Studio
Black Friday

Up to 70% off kits30% off services

Ends in
All Tools
AI product · Free scorecard

AI Use Case Scorecard

Got a list of AI feature ideas for your product? Rate each one on six questions and see which to build first, which to prototype, which are cheaper as plain software and which to park. Four sample ideas are filled in; rename and rescore them, or start blank.

  1. Use. How often would a typical user use it?

    1 = A few times a year5 = Several times a week

    Value. How much time or effort does it save them each time?

    1 = A few seconds5 = Many minutes, or a task they skip today

    AI fit. Does it need reading, writing or judging messy input like text, documents or images?

    1 = No: clean data and fixed rules5 = Yes: that is the whole job

    Data. Do you have the content or data it needs today, in one place?

    1 = No, it would have to be created5 = Yes, clean and current

    Risk. How easy is a wrong answer to spot and fix?

    1 = Hard to spot, costly or public5 = Obvious and harmless

    Cost. Would the model cost per use stay small next to what the user pays you?

    1 = It could eat the margin5 = A tiny share of the price

    87/ 100Build first

  2. Use. How often would a typical user use it?

    1 = A few times a year5 = Several times a week

    Value. How much time or effort does it save them each time?

    1 = A few seconds5 = Many minutes, or a task they skip today

    AI fit. Does it need reading, writing or judging messy input like text, documents or images?

    1 = No: clean data and fixed rules5 = Yes: that is the whole job

    Data. Do you have the content or data it needs today, in one place?

    1 = No, it would have to be created5 = Yes, clean and current

    Risk. How easy is a wrong answer to spot and fix?

    1 = Hard to spot, costly or public5 = Obvious and harmless

    Cost. Would the model cost per use stay small next to what the user pays you?

    1 = It could eat the margin5 = A tiny share of the price

    69/ 100Prototype and test

  3. Use. How often would a typical user use it?

    1 = A few times a year5 = Several times a week

    Value. How much time or effort does it save them each time?

    1 = A few seconds5 = Many minutes, or a task they skip today

    AI fit. Does it need reading, writing or judging messy input like text, documents or images?

    1 = No: clean data and fixed rules5 = Yes: that is the whole job

    Data. Do you have the content or data it needs today, in one place?

    1 = No, it would have to be created5 = Yes, clean and current

    Risk. How easy is a wrong answer to spot and fix?

    1 = Hard to spot, costly or public5 = Obvious and harmless

    Cost. Would the model cost per use stay small next to what the user pays you?

    1 = It could eat the margin5 = A tiny share of the price

    65/ 100Plain software

  4. Use. How often would a typical user use it?

    1 = A few times a year5 = Several times a week

    Value. How much time or effort does it save them each time?

    1 = A few seconds5 = Many minutes, or a task they skip today

    AI fit. Does it need reading, writing or judging messy input like text, documents or images?

    1 = No: clean data and fixed rules5 = Yes: that is the whole job

    Data. Do you have the content or data it needs today, in one place?

    1 = No, it would have to be created5 = Yes, clean and current

    Risk. How easy is a wrong answer to spot and fix?

    1 = Hard to spot, costly or public5 = Obvious and harmless

    Cost. Would the model cost per use stay small next to what the user pays you?

    1 = It could eat the margin5 = A tiny share of the price

    48/ 100Park it

Your ranking

  1. 1

    Answer "how do I" questions from our help docs

    87/100Build first

    Frequent, valuable and a good fit. Test it on 30 real examples, then build it.

  2. 2

    Summarize an account before a renewal call

    69/100Prototype and test

    Promising but not proven. A prototype and a small test set will show whether it holds up.

  3. 3

    Show a "top customers" chart on the dashboard

    65/100Plain software

    Nothing here needs a model. Build it with regular code: it will be cheaper, faster and always right.

  4. 4

    Approve refund requests automatically

    48/100Park it

    Mistakes would be costly and hard to catch. Let AI gather facts for a person instead of deciding.

Next, for your top idea

Test it on 30 real examples with the LLM evaluation template, then price it per user with the AI app cost calculator.

Want a second opinion on your list?30% OFF

Our $349$499 AI Feature Discovery scores up to 5 ideas with your team, tests the best one on 30 real examples with two models, and ends with a prototype and a fixed build price. The fee comes off the build.

How to pick the first AI feature for your product

Lists of AI use cases by industry are a fine place to get ideas. They can't tell you which one fits your users, your data and your margins. These six questions can.

  1. Use: How often would a typical user use it? Features people use weekly are the ones that build the habit.
  2. Value: How much time or effort does it save them each time? Small savings on rare tasks don't change how people feel about your product.
  3. AI fit: Does it need reading, writing or judging messy input like text, documents or images? If not, it is a regular software feature, and that is good news: cheaper and always right.
  4. Data: Do you have the content or data it needs today, in one place? Assistants are only as good as the content they answer from.
  5. Risk: How easy is a wrong answer to spot and fix? Features that show their source and let the user check are far safer to start with than ones that act alone.
  6. Cost: Would the model cost per use stay small next to what the user pays you? Most features cost cents per user a month, but long documents, images and heavy users can change that.

Frequently asked questions

Good AI use cases, the scoring, plain software and what to do next.