Scorecard

Is your problem a fit for AI?

Twelve questions about the problem you want to solve, not the technology. You'll get a verdict, the signs working for and against you, and what to test first.

Based on How to Tell If Your Problem Is Actually a Good Fit for AI. Takes about three minutes. Nothing is stored or sent anywhere.

1 of 12 · Unstructured inputsWhat do the inputs look like?
Clean, structured fields: rows, columns, dropdownsA mix of structured fields and free textMessy and varied: documents, emails, the same thing said fifty different ways
2 of 12 · Judgment you can't write downCould you write the decision down as rules?
Yes, a flowchart would cover almost every caseMostly, but there are plenty of exceptionsNo, experienced people just know when something looks off
3 of 12 · High volumeHow often does this decision get made?
A few times a week or a monthDozens of times a dayHundreds or thousands of times, more than people can keep up with
4 of 12 · Room for errorWhat happens when a single decision is wrong?
Serious harm, legal, safety or a lost customer, and nobody would catch itIt hurts, but someone usually catches itSmall and recoverable. The cost only adds up across many decisions
5 of 12 · Inconsistent expert judgmentDo experts agree on the right answer?
No, it's mostly taste. There isn't a right answerYes, and they apply it consistently alreadyThey agree on what good looks like, but apply it differently day to day
6 of 12 · Examples and feedback loopsDo you have past cases, and do you find out whether decisions were right?
No real examples, and we rarely find outSome examples, but outcomes aren't recorded anywherePlenty of past cases, and approvals, rejections and corrections are captured
7 of 12 · One side knows moreDoes one side of the decision know much more than the other?
No, everyone has the same informationSomewhatYes, one side writes it and the other struggles to judge it
8 of 12 · Hidden structureIs there a pattern hiding in the mess?
No, it's already well organizedPossibly, we haven't looked closelyYes, categories and trends nobody has had time to pull out
9 of 12 · Repeated thinkingWhat kind of repetitive is the work?
The same clicks and steps every timeSome judgment, but mostly routineThe same kind of thinking, applied to different inputs all day
10 of 12 · AI suggests, a person decidesCould AI make the suggestion and a person make the call?
No, it has to run with no person involvedMaybe, but nobody has time to review suggestionsYes, AI flags it and a person decides
11 of 12 · Small gains, big moneyWhat would catching a few percent more be worth?
Not much, the total is smallNoticeable, but hard to put a number onReal money, because the total spend or volume is large
12 of 12 · The outcome is the productIs a person's judgment the thing customers are paying for?
Yes, they're buying an expert's opinionPartlyNo, they're buying the outcome, not who produced it

0 of 12 answered