Skip to content
McCor Labs

Should This Be AI?

The wrong-tool test. Find out which steps of your work need AI, and which need a simple program, a checklist or a person.

How it works

  1. Name the steps of one workflow.
  2. Answer a few short questions about each step.
  3. See which tool fits each step, and the rule that decided it.

Load an example workflow

Each example replaces your steps with four answered ones.

Your workflow

Up to 8 steps. Names stay in this browser tab and are not stored.

  1. Do the information it uses and the result it produces both come in a fixed form, like spreadsheet rows or form fields, rather than emails, PDFs, photos or written paragraphs?

    Yes: an order form with the same fields every time. No: people describe what they want in an email, or the result is a written reply.

    Could you write the decision as a checklist a new hire follows without asking you?

    Yes: refunds under a set amount are approved. No: it depends on how the request reads.

    If it gets one wrong, does it cost money, a customer, or a legal problem?

    Yes: paying the wrong supplier. No: tagging an email with the wrong topic.

    Can a person check the result before it takes effect?

    Yes: a draft reply waits for approval. No: the payment goes out on its own.

    Do you have past cases where you know the right answer, say 200 or more?

    Yes: two years of quotes with the final price. No: each case is new.

    How often does it happen? (optional)

    Optional. A later version uses it to estimate cost.

    Worth a conversation

    Answer the not-sure questions to sharpen this

Your workflow at a glance

This is also what "Send us this result" sends.

Each step, its verdict, the reason and the rule that decided it
StepVerdictReasonRule
Step 1Worth a conversationWe need to know what a mistake here would cost before we can suggest a tool. Answer question 3.R1

Send us this result

Opens our contact page with these results in the message. You can edit them before sending.

Exactly what we will receive
Results from Should This Be AI? (the wrong-tool test)
Step 1: not answered yet (rule R1, rubric 2.0.0)
What I would like help with: 

How the verdicts are made

Every verdict comes from a fixed set of rules written in code. For any set of answers exactly one rule matches, so the same answers always give the same verdict. Nothing you type leaves this page unless you send it to us, or run a live test, which sends that test's text to Amazon Bedrock.

See every eval run

Rubric version 2.0.0