How to Pick an AI Use Case

IMAGE CREDIT: Fortiche Studios via Midjourney /imagine "an astronomer looking through the telescope at the planet Saturn, in a style of arcane anime."

Greg has led 20+ projects designing UX for AI applications. In all instances, one of the most important determinants of project success was picking the right use case to focus the AI project efforts on. Picking the wrong use case, on the other hand, was also quite often the first thing to doom the AI project. In this installment of our newsletter, we will discuss how to pick your AI use case… And how not to. 

The following story comes from the time when Greg was consulting for a precision irrigation company. This company was trying to use AI to tell farmers when and how much to water their crops so they were sufficiently watered. 

Farmers, whose fathers and forefathers tilled the very land they were caring for. 

The land these farmers knew and loved the way they loved their children.

I hope you see where we're going with this. 

When Greg did his research, he discovered that farm operators like the ones the company was trying to sell their AI solution to had no difficulty whatsoever determining adequate soil moisture levels. They did not need AI or fancy sensors, or computer models. What they used instead was a time-honored “kick test” performed exactly the way it sounds: the farmer goes into the field and kicks the heel of their boot into the soil. The feel of the soil resistance and how far the boot sinks into the soil tells them that the crops were sufficiently watered. 

And that was that.   

In fact, the whole idea that a third-generation farmer would rely on AI to tell them that their crops were sufficiently watered was insulting! 

In essence, the feeling among the potential customers was that this AI company was presuming to tell these farmers how to run their businesses. 

This is an important red flag. 

In general, whenever you are trying to pitch a human expert vs. a machine directly, it’s just not going to work. Not ever. As simple as that. 

No matter how many training data sets you have or how many rules your experts crate, there will always be some edge case you have missed. 

Natural human suspicion, pride, and prejudice of a trained expert toward a machine that would presume to tell them what to do will immediately get the way of your project, and your goose will be cooked. 

Our simple advice? Look for another use case.

Which is what Greg has done. 

Greg ASKED these very same farmers what was really keeping them up at night. And, surprise! It was not whether their crops were sufficiently watered. Instead, he found out through multiple user interviews that what really stressed these farmers out was more weighty considerations: like dwindling supplies of fresh water. Like new, much more stringent government regulations governing water use. Like Climate Change, bringing with it dry conditions that were slowly but surely turning their beloved California land into a lifeless desert.

That’s what was keeping them up at night. And those then were, in fact, the right use cases on which to laser-focus the power of AI.

It was immediately clear that the right use case was not: 

“As a farmer, I want to use AI to ensure my crops are sufficiently watered so I do not lose my harvest.”

Instead, it was:

“As a farmer, I want AI to make recommendations about which parts of the field actually need LESS water than I am currently applying so that I can save water without compromising my optimum yield, thus padding my bottom line and complying with government regulations.”

It was a very subtle difference, but it turned out to be critical:

That key insight about the right use case to focus the company’s work on was what made AI-based precision irrigation a viable play.

And in the words of Jakob Nielsen, “If you point your telescope at Saturn, you will see that it has rings” (https://www.nngroup.com/articles/banner-blindness-original-eyetracking/). Point the telescope in the right direction (ask the right question), apply sufficient magnification (interview enough people with focus, empathy, and an open mind), and you will get the same result. 

In other words,

UXers ask good questions.

Existing UX methods like contextual inquiry, field studies, user interviews, and the like, all basically come down to systematically figuring out what the customers want. Then UXers can use tools like affinity mapping and customer journey modeling to tease out those nuggets of wisdom that are solid gold to an AI project: profitable use cases, revenue improvement, new market opportunities, and so on. 

The good news is that those time-honored UX research methods, like those old telescopes made by Galileo Galilei, work just fine for AI projects – you just have to use them! 

But if you fail to ask questions, if you instead rely on the arrogant “blue ocean/red ocean” bullshit to determine your AI use cases, you might as well be trying to catch a black cat in a pitch-black room… While wearing huge welding gloves. 

No fun for you. 

No fun for the cat. 

No fun for anyone.

Now, besides running out right now and buying the best, most powerful UX telescope you can afford, you should smush that [Sign up] button below – because we’ll be covering the all-important topic of staying lean while working on AI projects in the next installment of our newsletter: “UX for AI: How to avoid searching for a black cat in a dark room, with giant welding gloves on” … Ahem, actually, our newsletter is called “UX for AI: UX Leadership in the Age of Singularity,” and we hope to see you again soon.

Peace,

Greg Nudelman with Daria Kempka