Skip to content

Free checklist · by Eromonsele Oboh

15 questions to ask before you buy an AI tool

Send these to the vendor before you sign, and ask for the answers in writing. If they can't answer clearly, that tells you something too.

Does it solve your problem?

  1. Can we run a paid pilot on our own data before committing?

    Good answer: yes, for a set number of weeks, with success criteria we agree on in writing.

  2. What results have customers like us measured? Can we speak to one?
  3. What will it cost in year one, including setup, training, and usage fees?

Your data

  1. Where is our data stored?

    Good answer: a named country or region. Canada, if you need it.

  2. Do you use our data to train your AI models?

    Good answer: no, and it's in the contract.

  3. Which other companies can access our data?
  4. How long do you keep our data, and will you delete it when we ask?

Security

  1. Do you have an independent security report, such as SOC 2 or ISO 27001? Can we see it?
  2. Can we control who on our team has access, and use our own login system?
  3. If there's a breach, how fast will you tell us?

When the AI gets it wrong

  1. How often is it wrong, and how did you measure that?
  2. What happens when it's wrong? Can a person review or override it?
  3. Will you tell us before you change the model or how it behaves?

Leaving

  1. Can we export all our data, in a standard format, at any time?
  2. What does it cost, and how long does it take, to leave?

In Quebec: if the tool handles personal information, you need a privacy impact assessment before you buy it, and another if the data leaves Quebec. Questions 4 to 10 give you most of the technical facts for it. This checklist is not legal advice.

Want a second opinion?

I'll review the tool for you and give you a clear answer, plus the list of conditions to send the vendor.