The Real Version Of The Job Only Shows Up Once
The last time I wrote about why AI training fails for trades workers, I ended on a line that I knew was going to come back to me. I said that most of these failures come from building the course without anyone sitting with the actual learner in the actual workflow first, and that an afternoon of doing that changes what you build. A few people read that and asked the obvious follow-up. Okay. How do you actually run that afternoon?
Fair question. It sounds simple until you're standing in a shop next to a finishing technician who clearly wishes you weren't there. Here is what I actually do, and the first thing is that I leave the clipboard in the car.
Go To Where The Work Happens, Not A Conference Room
The first rule is that you don't do this at a table. If you pull a distributor sales rep into a meeting room and ask them to describe their day, you will get a clean, tidy version of their job that has been sanded down for an audience. It will be accurate, and it will be useless. People narrate their work as the official process, not the real one, the same way most of us would describe our morning routine as calmer and more sensible than it actually is.
So I go to the counter, or the bench, or the mixing station, and I watch the work happen in front of me. The gap between what someone tells you they do and what they actually do is the whole game. You only see it by being there while they do it.
Watch More Than You Ask
When I do sit with someone, I'm watching for the moments where they pause and decide something. A rep takes a contractor's question, thinks for a second, and pulls a product. A technician looks at a panel, looks at the spec sheet, and adjusts a ratio. Those pauses are where a tool either helps or gets in the way, and they're the only part of the job worth designing around. Most of the rest is muscle memory that no AI tool is going to improve.
This is a lot like learning a delivery route by riding shotgun with the driver who's run it for years. The printed directions get you to the right streets. Sitting beside them, you catch the small thing they do that the directions never mention: the corner they take wide because a truck is always parked there, the read they make on a loading dock before they commit to it. That judgment call is the part you could never have gotten from the written-down version, and it's the part that actually matters.
So I ask less than people expect. When I do ask, I ask about the specific thing I just watched. Why did you pull that one and not the other? What were you checking when you paused there? What happens if you get that call wrong? Those questions get real answers because they're tied to something concrete that just happened, not to an abstract version of the job.
One thing worth knowing going in: people perform a little when they know they're being watched. The first stretch of any visit, you're seeing a slightly cleaned-up version of the work, the same way you'd drive more carefully with a passenger in the car. That's why you stay. Give it long enough, and they forget you're there, fall back into the real rhythm, and start taking the shortcuts and making the quick reads they'd never have shown you in the first twenty minutes. The honest version of the job is the one you have to wait for.
The Questions That Get You Nothing
There are questions I've learned to stop asking. "Walk me through your typical day," gets you the brochure. "Where do you think AI could help?" gets you either a blank look or a guess shaped by whatever they read last week, neither of which tells you anything about their work. And "What's frustrating about your job?" makes people defensive, because it sounds like the setup to a performance review.
The worker has to feel like a guide showing you around, not a specimen under glass. The fastest way to ruin that is to take notes the moment they say something that confirms what you already suspected. People notice. So I keep the notebook mostly closed and write things down after, in the car, while it's still fresh.
It Takes An Afternoon, And That's The Point
People ask how long this takes, hoping I'll say it's a quick thing they can skip. It's usually one good afternoon per role you're designing for. You won't get to every role on the org chart, and pretending otherwise just sets you up to do all of them badly. So you make a call: you spend the afternoons on the two or three roles where the tool decision carries the most weight, and you accept that you'll know the rest less well. I'd rather be honest about that than imply the lessons from one bench transfer cleanly to another. They don't, which is the whole reason you go sit down in the first place. You're not running a research project or a focus group or convening a survey that comes back as a spreadsheet of averages. You're spending one afternoon watching one person who does the actual job, with your mouth mostly shut.
What you walk away with is the two or three decision points where a tool would genuinely help, and the handful of places where it would just slow a skilled person down. Build the module around those, and you're designing for the day a worker actually has. Skip it, and you're guessing, and your completion rate will be the only thing that looks good.
