The Importance of Staying Lean

If picking the wrong use case is undoubtedly the surest way to sink your AI project, the second best way is… By having too much money. Seems like a paradox, no? Read on.

A few years ago, Greg was hired to create and lead a large UX research team of 34 people at one of the world’s largest precision industrial companies.  The reason was the company’s transition into a purveyor of industrial AI. 

The team never materialized. 

Instead, in just a few weeks of joining, Greg found himself in a company that was seriously struggling to monetize the enormous investment they made into the AI system. A short year later, the company narrowly avoided bankruptcy by selling off multiple divisions and laying off more than 15,000 people. While the reasons for failure are always complex, there is no doubt that over-the-top investment into their AI system carried a lion’s portion of the blame.

The reason?

Instead of being scrappy, lean, efficient, and operating with a co-located tiger team focused on solving specific business use cases with the help of AI, the company chose to sync billions of dollars into a centralized “AI platform.” 

And, of course, having invested billions, they expected tens of billions of dollars in return. The company executives expected the AI team to deliver these sweeping transformational initiatives, patents, innovations, and so, obligingly, Marketing took over and created enormous hype about the project, with these amazing visuals, with data flying around, bits adding up to wealth, and silicon rivers flowing with gold coins. All celebrating this huge enormous success to come… Because how could it possibly be otherwise, given the coolness of AI and this giant investment they were making?

The marketing hype was a bit like watching the movie Ex Machina – sexy robots strolling around with power, majesty, and purpose, setting Jackson Pollock paintings on fire.

What the executives received from the AI Platform team was closer to what Allie Brosh describes in her brilliant book, Hyperbole and a Half as a “Simple Dog:” a dog that eats dog food, then barfs it up… and thinks it made food and therefore assumes it is now magical. 

Replace “dog food” with “data,” and you get the right idea: the best AI the team produced ingested simple training data, looked at it for a good long while, then regurgitated it back to the customer in an exceedingly simplistic way, including all warts and biases.

There was no value there because there was no use case and WAY too much money.

A much better approach is to start small and stay scrappy. 

The ideal team should consist of the following:

• Product Manager 

• UX Designer

• Engineers

• SMEs

• … and finally, a Data Scientist.

Data Science should be well-tempered by other disciplines and ideally co-located with the business, UX, engineers, and customers so that the problems of the customer and business use cases are at the forefront of the Data Scientists’ minds.

While staying lean is key, it does not mean short-changing your UX efforts – quite the opposite! UX can save you money and time and help ensure the success of your AI project. Here are some tips from Daria (Siegel) Kempka to help incorporate UX into your lean AI project:

1. Last week we discussed why it’s crucial to talk to your users to make sure you understand their problems. You need to hear -- from them -- what their pain points, needs, and goals are before you decide what kind of product to build. This is true for both AI and non-AI products. If you don’t understand the problems they face, you won’t solve the right problem. You’ll just be blindly following the latest trends straight down the rabbit hole of hype and wasted money. 

2. Are you an internet user of a certain age? (Ahem.) If so, remember those walking video overlay people that every boss wanted on every homepage in the halcyon days of Web 1.0? The ones that every user immediately dismissed? Don’t do that. Another common anti-pattern we see is companies rushing to copy each others’ (often terrible) ideas. Needless to say, this isn’t innovative. And even if the company you’re copying is an industry leader, its users are probably different from yours. It is essential to understand your users and their context to come up with truly innovative solutions they’ll be willing to pay for. 

3. Avoid pursuing product or feature ideas that are based on unexamined assumptions. Quickly test your ideas with your users to see if they’re worth pursuing. Greg covers a lightweight process for this in his book, The $1 Prototype. 

4. Also, reality test your concepts with experts and cross-functional stakeholders. You need diverse perspectives on the feasibility, desirability, and potential ethical or legal issues connected to your idea. 

5. A quick reality check for desirability is to ask yourself: is it actually easier to accomplish this thing with some existing tool? For example, a Figma plugin lets you enter text to render a design. In some cases, this is helpful. But if you’re whipping up a couple of wireframes, it’s easier just to draw a bunch of squares rather than describe their sizes, positions, and locations in the text. It’s important to understand what use cases you’re working against. 

Here’s the thing: your skills as a UX practitioner still apply. You’re still designing tools that humans need to use. That requires a deep understanding of your customers and the business, collaboration with experts and stakeholders, and a user-centric approach. Blindly following trends without taking time to ensure you’re solving a problem your users care about is a recipe for failure. As a UX designer, you’re better than that. You bring a strategic approach to generating product ideas that meet users' needs, solve real problems, and add value to the market.

In the coming weeks, we’ll explore some of the design tools you can use to speed up your workflow, as well as the ethical implications of using and designing with AI.

We hope you enjoyed this lean installment of our newsletter: “UX for AI: Staying Lean to Win,” ahem… Actually, we call it “UX for AI: UX Leadership in the Age of Singularity,” and you should hit that [Subscribe] button now because every week we bring you real-life examples and actionable advice to help you thrive in the new normal.

Peace out,

Greg Nudelman & Daria (Siegel) Kempka