I listen to podcasts about AI all the time. So when I get pitched AI guests for this show, I’m usually just not in the mood.
But I’m always in the mood to talk AI with my friend Damon Pistulka. He’s different from so much of the AI noise out there and he’s just fun to talk to. Damon is a farm kid turned mechanical engineer who’s spent his whole career in manufacturing. Since 2010 he’s helped small manufacturers grow and sell their businesses. These days he’s busy showing small manufacturers how to make AI do the stuff they don’t know how to do, or just don’t want to do. Inventory counts. Dashboards. You know, useful, life-changing stuff.
Link to Graff-Pinkert’s Acquisitions and Sales promotion!
*************
Listen on your favorite podcast app using pod.link.
Facebook: https://www.facebook.com/swarfcast
Instagram: https://www.instagram.com/swarfcast/
LinkedIn: https://www.linkedin.com/company/todays-machining-world
Twitter: https://twitter.com/tmwswarfblog
Interview Highlights
Start With the Pain, Not the App
Every day there’s a new AI app promising to fix your business. Damon doesn’t start there. He looks for the job that’s a pain in the butt, eats up time, and that only one person in the company knows how to do. That’s where AI goes first. Solve one problem, then the next. Crawl, walk, run.
Connect AI to Your Systems
Take ABC cycle counting. Deciding which parts to count most used to take real analysis. Now you can connect AI to your inventory system and ask which parts move the most and which have the most count adjustments. A slow-moving part with a lot of adjustments might be critical, and it deserves your attention.
If you’re buying new software, make sure it has an API. That’s the connection that lets two programs talk to each other, including your AI. With it, you can tell AI to pull a report from your ERP, build a spreadsheet the way you want it, and ask you questions first so it gets it right. Once you’ve checked the results, have it run the report every Monday morning.
You can check whether part dimensions vary more in summer than in winter. You can build real-time dashboards for uptime, shipping and sales, so each department doesn’t have to put its numbers together every week. The bigger payoff isn’t the time saved. It’s that you can make decisions faster.
Teach It What Good Looks Like
I lamented to Damon that before our interview I had spent an hour and a half perfecting an email blast for used machinery. Damon’s advice was to ask AI to study the past blasts I thought were good, pull out what made them work, and then build the new one. If you just ask AI to write one, it makes up what good looks like.
Damon also has AI keep a knowledge base about him and his business. He spent a couple of hours letting it interview him, and he adds stories as he remembers them. He also set a rule that it saves what it learns after every task. One result is an email drafter connected to Outlook that he uses 20 to 30 times a day.
Tag-Team the Brainstorm
I brought up advice from Ethan Mollick, a Penn business professor who writes books about AI, to think on your own for ten minutes before going to AI. It’s the same reason group brainstorming falls short: people come up with better ideas alone first. Damon works this way too. He writes down his own ideas, then asks AI how it would solve the problem. He asks why it chose that answer and has it look from another angle. Then he feeds in his own ideas and asks how to combine the two. One plus one equals three.
If you show it your answer too early, it’ll tell you that’s a great idea, and you’ll end up two steps short. People tend to reach for solutions they already know. AI can suggest a whole different process. One example is hooking up a sensor instead of writing down a tank’s temperature by hand every day.
When It Has to Be Right, Use Code
AI still hallucinates. Damon wouldn’t let it cut into a $100,000 titanium casting unless a person had checked the program. He also wants an AI working on accounting data to check its own work and produce a spreadsheet that people can review.
A year ago he couldn’t write code. Now AI writes programs for him four to six hours a day. Code either works or it doesn’t, so it can’t hallucinate. His approach is to build a dashboard in AI first, then turn it into a program that runs on its own. Code can run 90 percent of a process, with AI handling only a small piece.
More Time for People
Damon says AI has him talking to more people, not fewer. To automate a process, you have to know it really well, and that means talking to the people who do it. He expects companies to hire fewer people as workers retire or leave. But the goal is to free people to do what they’re good at, including building relationships.
Question: What’s one task in your business you wish you could have AI do for you?
You can find Damon on LinkedIn, where he writes a newsletter called AI Without the Hype.

