The Line Outside Elmer's Door (When the Whole Plan Waits on One Expert)
The Problem
The roadmap is green and yellow except for one red line stretching almost the full width of the screen. Eighteen months. Everyone assumes it's a technology problem. Then you watch for an afternoon. Every question in that part of the business routes through one person. Does this fit? Which year? Is this right? Three hundred unread messages, half of them the same question as last week from someone else. He answers maybe twenty a day. And if he got hit by a bus tomorrow, all of it would go with him. The red line on the roadmap is the length of the line outside his door.
The Principle
Codify your Brent. Write down what's in their head, in plain English, and give it to everyone. Every company has a Brent, the engineer in The Phoenix Project who is the only one who knows how anything works. He isn't the villain. He's the most valuable person in the building and the reason everything is stuck. AI makes it practical to fix that, for the first time, without six months of documentation nobody finishes. Let the expert talk while the AI listens like an apprentice and asks questions. Turn what they say into rules. Then hand the expert the red pen and make them the owner. The point is to stop them answering the same question five times a day, so they can answer the ones nobody else on earth can.
5 Steps to Take Action
- 1
Find your Elmers
Where does work pile up waiting on one person? Look at whose inbox the whole team routes through, and whose absence would stop a project.
- 2
Let them talk
Two or three hour-long sessions. The AI asks questions like an apprentice. No typing required.
- 3
Turn it into rules, then hand over the red pen
Plain-English rules, with anything uncertain flagged "ask the expert." The expert reviews every line and owns every change.
- 4
Share it as one skill
One shared set of instructions everyone uses, so the answer is the same no matter who asks.
- 5
Ask what they'd rather spend their time on
Being the one who knows is part of who they are. Help them get to the questions only they can answer.
💡 Inspired By
Chapter 5 of The Second Factory, "Elmer." And Brent, from The Phoenix Project by Gene Kim, Kevin Behr and George Spafford.
This is what Analytics Pitfalls & Principles is about: people helping people.
In Practice
ClassicMotors was building a parts business: reproductions of classic car parts that no longer exist. Everything on the plan could be sped up except one line. Catalog and fitment, which part fits which car, which years, which trim: eighteen months. The fitment data lived in Elmer Novak. Sixty-eight, reading glasses on a cord, thirty-one years behind the counter of his own parts shop in Fresno. "The '66 coupe and the fastback use different hinge pins," he told a caller. "Every catalog on earth says they're the same. They're not." His inbox had 340 unread messages, almost all from his own team. He hated typing. So Nora Berg set a laptop on his workbench, turned on the microphone and told the AI it was his apprentice: listen, ask follow-up questions, never assume. Elmer started calling it Kid. He talked for almost thirty hours in two weeks, and the AI turned it into rules. When Nora printed the first sixty-two pages, he read every one with a red pen. On page nineteen he wrote, in capitals, NO. THAT IS THE CANADIAN CAR. "If this thing's going to have my name on it," he said, "it had better be right. I want to see every change."
