In this instructional video, Thomas "Tom" Zoebelein, an AI practitioner for collision‑repair shops, outlines the five most common ways shop owners misuse artificial intelligence and provides concrete remedies. He first explains why treating AI like a Google search (seeking quick facts) limits its value and advises asking for step‑by‑step processes instead. He then warns against using AI merely as a writing assistant, urging owners to view it as a thinking accelerator that can analyze, plan, research, and automate repetitive tasks. The third mistake is weak prompting: failing to give the model a clear role, context, and goal, which leads to vague or incorrect answers. Tom demonstrates how to structure prompts with explicit role, context, and objective. The fourth mistake involves hallucinations; he clarifies that models guess when lacking information and recommends feeding verified sources, demanding citations, and explaining reasoning to curb false outputs. The fifth mistake is poor data hygiene—messy, inconsistent spreadsheets that produce garbage‑in‑garbage‑out results—so he stresses cleaning and standardising data before AI analysis. Finally, Tom shares a “professional prompt template” that bundles role, audience, objective, constraints, background, output format, structure, rules, tone, word count, citation requirements, do‑and‑don’t lists, task ordering, and acceptance criteria, enabling shop owners to generate production‑ready AI results. The video’s purpose is to elevate AI adoption in collision shops from superficial use to strategic, reliable collaboration.
Chapters
You know, I've been working with AI every day for the past few years, developing tools and testing prompts and learning how to make this stuff actually useful for collision shops. And after a ton of trial and error, I've realized that there are common mistakes that people make when trying to use AI. Most people using AI right now are doing it wrong, not because they're lazy, but because they don't understand how it actually works because nobody's taught them how. So today I'm cutting through the noise and breaking down the five biggest AI mistakes that I see shop owners make and and what to do instead. So if that sounds interesting to you, if you think that might be helpful, then come join me in the labs. Hey, I'm Tom. I got my lab coat on and I'm back here at the Hero Group AI Labs with yet another experiment. If you've been using AI tools, whether it's ChatGPT or Claude or anything else, you've probably seen a mix of results. Some amazing and maybe some not so much. After working with AI for years now, building tools for the collision repair industry and working with developers and experimenting daily, I've realized that most people aren't getting anywhere near the full value of AI. And it's not because the tools are weak, it's just because of how they're being used. So today we're fixing that. Here are the five biggest mistakes that people make with AI and how to fix them. So let's dive in. The first one is treating AI like Google. You know, most people use AI like a search engine. They type in a question, hit enter and hope for an instant answer. It can be used as a search engine or an answer engine, and in fact, it's slowly replacing Google. But it's so much more than that. AI isn't built to just fetch, it's built to think with you. The real power comes from conversation, from iteration and reasoning, not just quick facts. Although again, you can use it for quick facts. I do. The fix is don't ask for just answers, ask for processes. For example, instead of saying what's the best way to grow sales, try walk me through a step by step process to increase SAL at a collision center, including milestones and metrics. You see the difference. The first asks for a fact, the second asks for a strategy. The second mistake that I see people make is thinking that it's just a writing tool. You know, a lot of people use AI to help them write emails or to help write a letter. And, and it's fine for that. I Use it for that too. But that's like using your smartphone just to make phone calls. AI isn't just a writing assistant. It's a thinking accelerator. It can analyze, it can plan, it can research. It can automate all the stuff that eats your time. The fix for that is to look for anything that's repetitive or tedious in your day or your job and ask it. Could AI handle this? The chances are that it can. That's how you go from, from using AI for copywriting to using it for profit. Building number three, and this is a big one, is giving it weak or bad prompting. No context, no role. AI is only as good as the directions that you give to it. If you don't define who it is, who you are, and what success looks like, it's going to guess and it's probably going to guess wrong. So the fix is to always assign it a role, context and a goal. So remember that role, context, goal. Here's an example. You're a financial analyst for a body shop. Review this P and L and identify three opportunities to improve profitability. So the role you're a financial analyst, the context for a body shop, the goal. Review this P and L and give me some opportunities to improve profitability. Now, it's not pretending to be an encyclopedia. It's working like an expert on your team. Number four is not knowing how to prevent hallucinations. I'm sure you've all experienced hallucinations, or I've heard people say AI is just full of hallucinations. Well, AI doesn't lie. It just guesses when it doesn't know. And that's how language models work. It's trying to guess what the outcome that you're searching for is. The fix is when accuracy really matters. Always give it verified context or source documents. Tell it to cite where its answers come from or explain its reasoning. The more evidence that you feed it, the less it hallucinates. So think of it as garbage in, guesses out. Number five is ignoring the data problem. Most people want insights from AI, but their data is a mess. Their P and L is inconsistent, their estimates don't line up, their spreadsheets are half complete. AI can't find patterns and chaos. So the reality is garbage in, garbage out, always. What's the fix for that? Just good data hygiene. That means clean exports of your data, consistent column names, and no missing values. AI analysis only works when your inputs are structured and labeled. Those are the five biggest mistakes that people make with AI and how to fix them. If you treat AI like a collaborator and give it a clear role and keep iterating until it nails the result. You'll get outcomes that are actually moving your business forward. Now, before we wrap this up, here's a bonus prompt template for you. And this is the one that I use every single day. The professional prompt template Role URA blank context the audience, the objective, the the constraints, the background, the output. What do you want it to give you? Define the structure. Is it bullet points? Is it a table? Is it a report? Is it code rules? Give it a do and don't list tone word count, no emojis. Include citations. So what to do and what not to do tasks 1, 2, 3. So give it the task that it needs to do and in the order that it needs to do them. What steps should it take in order to arrive at this thing that you want and that quality bar? Well, what acceptance criteria is what? What does good look like? You know, clear, accurate, 8th grade reading level, whatever the quality bar is that you want. So that's it. If you use this structure, you'll instantly sound like a pro and your AI results will jump from pretty good to production ready. And I'm Tom and I'm here at the Hero Group Labs. Thanks for watching me and I'll see you on the next one.
