Engineers have long split their problems into known unknowns (gaps you can see) and unknown unknowns (gaps you can’t). Unk-unks for short.
AI will bite you in your unk-unks. So before you build anything with it, turn them into known unknowns.
I was reminded of this the hard way recently.
Focus on what you know and you'll miss those unknown unknowns lurking to the side
My boyfriend and I are in the middle of a house build. We live in Florida, where hurricane season peaks in September and October. In other words, Floridians brace. As it happens, our house build hit "peak bracing" at the same time this year. Decisions-per-day: coming with hurricane-force speed.
So I took a few weeks off work to address the urgent matters of the month, beginning with the design of our bathrooms. We needed to know where the lights would go, so *I* needed to know where everything would go to be sure we put the lights in the right places.
I asked ChatGPT to help me visualize the rooms. I was quite proud of the flow I created. From framed room to "blank slate" room to designed room.
Framed room -> Blank slate room -> Designed room
That designed room looks so cool, doesn't it? Someone said it’s giving "modern Golden Girls." I love that (but don’t tell my boyfriend...living in Blanche’s dream bathroom may not be his ideal).
8am, first day back at work. I'm just "grabbing the measurements" for the electrician, who is already on site.
And... that's when it hits me: The mirror and the lights in real life would look like this.
Sconces on mirror? Maybe Blanche's vibe, but not mine.
Fail.
Cue panicked redesign, tossing out careful plans, and guessing-while-stressed.
My unk-unk was a render gap. I’d designed on a blank slate AI made up, with fictitious measurements. AI rendered the room I imagined. The electrician needed the room I actually had.
That’s a bathroom. I’m also building a new digital product, where the same mistake would cost a lot more than a last-minute electrical change.
With AI, we can reinvent our businesses and build convincing new products in an afternoon. We can also build them on a hallucinated blank slate, and AI won’t mention that it bent reality to get there.
Customers live in the framed room. They'll happily upgrade to the designed room, if it's not hallucinated. In my bathroom, the constraints were measurements. In your business, they’re your customers’ needs and behaviors.
So before you reinvent your business with AI or build that new product, here’s your light action:
Take a week and talk to 6-8 customers. Ask them to walk you through how they work today, especially the part your product would change. These are curiosity conversations. Notice how your customers function in that space. Start with them. Your product can wait.
Now some of your unknown unknowns are known unknowns.
You have your measurements. Now it’s time to build.
A personal newsletter from Amy Bonsall, founder of Light Actions. Light Actions helps companies use AI to grow: new offerings, new ventures, the next business. But light actions (thankfully) work anywhere you're uncertain.
I'm Amy Bonsall, ambiguity architect and the on-ramp for leaders ready to move further with AI. My superpower: helping leaders make heavy things light and fuzzy things clear. Former IDEO and Old Navy exec, Harvard Business Review author, and secret-back-pocket resource for leaders who want to do more (with AI) while working less.