The actual version of the job is displayed only once
Last time I wrote about why AI training fails for trade workers, I ended with a sentence I knew would come back to me. I said that most of these failures come from building the course without anyone sitting down with real learners in a real workflow first, and doing that in an afternoon changes what you build. Several people read it and asked for an obvious follow-up. Understood. How do you actually run that afternoon?
Fair question. It seems like a no-brainer, but when you’re standing next to a finishing technician in the store, you obviously wish you weren’t there. Here’s what I actually do: First, I leave my clipboard in the car.
Go where the work is done, not the conference room
The first rule is don’t do this at the table. If you bring distributor salespeople into a conference room and ask them to explain their day’s work, you’ll get a stripped-down, clean-cut version of their work for your audience. That would be accurate, but not useful. Just as people talk about their work as a formal process rather than an actual process, most of us describe our morning routines as calmer and more sensible than they actually are.
So I go to the counter, bench, or mixing station and watch the work being done right in front of me. The gap between what someone says they are doing and what they actually do affects the entire game. You can only see it by being there while they are doing it.
look beyond what is asked
When I sit down with someone, I observe the moment when they stop and make a decision. The representative answers the contractor’s questions, thinks for a moment, and pulls out the product. The technician looks at the panel, looks at the spec sheet, and adjusts the ratios. Such pauses are the only parts of the work that are worth designing for, where the tool either helps or gets in the way. Most of the rest is muscle memory and cannot be improved by AI tools.
It’s a lot like riding shotgun with a driver who’s been driving for years and learning delivery routes. You can follow the printed directions to get to the right street. Sitting next to them, you notice the little things they do that are never mentioned in the directions. Things like widening corners because trucks are always parked there, and the reads they do before they start work at the loading dock. That judgment is the part that you can never get from a written version, and that is the part that really matters.
So I ask fewer questions than people would expect. When asking questions, ask about specific things you just saw. Why did you pull one and not the other? What were you checking when you stopped there? What if I make a mistake on that call? These questions will give you real answers because they are tied to the concrete thing that just happened, rather than an abstract version of the job.
One thing to know when participating is that people perform worse when they know they are being watched. For the first part of your visit, you’ll see a slightly cleaned-up version of the piece, similar to driving carefully with a passenger in your car. That’s why you stay here. If you go on long enough, they’ll forget you’re there and get back into their real rhythm, taking shortcuts or speed reading that they never showed in the first 20 minutes. The honest version of the job is what you have to wait for.
Questions that don’t yield anything
There are questions I’ve learned to stop asking. You receive a pamphlet that asks you to describe your typical day. “Where do you think AI could be useful?” A guess formed from what they read last week will either give you a blank look or tell you nothing about their work either. and “What frustrates you at work?” It makes people defensive because it sounds like preparation for a performance review.
Workers must feel like guides, not specimens under glass. The quickest way to ruin it is to take notes the moment they say something that confirms what you already suspected. people notice. So I mostly keep my notebook closed and then write it down in the car while it’s still fresh.
It takes an afternoon, that’s the point.
People ask how long this takes and expect me to answer that it’s a simple task that can be skipped. Typically, it will be one afternoon for each role being designed. Not every role on the org chart can be filled, and pretending otherwise will only lead to a poor performance in every role. So you make a call. Spend the afternoon working on two or three roles where tooling decisions are the most important, and accept that you don’t know much about the rest. I’d rather be honest about this than imply that lessons learned from one bench transfer neatly to another. They don’t, which is the whole reason you sit down in the first place. We’re not running research projects or focus groups or convening surveys that come back as spreadsheets of averages. You end up spending an afternoon watching one person do the actual work, mostly silent.
There are only two or three decision points where a tool is really useful, and only a few places where it only slows down the skilled person. When you build modules around these, you’re designing around the actual day your employees spend. If you skip it, you’ll be making guesses and only your completion percentage will look good.
Share with
