difference?

A chatbot returns an answer and stops. An agent owns an outcome: it takes multiple steps, uses tools or takes actions in your systems, and keeps going until the job is done. Same model can power both. The difference is permission to act, not smarts.

Why do I need a card for this?

Because the word 'agent' is on everything now, and most of it is marketing. The card turns a fuzzy buzzword into a test you can run on any task in seconds, so you spec what you build and see through what you are sold.

Is this only about Claude?

The test is universal, it works for any AI tool. The examples use Claude because that is what we build on, but ACT vs ANSWER classifies a task no matter who makes the model.

Is it really free?

Yes. Drop your email and it is yours, and a copy lands in your inbox so you keep it.

Get it free. Where should we send it?