I’m of two minds regarding Artificial Intelligence. One: I don’t like today’s generative AI tools like ChatGPT and Gemini and others, but I’m finding ways to use them. You can read about that on p. 8, though I don’t even get into the “huge problem” aspect of AI’s electricity needs or the likely investment bubble. Two: I assume that AI agents and machine learning technology have potential to improve people’s lives and businesses. Because I don’t know much about this side of AI, I spoke with experts who do.
Sean McLaughlin owns Craig Taylor Equipment, a dealer with four locations in Alaska. He also created Flyntlok, a cloud-based DMS for OPE and tractor dealers, following earlier success in software development. I turned to McLaughlin to learn what AI might do for power-equipment dealers.
OPE+: I want to help dealers know how they can use AI. I have a little bit of an idea, but I really don’t know anything about this type of AI, so help me understand.
Sean McLaughlin: Since I last talked to you, which was literally a year plus ago, it’s taken shape. Back then, AI was this conceptual thing that we were chatting and spitballing on; now we’re going to be releasing an AI tool we’re calling Yosemite Sam. Every AI you kind of give a persona to. We’re coming out with Yosemite Sam in the second half of this year.
What we’re working on for customers falls into two camps. One that’s sexier and one that’s less sexy. I tend to like the less sexy one. The sexier one is what people want to talk about. Basically there’s a revenue side and what AI can do to help you grow revenues. And then there’s the cost-cutting side of what AI can do.
On the cost side, it comes down to the use cases; this isn’t guesswork anymore. The first one is reconciliation and automation of part receiving. Right now receiving parts is kind of a pain in the ass. You order them, the parts come in and you’re verifying all the pricing, you’re verifying everything. You get it into your DMS, and then you pay the vendor. In the new world, you’re going to take your invoice and scan it and drop it in AI Flyntlok. And we’re going to pull all the data off there about invoice number and when it’s due and what parts were received and what their cost was and what the fees were. And we’re going to auto-reconcile it against what we know we’re receiving. We’re just going to do the whole thing automatically for you and show it to you and you’re going to bless it and go. It’s going to save something like 70 percent of parts receiving time; it already works.
OPE+: You said, “scan the invoice” and then it’ll do this and this and this. Do you literally mean the dealer has to take paper and scan it?
McLaughlin: Great question. In almost everything in a dealership now, they’re getting the invoice via email or a portal. They grab the PDF out of one app and drag it into Yosemite Sam. And the AI engine parses out that PDF. It builds the left-hand side of what’s being received. The right-hand side is what it knows is going to be happening and it just starts automatically matching everything and saying, you’re good.
OPE+: Does the PDF need to be formatted in a very specific way or it can read a PDF from Deere or a PDF from Bobcat?
McLaughlin: It’s even better because as you learn, it’s learning your session too. So it’s learning different dealers. It’s learning every time it makes a mistake. It will also look for fraud. If you put in an invoice that doesn’t make any sense, we’re looking for fraud. Fraud’s a big thing in dealers, people trying to get paid for parts they didn’t ever send. We’ll do a fraud scoring on the PDF. This one is a pure time saver. It doesn’t create more revenue.
OPE+: Okay, talk more about the second one, for service, right?
McLaughlin: Yes. Everything in the dealership on the service side is “complaint-cause-correction.” That’s the basic math of a service ticket or a repair order. As soon as you put in a complaint, we’re going to recommend to you what we think the cause and corrections are and the parts you need. And you’re going to say we did a ‘thumbs up’ job. Or you’ll say we did a ‘shit job’ and that’s going to retrain the model. It’s going to start just giving you the cause and corrections and parts you need like right off the bat. And we’ll write them all for you.
The only reason we know what to recommend on a complaint-cause-correction is that the data from our vendors is being fed into an AI engine of complaint-cause-corrections that are complete. So we feed it all in, and kind of turn into an English sentence basically. Then the next complaint comes in and we go, ‘Does this look like anything that any of our 200-plus customers have done? Is it the same machine model make?’
Then it’s pulling back and saying, ‘All right, this would be my recommended cause-correction. And these would be the parts to fix it. Before the only way to do that was you had to build that manually. Then the tech would pick it. And then you could have it all predefined. The thing is, everything has to be tested.
(McLaughlin entered prompts into an AI agent as we talked)
I just wrote, “My lawnmower is running rough. What should I do?” So the generic Gemini is giving me a generic Gemini answer. It’s talking about fuel, air filter, spark plug. Those are the three things you get taught, like in second grade.
Well, if it actually knew complaint-cause-correction and it knew the type of mower, what could it do? Again, if I just do this and say a Toro, but what should I do? I have all this data exhaust that gets really granular about this exact problem. It just gets smarter and smarter.
OPE+: Are these things a dealer can do right now? Without a Flyntlok DMS account? They can start to use ChatGPT or Gemini and do some work?
McLaughlin: So the easiest thing to do, the organic thing we see is people building their own AI chatbots. They go to ChatGPT, they build their own bot, they give it a purpose, and then they try to use it to solve some problem in their organization.
Here’s one I’ve seen. This is actually a pretty good one. You work in the front counter and you’re a brand-new employee. You sell Toro, Kubota, and Grasshopper. The customer comes in and asks, ‘Tell me what’s different about these mowers?’
A lot of them will put up Chat GPT on the right and they’ve got their point-of-sale on the left. If a customer comes in and they’re just like blank stare, I have no idea what to say, you know, for the new employee, they’ll go ask Chat GPT which will give them a voluminous answer about what the differences are. Then they can even do follow-ups. Every session it gets smarter and smarter. You can say, ‘The customer has one acre of grass and it’s hilly.’ You can get answers that make you sound, at point of sale, like you know what you’re doing when you really don’t.
OPE+: I’m less optimistic about AI, mainly because I’ve seen far too many marketing people use generative AI as only a time saver. You’re a unique case as both a dealer and software expert.
McLaughlin: The velocity of that change in our world is astounding. It’s absolutely astounding. At the end of the day, it’s going to help this industry. Humans still have to fix everything. But if AI can pull minutes and hours out of your day, it does add up. And it’s not going away. It’s changed how every executive looks at what they’re doing next in their business.
This is a brief and edited story from a longer interview with McLaughlin and his teammate Mike Wasserman, chief revenue officer for Flyntlok.