How to Use AI to Analyze Reviews and Win | Hero Group AI Labs Ep. 3

Body shop owners. In my 20ish years working in this industry, I've worked with hundreds of body shops and almost all of them can recall a couple of bad reviews and what caused them. But here's the thing. A couple of bad comments doesn't define you, but your patterns do. So today I'm going to show you how to uncover those patterns using a reviews tool and some AI to discover what your customer experience is really like through the eyes and the words of all your past customers. If that sounds interesting, then come join me in the lab. When I ran my marketing agency, my job was to get you more cars. To do that, I needed to see exactly what your prospective customers saw when they checked you out your reviews. Very quickly the patterns began to emerge. I would see some shops that killed it with clear updates and on time delivery, while others struggled with communications or surprise delays. And here's the truth. Sometimes you don't actually have a marketing problem. Sometimes it's a sales problem or even a reputation problem or even just the appearance of one. Today we're going to use AI to do that review analysis at scale. We're going to pull some reviews with a simple Chrome extension. We're going to classify them into collision specific themes like communication, timeline and delays and quality and compare that to competitors to produce a short clear action plan. We're going to identify the top three fixes that will move the ratings and referrals the most. And I'll show you the prompts that I use. Does that sound useful? Then let's get into the experiment. The first thing you're going to do is you're going to Google Phantom GMB for Google, my business audit tool, Chrome extension. And you're going to grab this one right here, Phantom local dot com. So you're going to go here, You're going to come here, you're going to install Phantom, which I already have on mine. You click that and install it. And I already have it on Chrome. So but you would follow the prompts here to install it on Chrome. So once you have that, the next trick is to then go to Google and grab your review. So I'm going to pick a shop here with three locations so that we can, we can do an MSO version here. So I'm going to go to Cherry Hill Collision, New Jersey, they've got three locations. So you'll see here, these are all the extensions that I have. So back in my marketing agency days I would do search engine results. So some of these you may see, some you won't but what you're going to see right here is audit. You're going to want this one right here, the review audit. So you can click that. And this is, by the way, this is one of the locations. I'm going to do this for all three locations. So then you're going to hit review audit. And you can see in the background here, it's performing all the reviews, it's grabbing all the reviews. There's 263 reviews. Only takes a minute or two. All right, it's got that one. Now we're going to go grab the next one. Cherry Hill, Collision, Millville, we're going to do the same thing. Give it a second. It pulls up Review audit. We let it grab all the reviews here for Millville. While that's working, I'll go ahead and get that Laurel going. Review audit and we'll let that go. So there are 114 reviews from at Laurel. Okay, now I've got all my reviews. You're going to come in here and you're going to export your reviews. That one and then here ushers and then Cherry Hill. All right, so that's downloading all those reviews. So now I've got that. Now the next thing is to go into ChatGPT and put in our prompt. Now before we do that, I'm going to read you my prompt that I have. So this is the actual prompt and I'm going to put this in the notes so that you can copy and paste it and use it yourself. And I'll also show you where you can modify some things. But this is the prompt that I'm using. So, you know, here's an area again, why I'm doing this series is because I think most people would just put in chat GPT all their stuff and just say, hey, you know, go through my reviews and tell me what you see. Something like that. But good prompting will get you better results. So here's, here's an actual way you should do the prompting. So system, it's broken up into these sections here. So system who the model is and how it must behave. So I'm telling it that you're a collision industry analyst. Produce a readable executive narrative with and clear tables. Use only the data provided. Be specific, concise and evidence based. So I'm telling it to stick to just what I upload. And then this is why it's built this way, because it sets the role so that the model uses the shop's claims language, not generic marketing fluff. It forces a Narrative first style so that you can read it uses only provided data prevents invented facts, context, scope and desired outcome. So I have objective business type, collision time range, last 24 months and source the phantom export. And it's built this way because it narrows the domain and in the time frame and it names the exact data source that I want it to pull from. And then the next section is data schema what the columns might represent. So I can indicate in here what this these CSV files will have. So review id, date, reviewer rating, text, owner response, so on and so on. The classification schema, the shared vocabulary for the analysis. So I'm identifying that a promoter is a five star, you know, reviewer passive is a four star and a detractor is anybody one to three. The themes I want to look for Communication timeline and delays, quality, price and surprises. Insurance process, rental, loaner, staff experience, cleanliness, delivery, convenience. You can add to this whatever you want. This is just what I'm analyzing it for. Keywords, insurer mention, yes or no. And then I give it some. Then I add rules which the next section is Rules guardrails that keep the results trustworthy. This is where I tell it, you know, don't. Don't interpret things, stick to what's said. Use only the provided text. Don't guess on the names and locations. Person detection from explicit mentions or owner responses. So identifying if there's anybody that gets mentioned in the reviews. Ongoing versus isolated definitions, accounts and time windows Time analysis by day, month data allows neutral tone, site small counts, no updates and then finally tasks. This is what I wanted to do. So I wanted to tag the reviews internally. I wanted to aggregate overall by location and then by person because I wanted to look for a specific if anybody gets mentioned that's doing a good or bad job and produce the narrative in the tables and finally output. This is the exact report that I want. So I want an overall assessment of two to three paragraphs, a quick, quick metric snapshot. I want a leaderboard to see who's being mentioned in the reviews and what's being said about them. Location summary so there's three locations. I want a summary by each location. I want it to not just have an overall company, but I want by location and then I want deep dives by location. So bullets with literal keywords and counts. I want to build out next steps for improvement. And then I want to search for patterns and timing. Like if it sees that there is an issue, was it isolated to certain or is it an ongoing thing? And then input and then that's what I'm inputting in here. Okay so having said that we're going to go over to chat GPT. So I already created a project here, a new product project called Auto Body Reviews homework. But we can start fresh from this and say Auto Body. Body Shop reviews why I'm doing this, why I'm doing a project and not just the new chat is we may want to come back into this later. So this is an area that I again that newbies to using AI may not realize is that you can create projects and then in uploading, in creating products projects you can upload documents that you want to go back to and refer. Instead of having to keep a chat history, you can create projects to do different things with the same data over and over again. So I'm auto Body Shop reviews creating a project. So now I've got that project. I'm going to tell it I'm going to upload three three. CSV files containing google reviews of. Okay so this is just kind of giving an overall what I'm getting ready to upload. I'm going to add the files. So I'm going to go to my downloads and here's my reviews. So. Cherry hill mill bill melt one. Okay now it's got all the files and it's already vectorized this data. So now I just come over to my prompt that I wrote and I'm just going to grab this here foreign. Now I'm using five. Five has the deepest reasoning model at least as of this video here. So it's going to take longer to analyze it. There's a lot of things in here for it to look for and I'm telling it to do a lot of things but with the magic of video I'm going to cut to the results but this will take a few minutes and I'll actually track the time here to tell you how long this takes. While that's working I'm going to go ahead and show here's how you can do the same thing for your competitors. For the competitors. Let's just pick a competitor here. We all know and love Caliber collisions so I'll do caliber go ahead and Cherry Hill. I'm only going to do one so that you get an idea but I can do the same thing. I can do a review on it. 403 reviews for Caliber and to find all your competitors you would just do, you know auto body shop near Cherry Hill and this is a competitor so you would do review audit. May have to upgrade to paid if you do A bunch of these. I just had to do that myself. It's 20 bucks a month or 120 for the year. So Cherry Collision and then we go back to Virgo. So grab their reviews and just so on and so on and so on. Right. You just grab a bunch using the maps. So you just come in here to your maps and then just start grabbing all the ones you know, we can grab. You can determine what you think is a real competitor or not. Cherry Hill Motors review audit, See how many they got. 65. So just grab all your competitors that you want to track. Right. So now I'm going to jump back over here to chat GPT. All right. Our assessment is done. So that took about, I don't know, about six or seven minutes to do that. So detractors primarily cite communication breakdowns, missed updates, misunderstandings about timing, and a small number of quality and fit issues. Bumpers and paint. Mount Laurel shows the highest recent detractor concentration, while Cherry Hill is strong, but had a late summer uptick in communication complaints. Millville remains largely positive with few detailed negatives. So if I'm the owner of this company, it's. It's just drawn a pretty good distinction of what I need to go over on my next meeting. I need to study, which this has done for me, the highest concentration of why they got bad reviews. So there's a lot of quality fitment issues. So I need to get with that location leader or manager, whatever you call them at that location and understand what's going on at that location. Are we understaffed? Do we have the wrong guys there? Do we need to change some SOPs? And then I need to figure out what happened at Cherry Hill because all of a sudden during the summer we had an uptick in communication complaints. Did we lose somebody on the front office? Was it vacations? So we need to drill down and figure out what happened there. So here's a snapshot. Average rating 4.26 vs. 4.41. Top detractor Theme communication issues. The most cited detractor keywords rude insurance, bumper paint, told insurer mentions 17% of the time. Here's the leaderboard. The people that are praised the most saw Hamza mentioned 16 times. Paige, so on and so on. So you can see that certain people are really doing a good job versus those that are, you know, at least not getting mentioned doesn't mean they're doing a bad job. Location summary Detractor share so we can see that Mount Laurel has the highest amount. That's a pretty significant amount 34.4% of the reviews are negative. Deep dive by location what customer said is good what needs improvement Patterns and timing Hamza and Paige repeatedly praised by name Millville what customers are saying is good. You know you can read that yourself here. What needs improvement Negatives lack consistent phrasing. No complaints about paint, finish or delivery Mount Laurel which I know has a problem Friendly professional helpful these are these are the goods. What needs improvement is they're a bit rude at that location apparently Communication and told and tone expectation setting that's isolated recurs across months but less than three in any two months within a six month span. Bumper and paint was mentioned a lot four to five times Quality follow ups occasional insurance issues quoted patterns and timing Communication detractors appear in November, March, September, February multi month presence but below the ongoing rule threshold. I told it needs a greater than three detractors in two month span. And again you can change all the stuff in my prompt. By the way, no single person was criticized but no positives were named. So here's the action plan. Top three shop wide fixes this is the beauty of AI is just tell it hey, give me a plan to fix it so you don't even have to always figure out what to do to fix it. It'll help you why Communication is the number one detractor so we need to work on communication at all three locations. We need to standardize the update cadence every three business days or upon any status change with the with the vehicles in the shop. Here's some steps to do that. Use a single script log into each outbound update in the RO. Create an escalation path for stalled jobs spot check 10 active ROS per week for compliance and then who needs to do that when we need to get started Post repair QC snapshot Find out why bumper and paint shows up in detractors Add a six point QC checklist photograph before delivery attach the RO if a mismatch is found hold and rework and whatever your process is that you want to do for a QC overview Insurance expectations script by 17.8% of reviews reference insurance claims confusion so this is a clear example of where at this location the front office is not explaining the claims process well. So this is an easy fix but probably something that you wouldn't know managing multiple locations if you're not doing a review like this and then per location fixes, you know tighten up cadence on long lead jobs Millville capture specifics and Mount Laurel tone and expectation coaching so there we go. Now we can upload those. Here we go. So here's Cherry Maple, cherry collision, Cherry Hill caliber. All right, so now we got, we're gonna put these three in there. Okay. Now I want to compare my competitors to see what we do best where they're weak. Add to the report a spot analysis. We want to understand what our. Better advantages. Sam. Okay, we'll let that run. Let's see how long that takes. Okay, here we go. Here's the head to head analysis of how Cherry Hill stacks up against some of its competitors. So Cherry Hill has 9.1% detractors at 24, Maple 7.7. They have 5. Cherry Hill Motors Body Shop 43.1. So they have a lot of bad reviews. 43% of their reviews are bad. And then we've got caliber collision shockingly at 6.9%. So you and Maple are neck and neck on ratings. So they have, we have similar ratings. Caliber leads slightly on the rating. But it has the lowest detractor share. So people are generally happy with that caliber location. Cherry Hill Motors body shop is the weakest by far. With communication and timeline failures dominating. Your detractors most often cite communication. Your reviews are also most likely to mention insurance signal that expectations around adjusters and supplements influence satisfaction. It's an interesting finding there. Most cited detractor keywords top signals at Cherry Hill Insurance told time back company at Cherry Hill Motors weeks days time insurance shop caliber it's insurance back time and work executive interpretation here where you win today. High satisfaction at scale, staff experience and promoters Positive reviews frequently praise professionalism helpfulness when communication is proactive where your competitors are weaker. Cherry Hill Motors body shop is consistently weak on communication and delays. Maple has few negatives but they occur when they occur. They reference estimates and what was said suggesting an expectation drift. Usually when I see that estimates and people complaining about price or you'll see in a review that they're saying oh they're charging, they're trying to screw my insurance company. That's a sign to me that the front office is not communicating the process for getting an estimate. That initial estimate versus what the final estimate is going to be after teardown are going to be two different things. Caliber subtractors lean timeline and delays. Their insurer mention rate is low indicating they may deflect that context to their customers or their customers mention it less creating an opportunity for you to differentiate on transparent insurer handling your biggest improvement. Letter communication around insurer steps and time frames. Your detractor language clusters around told time and insurance indicating that how and when expectations are set matters as much as final quality targeted SWOT analysis Cherry Hill versus your competitors here's what the strengths are consistently high rating staff experience praised insurance fluency is visible weaknesses communication driven detractors insurance context often becomes the customer's problem High insurance mention rate correlates with the tractor language around timing and approvals when updates aren't proactive opportunities differentiate on insurance transparency exploit Cherry Hill Motors body shop's weaknesses message against Maple's soft spot the with written timestamp update commitments and quote clarity and then the threats caliber's slightly stronger headline and much larger volume gives them a broad market credibility. Well we all kind of know that, right? Maple's parity on rating with a marginally low detractor rate could cipher customers if your communication issues issues persist and insurance cycle volatility will keep creating detractors unless expectations framing becomes bulletproof what you do best versus where they're weak actionable takeaways you versus caliber caliber are strong on timelines lower detractor share tied to delays because they are a high volume shop that moves quite quickly. But customers rarely mention insurance exploit make insurer made clear your signature every job gets a claims roadmap approval steps typical wait times you know explaining how long this thing's going to take from start to finish and auto updates every three business days. I should mention this is a heavy DRP shop so we are comparing apples to apples here. This turns competitive silence into differentiation. You versus Maple Maple's ratings are close but the detractors reference estimate friction and then you versus Cherry Hill motors they struggle heavily with communication abilities. They're not even a really a competitor in my opinion after reading this. And then we have your 30 day action plan here. Then an appendix. So that's it. That's the review analysis. I could go on and check different models here, but I think this is doing a pretty pretty thorough job. You know we could put this through Claude and see what happens. I don't know we're gonna get that much of a difference. I can jump in and try it and we'll see. But, but anyway, there you have it. That's my review analysis. This is this week's experiment and hopefully you found this useful and you can put this to use in your shop and get a little bit of a leg up on your competition. Thanks and I'll see you on the next one.