The theme of the summer has been “Automate Last.” We kicked off June with Jon McNeill talking about his new book, The Algorithm. Step 5 of The Algorithm is “Automate Last” which means don’t jump to AI to solve a complex problem.
“Don’t automate until you know exactly what you’re doing— every motion, every turn, the pathway of every component.” (p. 106)
You learn from your mistakes, right? As much as we’d like to say we took Jon’s advice, we didn’t initially.
We’ve been trying to organize our HubSpot. Thousands of contacts, each with notes and characteristics we want to analyze, and far too much information to sort through on our own. So we went to Claude: “Can you organize these contacts by… and list them in… and…”
Claude tried. It gave us a glimpse of hope for what could be.
So we went back. “Can you change… and make it look like…”
One hour later.
That’s when the lightbulb went on. We needed to do the organizing ourselves to find the pain point. We spent two hours manually organizing the list to our liking, and finally understood what needed to be done. Rather than asking Claude what we should do, we told Claude exactly what we needed and how to get there. We then got the result we wanted.
Jon shares a prime example about DoorDash in his book:
“Rather than building an ordering system for multiple restaurants, it went manual instead. The three founders scanned menus from local restaurants within three miles of their dorm. They put those menus up on a website with a Google Voice phone number to order. As calls came in, they would fulfill orders by hand to learn the restaurant ordering and delivery process. Only after studying it and optimizing it did they begin to automate” (p. 107).
So what’s the lesson? Do the work. Before you prompt AI to build something, do the task manually once and map every step. That turns your story into their instruction.
For a deeper dive, you can listen to Jon’s panel or read the book.
-Nicole
