Skip to content

GuidesHow to2 min

How to tell AI is the wrong tool

A filter you can run in a meeting: undo-by-Friday, named job, owned source of truth. Fail any line and you are buying an incident.

Buyers ask this before they ask which model. The honest answer is a list of nos, not a capability matrix. If being slightly wrong costs money, a license, a body, or a customer you cannot get back, AI is the wrong tool — and a human has to stay on the send.

SINLE will write those nos into the SOW. Diagnosis, unsupervised legal advice, silent refunds. If a vendor will not sign that, they are the wrong partner. If the site cannot take a booking and the FAQ is empty, fix the page first. A model on top of a mess is a faster mess.

Run the filter below in a meeting. Fail any line and you are buying an incident, not a clerk.

A no you can read in the meeting

Fail any line and stop. You cannot undo it by Friday. The site or the sheet is still the leak. The truth lives in someone’s head. Nobody can turn the feature off. The vendor will not sign the nos: no diagnosis, no unsupervised legal advice, no silent refunds.

Those nos belong in the statement of work. A capability matrix is how a wrong tool gets bought. A list of nos is how it does not.

The sequence

Do these in order. Skip a line and you are back to a workshop.

  1. 01

    Ask what you cannot undo by Friday

    Money, licenses, bodies, a customer you cannot get back. Those stay human.

  2. 02

    Ask if the site or the sheet is still the leak

    A chatbot on an empty FAQ is a slower phone. A model on a racing spreadsheet is a faster mess.

  3. 03

    Ask where the truth lives

    If it only lives in someone’s head or a consumer chatbot, you are not ready.

  4. 04

    Ask who can turn it off

    If you cannot fire the feature and still run Monday, you bought a boss, not a clerk.

  5. 05

    Write the no into the SOW

    Diagnosis, unsupervised legal advice, silent refunds. If the vendor will not sign that, they are the wrong tool.

Keep these

  • Wrong on purpose: unsupervised refunds, invented hours, medicine, law.
  • Wrong because the job was never named: “AI transformation.”

Do this before you spend

  1. 01Run the five questions on the AI idea already in the room.
  2. 02If any line fails, write the no into the draft scope.
  3. 03Fix the page or the sheet before you fund a model.

Send the idea you are about to buy. We will tell you if it is a clerk or an incident.

Questions

When is AI the wrong tool?
When being slightly wrong costs money, a license, a body, or a customer you cannot get back. Draft, tag, extract — with a human on the send — can be right.
All guides

Next

Ask for the no before the model.

Spark writes where AI is the wrong tool on your job. $750. You keep the note. We will decline a brief that is only “AI transformation.”