Diesel Goes Digital: AI in Heavy-Duty Repair Shops

AI is everywhere.

Seriously. Head to a news site (or a newspaper, if you’re old school) and you’ll see some story about the latest artificial intelligence (AI) advancement. Wander over to social media and you’ll find a blitz of AI-generated images and a mixture of people praising and denouncing it. Everyone’s got an opinion on how AI is impacting small businesses or turning the world against the em dash.

(Editor’s Note: Suz is very very bitter about the em dash.)

In the nearly three years since ChatGPT exploded into the public consciousness, white-collar work has gotten something of a crash course in how AI functions and how it can assist (or not) in their roles. The trades have, at least in theory, largely been left alone; AI can do a lot of stuff, but it certainly can’t put an engine back together or even really reliably drive long distances (they are working on that, though). 

Fullbay was built around the idea of making life easier for heavy-duty repair folks, so naturally, we’ve been curious about AI and how it might eventually make its way into the industry. So the marketing team followed its usual pattern: we guzzled a lot of coffee and started bugging other departments and a few customers to get their take on AI. 

In a nutshell: while everyone has a slightly different opinion on AI and what it can/should do, the industry as a whole is very interested in anything that can minimize or take over repetitive tasks and free people up for more meaningful work, like actual repairs or customer service. Pretty much everyone we’ve talked to is well aware that we are years away from directing robot apprentices to do oil changes or clean up the parts room, but there are other parts of the job a solid AI platform could assist with or even eventually manage entirely.

So yes, we’re going to discuss that. But first, let’s set a baseline for what we’re talking about when we say “AI” (hint: it ain’t Skynet…yet). 

WHAT IS AI, AND WHAT DOES IT DO?

AI stands for artificial intelligence, but generally speaking, our current crop of AIs are not AIs at all — they’ve just been labeled that because it’s easy to understand. To be clear, we are not referring to a Hal or a T-1000 or even an Agent Smith. ChatGPT, Gemini, Grok, and other “AI platforms” are in fact Large Language Models, or LLMs. They are gigantic computers. They do not “think,” per se; when asked a question, they consult the ginormous amount of data within them and come up with an answer. 

In the words of this writer’s free ChatGPT account, these machines are “giant pattern-matching engines.” 

This is important enough to repeat: They are not thinking. They are predicting patterns. And for the purposes of this article, we’ll be using “AI” and “LLM” interchangeably, even though they are technically the latter and not the former; everyone’s calling them AI, so we’re rolling with it.

How Large Language Models Work, in Plain English

How does it work, exactly? Let’s say you have a tech that has read (and memorized) every repair manual, every shop forum thread, and every snide social media comment about trucks. Ever. Over and over again. So, if you ask that tech, “Why is my truck blowing white smoke?” the tech will not look it up. Instead, they’ll remember the patterns they’ve seen in the past:

• Mentions of white smoke are often accompanied by discussion of coolant.
• Other mentions of white smoke may be accompanied by a conversation about Duramax engines (this is straight from ChatGPT: “When it’s early morning and cold, and a Duramax is involved, there’s a different pattern.”) 

The AI will then essentially generate an answer that just sounds right based on all the data it’s internalized.

That’s a verrrrrrrrrrry basic explanation of how LLMs work. When you ask one of these platforms a question, it analyzes what you’ve said or written and predicts the next likeliest word, one after the other, until it has an answer, or a diagnostic, or an article. And it does this all at approximately 5.2 billion times the speed of light.

(Editor’s Note: That’s hyperbole.)

So, that’s a basic idea of how the current generation of AI works. Does that info seem hopelessly nerdy? Perhaps. But if it’s a tool you’re considering using in your shop, you want to know how it works and what it can do, right? 

Right.

So, onto the potential use of AI in heavy-duty! 

WHAT CAN AI DO FOR THE HEAVY-DUTY REPAIR INDUSTRY RIGHT NOW?

We thought about being clever and asking ChatGPT what it thought it could do for a heavy-duty shop, but instead opted to ask real live people working in the field. 

“Chatbots are helping us clean up technician notes, draft emails and shop processes, but we are just scratching the surface,” said Bobby Chambers, President of Liberty Fleet Solutions in Fredericksburg, Virginia. “It is just the newest tool in the toolbox, not something to shy away from.”

Stacy Conner of Equipment Experts Inc. in Lakewood, Washington, thought similarly, citing AI can likely be very helpful in streamlining estimates and follow-ups, along with some customer communication. “This is moving at lightning speed,” she added. “If we ignore it, some will shoot ahead, some will be average, and it will keep some people completely out of the game.”

Almost everyone we spoke to, both for this article and at the most recent Diesel Connect, spoke of AI as a way to cut down on repetitive tasks. Over the last few months, we’ve heard people hope that it takes over: 

• Payment processing (particularly via email).
• Cleaning the shop.
• Diag work & basic troubleshooting.
• “Dealing with angry customers.”

First, let us say that we completely understand the desire for a cleaning robot. No one likes mopping up oil and grease. But AIs and LLMs, in their present condition, are largely confined to computers, tablets, and phones (for now).

So cleaning is largely out; you’ll need a cleaning service or maybe an industrial Roomba for that. The actual repair work is also out, since AI at present lacks limbs. We would gently advise against siccing a bot on angry customers, too; while that is technically possible, at least on your website, most folks still want to deal with a real person when they’re having problems. 

What does that leave?

Help with Data, Diagnostics, and Communications

Well, let’s touch again on what AI and LLMs are really, really good at: storing and parsing huge quantities of data. It won’t be turning wrenches for you, but you can ask it to:

• Analyze your business. You can feed spreadsheets, reports, and receipts to your pet AI and ask it specific questions about your shop. Think:
  • “Show me a summary of my inventory that sold this week.”
  • “Give me high-level detail on this report that shows everything sold during the time in the file name.”
  • “How much has Boromir the Tech billed out this week?”
• Help you with diagnostics & troubleshooting. Vehicle manuals tend to be…um…well, dense is the word. Heavy. Thick. Full of words and diagrams and information. Fun to flip through (or not). Fullbay Product Manager Ryan Magnum recalls a prior occupation where he uploaded dealer-provided PDFs of those vehicles to an LLM and then asked questions: “Hey, this engine is struggling to run; what could be some possible causes?”
  
  The LLM would reference what page to look at, and based upon symptoms produced, it could try to figure it out. “It wasn’t always going to be fully accurate,” Ryan said, “but it could at least give us ideas, and get out of that ‘first draft paralysis’ of trying to figure out what to do.” 
• Handle basic tech translation. AI’s transcription abilities have improved quite a lot over the last few years. Yes, you still have some weird slip-ups (Wistia, the software we use to transcribe our Zoom interviews, sometimes writes Fullbay as “Fullbe” and “Full bed”), but it’s overall pretty solid as long as you are speaking somewhat clearly. Along similar lines, you could ask an AI to polish up information that will go into estimates, invoices, and so on.

Verify, Verify, Verify

What do the three things above have in common? They’re ultimately verifiable by humans. 

This is important! We saw “Fullbe” and “Full bed” because we looked at the transcript afterward. The LLM referencing the PDF would provide page numbers so Ryan could take a look at them. And if you are feeding spreadsheets and reports to your ChatGPT account, you still have access to those reports to review the numbers. 

“AI is a bit of a novelty for this industry,” said Peter Cooper of Ascend Consulting, “[but] it’s not very good at understanding logic or sorting through things like opinions vs. hard data from a reliable source.” 

That actually brings us to some important points about AI safety, so let’s keep moving!

SOME GUARDRAILS FOR AI IN SHOP USE

Safety is paramount in heavy-duty repair — in the shop, on the road, and yes, on the computer. AI is pretty darn spiffy, but there are some precautionary measures you should take to make sure you’re a) getting the most out of it, and b) protecting your shop and your customers. 

Keep these things in mind: 

The AI models are changing fast. 

Companies update their LLMs at the speed of light. While this is great for advancement, it does mean they are not always ultra-consistent. A prompt that works today may not gain you the same results tomorrow. Beyond that, you may get different answers from the different platforms.

For a closer look at what each model of LLM excels at (and how they’re changing with updates), check out the AI Benchmarking Dashboard. It regularly puts AIs through their paces and ranks them based on various criteria. 

You have to use the right type of AI. 

If you do decide to play around with AI, make sure you review the terms and conditions carefully. Like any other internet-based platform, you want to make sure protections are in place so hackers/bad actors/the general internet doesn’t gain access to the sensitive information that may wind up on it. Obviously, you want to keep you and your customers safe, so do your research on what data protections and privacy levels you need and what you’re willing to pay for them. 

Don’t ask it to run your shop. 

It’s a robot. It should not be making decisions around what clients to take on or whether you should be offering different benefits. It can help you understand all the information behind those decisions, but you need to still be the one making them.

Specifically, don’t look to AI to tell you who to hire (or really involve it in hiring at all). Why? Well, for starters, you may end up getting sued. But to elaborate on the points we described above, while AI is great at crunching data and making predictions, it’s not going to be able to determine if someone is actually going to be a good fit for your shop. 

Treat it like a trainee.

How do you know if the machine is giving you correct information? By asking questions you can answer for yourself. If you’re asking it to analyze a report, for example, you should review the insights it gives you by matching the numbers to the report. Is this a pain and time-consuming? It can be. But if you want to introduce AI to your shop and make it part of your business, then you need to train it properly. 

Yes, we just said train. AIs/LLMs can be useful if you let them do what they’re good at. Treat them like a very smart apprentice or assistant. Assume the kid has an amazing brain but zero survival skills. Like any greenhorn, you’ll be cross-checking its work until you’re comfortable with its error rate. 

WHERE WILL AI GO IN THE FUTURE?

Something to keep in mind — particularly if all this talk of verification and training is making your head spin — is that this technology is still very much in its infancy. Like it or not, it’s probably sticking around. It will continue to evolve. You may find it becomes more useful as that happens.

And, y’know, if you’re not using Fullbay already but you want to see what we’ve been yapping about, we’ve got a free demo. Go check it out!