Field notes and playbooks on putting AI to work in traditional, hands-on businesses — from an owner who’s actually done it. No hype. No buzzwords.
Get the Starter Pack →Start Here · Why this exists
I went looking for a filing cabinet and found a coworker.
Early in 2025 I wasn’t looking for AI. I was looking for a place to put things.
My company had SOPs scattered across drives, inboxes, and people’s heads. Safety policies we reviewed every Monday but still couldn’t put hands on when someone needed them on the job. Training material that needed updating again the moment we finished it. I needed one organized home for the documents that actually run the business. So I did what most owners do: signed up for trials, clicked through demos, imported a folder, watched it fall apart by week two. Repeat.
After the third platform, I was ready to quit on the whole category. So I did what a lot of people do now — I opened an AI chat and described the mess. Not a clever prompt. Just the actual problem: what I needed it to do, what my team would tolerate, what had already failed.
The conversation went longer than I expected, and by the end it understood the situation in a way no sales demo ever had. It pointed me toward a knowledge-base tool. I tried it. It stuck. That was my fit — not the only one, and I’ll talk honestly about the alternatives here. None of it has to be perfect. It has to be affordable, flexible, and something your people will actually use.
That was the first time AI stopped being something I visited when I had a question and became something I built on — week after week, project after project. I didn’t know it then, but I’d already found the advice I’d spend the next year proving: build the context before anything else.
That’s what this is. Everything here is built to send you back to work with something usable — the templates, the guardrails, the first-week habits — not just something to think about.
From the field
It’s a coworker, not a search bot.
This morning, a task I set up once ran on its own — it read the latest documents in our shared drive, pulled my notes out of Slack, checked where every project stood, and had a briefing ready for my 3:00 meeting before I sat down. I never asked. That’s the difference most people miss: AI isn’t a faster way to look things up — it’s a teammate that does the work while you’re busy doing yours.
(And yes — it’ll also pull last quarter’s job costs, the right SOP, or where a deal sits in the CRM in seconds. That part’s just icing.)
A publication for decision-makers in hands-on businesses — owners, operators, and the people who sign off on how the work actually gets done. Mostly service businesses: work that happens on site, in the shop, on the phone, and in the truck. If that’s your world, you’re in the right place.
I’m Mike Roberts. I own and run a corrosion and pipeline-services company — where I actually did all of this before writing a word about it. This is where I tell that story in the first person: what worked, what broke, what it cost in time and patience, and what I’d do differently today. Not theory from the sidelines — field notes from someone who signs the checks and still has to get a crew to use the thing on Monday.
This isn’t a tool-review site chasing whatever launched last Tuesday. It isn’t a consultant pitch from someone who’s never run a P&L. It isn’t a hype funnel with a checkout at the bottom. And it isn’t a sales channel for my corrosion company — that business shows up here only as proof that a hands-on, technical operation can pull this off.
If you want someone to promise one magic app will fix everything, this isn’t it. If you want the unvarnished version of what it actually takes, that’s exactly what this is.
Here’s the trap I see decision-makers fall into, because I fell into it too: they ask “What can AI do for me?” before they have anything for it to work with.
That’s the wrong first question. AI is already here. Your people are already using it — in their browsers, on their phones, on tasks you don’t know about yet. The pressure is real. But the ones who get somewhere useful don’t start by shopping for a model. They start by writing the business down well enough that a machine — or a new hire, or you at 2 a.m. — can act on it.
Knowledge first, not the tool. Run AI like a coworker with a durable memory. Treat the documentation as the product.
You don’t have an AI problem. You have a writing-things-down problem — and AI is the first coworker that won’t pretend otherwise.
Which model you pick matters less than vendors want you to believe. Once the context exists, you can use one tool or several. Swap them. Outgrow them. The stack changes. The memory shouldn’t.
Flagship essays that go deep on one idea. Shorter notes when one observation is enough. Templates and guardrails you can copy. Honest write-ups of tools I’ve actually put in front of my crew. And the Context Starter Pack — the policies, card formats, and first-week habits that took me a year of doing it wrong to simplify.
My industrial peers — pipeline, oil and gas, heavy field services — will recognize the examples. But the principles travel: a contractor running HVAC and an operator running a corrosion company are fighting the same fight — the knowledge walks out the door faster than the documentation catches up.
The fundamentals you can repeat aren’t always the ones you already know. For most of us, the gap between “we should use AI” and “we actually use it well” isn’t a model gap. It’s a context gap.
Build the memory first. Then pick your tools. Then iterate inside your own company until the thing holds up on a bad week. That’s what Field Operator is for.