Design for Machine Learning

How Designers Can Catch the ML Wave that’s Changing Everything.

Picture this. It’s Wednesday afternoon, and we’re ideating new product offerings to solve customer problems we uncovered. As usual, ideas like an “AI-enabled personal assistant” or “automatic ___ powered by AI” finds their way into the hodgepodge of ideas.

Although ideation best practices forbids early judgment, I can’t help but notice how frequently the AI ideas get prioritized without anything asking, “what exactly do we mean and how is it ever going to work?”

Sometimes, especially when data-science is not included, the team spends time validating the desirability of a vaguely defined “AI solution” and building a business case to sell the idea. By the time the lack of feasibility is finally recognized, a lot of time have already been wasted working on the now-invalidated idea; time that could’ve been used more productively elsewhere.

My reaction: 🤦🤦🤦

At least 🤦 is how I feel on the inside. Remembering the advice I received from one of my early mentors,

The best way to complain is to do something.

I realized that complaining about the AI hype’s influence on our creative process alone wasn’t going to be helpful. I needed to get a better understanding of how AI (machine learning) works and how I can harness it as a designer.

I also noticed a big knowledge gap between those good at understanding problems (designers) and those good at building solutions (ML engineers). There was room to help the two disciplines better understand and work with each other. I thought, why not step up and be the bridge?

Given how complex this subject is, observing from the sidelines didn’t get me very far. So I spent months immersed in a cutting edge ML company named Dessa to lead US business development (update: we just got acquired by Square in Feb 2020). My goal for this adventure: to learn about the state of the art in ML and the team building it. My takeaways below.

A quick definition of machine learning

What are AI and machine learning? This question has already been answered too many times that I’ll just…