“This Is the Worst It'll Ever Be” – State of Dealership AI and What Dealers Should Be Asking

What's changed for Podium since Eric Rea's last appearance?

Eric Rea: If you think about somebody who's really good at selling cars, or really good at scheduling test drives, the number of paths a human brain that's really good can take is thousands and thousands of things. So it's not as simple as, "Hey, I want this AI BDC to be aggressive." There are all these things — if the car isn't on the lot, what do you want to do? If it's in transit, if it's on a boat, if it's on a train, if it's on the lot but there's a deposit on it, do you still want to schedule a test drive for it?

Yossi Levi: Today I'm joined again by Eric Rea, CEO of Podium. Eric returns nearly a year after his last appearance, and much like the AI industry, his company has made huge strides — now deployed in nearly a third of U.S. franchise dealerships. We dive into the current state of dealership AI and the key questions dealers need to ask to stay ahead. A big thank you to our sponsors for making today's episode possible: Qmerit, Lotlinx, and of course, Podium. And now, let's get into the show. Eric Rea, back on the CDG podcast. Eric, welcome back.

Eric: Thanks for having me, Yossi. Great to be back on.

Yossi: Last we spoke was about a year ago. A lot has changed in some respects, and in another, not much has changed — we'll talk about that. One thing I noticed prepping for this — and correct me if I'm wrong — is that Podium is today in one out of every ten dealerships. Is that correct?

Eric: Yeah, it's even more than that. Podium is being used by, I think, closer to 30% of OEM dealerships in the United States. It's our AI specifically that's being used in about 10% of dealerships now.

Yossi: Wow, that's a pretty incredible stat. I don't think I realized just how much you've scaled — obviously it's because you came on the Car Dealership Guy podcast here.

Eric: Yeah, it's all happened since the last one. It's funny — we serve home services and another vertical too, so sometimes people don't think of us as an automotive company. But we literally have, I think, 7,000 OEM dealers in the United States using Podium.

Yossi: So what's the secret?

Eric: There's no secret, honestly.

Yossi: But let's — for anyone who didn't catch the first time you came on and we chatted last year, can you give us the 30 seconds on Podium and yourself?

Eric: Yeah. So Podium — we've been going for 11 years. That's one of the secrets: you just keep going and don't run out of money. Eleven years ago, my dad ran a tire shop in Canada. I worked there growing up, didn't think anything of it, but I got exposure at a young age to how the car business works, and realized my dad had very little technology to run his tire shop. So I went off, studied computer science, had a short stint at the United Nations working for the International Atomic Energy Agency, realized that was a terrible match for me, and came back and started Podium to help my dad with his reviews problem. He's a great guy, but sometimes people would go on Yelp and Google and call him names, and he wanted to know how to fix that. I said, "Well, you've got like 99% of your customers who like you, so just get them to review you online and you'll be good." He said, "I don't have time for that." So I solved the problem for him, and it turns out millions of businesses — well outside the tire shop — have that same problem. That's what we started the business with, and today we do a lot more for a lot of businesses.

How did Podium get started, and why does founder mentality still matter?

Yossi: You know what I noticed? I can tell you still have that founder mentality, because the first thing you said is, "Don't run out of money." You're a massive company today, but it's such founder mentality — just keep going, keep compounding, don't run out of money.

Eric: Yeah, it is crazy, because we've been in business for 11 years and we've had plenty of ups and some downs, like every business has. You learn over time that if you just keep going, and you don't make fatal mistakes, it will become something great. Time is a great compounding function.

Yossi: That is absolutely true, and not just in business — I find it in life as well, and in friendships. Compounding is pretty incredible. So, before we get into the nitty-gritty — if I had to sum up my conversations from the last couple of months with dealers about AI, it's: "Hey Yossi, I think I need to use some AI, I don't know what, where, when, what's best for me." That's the oversimplified state of where we're at. The second interesting conversation is when dealers text or call me and ask, "What do you think about this tool versus that tool?" and I'll say, "This is great, that's great, but how are you thinking about this in terms of ROI?" It's interesting that dealers are looking at their SG&A on their P&L as opposed to their marketing budgets — meaning, can I save on headcount if I implement this AI tool? I find that fascinating, because we talk a lot about AI replacing people, but when I have real conversations with dealers who are doing the math on, "If I implement this, it's like I didn't have to hire two extra people," I think it's a pretty incredible place we're in today. So let's start there — since you last came on, about 12 months ago, what's changed the most about the state of AI in dealerships?

How is AI transforming dealerships in 2025?

Eric: Yeah, there are a lot more dealerships using AI today than in 2024. When I came on last time, I think it was August, and NADA earlier that year — AI was the big hot thing, everybody was looking for an AI BDC or whatever. So it was going, but what's happening — and we're seeing this with our customers — is everybody wants to try it out. Dealers will try anything, they're very exploratory. And just because somebody buys your product doesn't mean they're going to keep using it. One thing I've learned working with dealers is you've got to keep providing value, or they're gone — more than in any other industry out there. Everybody was excited initially, there are more products in the market today, and the big thing is it's so easy to have a really compelling demo — even for us. It's really not that hard to have success with AI, whether it's service or BDC, in a very controlled environment where the happy path is tight. But real life isn't like that — real life has a thousand variations of how these things happen, a thousand ways leads come in, a thousand ways conversations can go, inventory, all these things. Even for us — we have a thousand dealers using Jerry to take leads and convert them to test drives using AI — we're still working every single day to make sure Jerry can do his job across a broader and broader set of variables. Does that make sense?

Yossi: And what's Jerry?

Eric: Jerry is our AI employee — our AI BDC. We also have revenue mining and a couple of other things, but primarily right now everybody's using Jerry. Leads come into the CRM, then into Podium, and Jerry takes over. It's mostly text right now — we're starting to do some voice on service and other things — but leads come in, and Jerry works with potential car buyers to schedule them to come down to the lot and take a test drive.

What makes Jerry, Podium's AI BDC, special?

Yossi: Okay, now simplify this for me — a couple of general questions. What are you seeing as the highest-value use right now dealers are gaining from your AI? Is it service, is it sales? Where are you seeing the most incremental value?

Eric: There's a huge opportunity in service. We started in sales because that's where we've sold into the GSM and the GM for the last nine years — our reviews product, all of our communication tools, have mostly been targeted at the sales side of the business. We think it's the easier place to sell into, because the GM typically cares way more about sales than service, although it's up for debate whether that's actually the right thing. But we've focused a ton on sales. On the sales side, the best thing is: you get all these leads, people looking at cars, kicking tires — being able to take those leads and convert them into test drives without employing a massive BDC team is huge. Last year, our BDC agent was outperforming human BDC reps in certain cases, and today it's pretty clear that a really well-trained AI BDC can outperform most humans.

Yossi: Naturally, some people are skeptical, right? And it's changing so rapidly, which is why I like to know the specifics. I recently had Tatiana on — from Van Horn Auto Group, who's been a client of yours for a long time. She's one of the most impressive BDC directors we've had on the podcast, because I love the level of detail she goes into. We were chatting about how she's leveraging AI, and of course she's using your tool. She mentioned specifically being able to customize her AI to respond to customers for different situations — it's not just a canned response. I asked her, "Different brands, different types of questions — how are you tailoring this to the customer to make sure you're not losing someone who doesn't want to chat with a non-human?" How are you navigating through that?

How does Podium customize AI for each dealer?

Eric: Okay, first thing — every dealer is different, even within Van Horn. The Toyota dealership works completely differently than the Ford, it's crazy. We're working with a really large auto group right now, and if you go to the Beverly Hills dealership selling BMWs or Mercedes, they want a totally different Jerry than the Toyota dealership in Atlanta. So the thing we've learned, and the thing we've been forced to get good at as we get Jerry into more dealers, is that Jerry has to be able to adapt. A great example: at a Toyota dealership, high volume, they want Jerry to be really assertive and aggressive — they want Jerry to follow up a lot, to apply more pressure. Makes sense. But if you're selling Mercedes-Benzes or Porsches, you want to be way more consultative and concierge. Initially, our first version of Jerry just took a lead and turned it into a test drive, without thinking about any of that. So what we've built — and I think this is the big thing — is the ability for a dealer to train Jerry over the first 90 days, and they can keep adjusting it after that too. Over those first 90 days, you've got to train this AI just like an employee, tuning it: "I want more pressure," or "I want it to be less pressure." It's wild.

A good example — we're rolling out to a really large auto group right now, and it's going well. The other thing everybody knows is these huge groups have great corporate teams, but the GMs still run the business — it's very decentralized in many cases, which is great, because I actually think they make better decisions since they're closer to the business. A month ago, we were traveling around to all these dealerships where we were putting Jerry on, and one GM said, "I don't really like Jerry — he's not doing this well, he's not being compelling, I don't think he's good at selling. I don't want to use him, because right now he's not better than my BDC team." He gave us all this feedback. Luckily, we had one of our engineers there — he literally drove to a coffee shop after the meeting and spent the next six hours building in the feedback that guy gave. Fast forward to this week, we go back on-site with this guy, and he says, "Jerry's completely changed." Now he gets mad if his BDC reps or sales reps interrupt Jerry or take over the conversation, because Jerry does it better now — there's a very tight process.

The other thing is, if you think about somebody who's really good at selling cars or scheduling test drives, the number of paths a human brain that's really good can take is thousands and thousands of things. So it's not just, "I want this AI BDC to be aggressive." There are all these things — if the car isn't on the lot, what do you want to do? If it's in transit, on a boat, on a train, on the lot but there's a deposit on it — do you still want to schedule a test drive for it?

Yossi: How do you train for the long tail of those objections or questions? How do you even do that, without having to train every single second of every day?

Eric: Before the reasoning models came out, it was strict guardrails — basically a bunch of mini agents handling very specific use cases, and it was very labor-intensive because engineers had to build each one. Now what we're finding is we can give general policies and procedures. We onboard a dealer, and through interviewing them, come up with the general policies and procedures of how a BDC works at that dealership and how they want it run — that's step one. Then you turn Jerry on, and we tell people: if they don't expect this, they're going to be disappointed — Jerry is going to make a mistake, Jerry is going to do something you don't like, guaranteed. When that happens, they can go in and say, "Next time a car is in transit, schedule the test drive," or "tell them the car's not on the lot and they can come test drive something else." There are all these cases, and the nice thing is, with reasoning models now, we can just give Jerry more general policies to abide by, and because the LLMs can reason step by step, they can take those policies and do a really good job.

Will AI replace BDC teams?

Yossi: That's incredible, and it's definitely come a long way. I'm trying to digest all of what you're saying. So how are dealers implementing AI right now? We're talking mostly about text and chat — are you seeing them retain their BDC teams and add incrementally, slow down on hiring, or are you seeing them actually let people go? What's the return on investment, and how are they actually deploying it in their dealerships?

Eric: It's all sorts of things. There are actually two primary ways we're seeing BDC teams respond to AI coming into their business. We have some dealers opening new dealerships and trying not to have a BDC team at all, because they think Jerry can do a great job. And others are using it as a co-pilot — supercharging their BDC team — which is the most common thing. Most commonly, they're not going to get rid of their BDC, at least not right now. And if anybody's telling you that AI can completely replace your BDC, that's not true, because there are still things AI cannot do, like walk around the lot and check vehicles — sometimes that's what needs to happen. We have a human-in-the-loop feature: Jerry's going to go do his thing, he knows the outcome he's trying to achieve, which is schedule a test drive. But sometimes he needs assistance, because we've given him access to inventory, the DMS, the CRM, and all the data — but sometimes that data isn't in there, and we know dealers' CRMs aren't perfectly clean. So Jerry will ask for help — he'll send a notification to the BDC team or the GSM or somebody else, saying, "I need you to tell me if this car is actually on the lot." They'll walk out, verify it, say yes, and Jerry can continue on.

Yossi: Now, talk to me about actual usage and onboarding — where are you seeing the most friction? Is it training Jerry? Is that ultimately the most difficult part?

Eric: I think the GMs and GSMs want AI in every part of their dealership, because they're motivated — we're selling a lot of it, so I think they are. But the people on the ground, the BDC teams and even some salespeople, there are two ways I'm seeing them respond. Some see it as competition: "My job's going away, and I need to compete against this." We were on-site, talking with one of the BDC reps at one of these dealerships — I think it was in Atlanta — and she said, "I think I might quit." We asked why, and she said, "Because I cannot beat Jerry. I'm eating dinner at 7:30 p.m., a lead comes in, I check my phone five minutes later, and Jerry's already responded." So we're seeing some people look at this as an existential crisis for their job, and they're actively rejecting it. And then we're seeing other people see this as a way to make themselves ten times more productive. I think every BDC rep today doesn't want to be a BDC rep forever, so I'd actually look at this as an opportunity to get really fluent with AI and figure out how the world is going to work, because this stuff isn't slowing down. These dealers are getting more opportunity as AI makes lead conversion better and sends more people into the dealership — that's more opportunity for the rep too, to go into other parts of the dealership, or get into actual sales. I think people who look at it that way are going to do really well. But it's interesting to see the differences.

Yossi: Eric, you're coming from a position with a lot of power, because, as you mentioned, maybe AI is in 10% of dealerships in the U.S., but Podium is in 30% of dealerships to some extent. What's your vision? On one hand, I heard you say it's replacing human labor. On the other hand, it's making current employees ten times more efficient. What's the goal ultimately — where are you steering the ship to push the industry forward?

What's Podium's vision for AI across the auto industry?

Eric: I think there are a lot of things AI is going to be able to do better than humans, because of its ability to be on 24/7 and have unlimited access to data and knowledge. Our vision is to build a system of agents across auto — sales, service, parts, finance, all the different areas. A year ago I would have said just really repetitive, basic tasks, but now our view is that it can do a lot — it can do financial planning, it can do almost anything. I think that's the future, and we have a choice: we can embrace it ourselves and build for it. I think dealers, especially principals, are pumped about this. Our job is basically to build these products, make them great, make them able to cover all these unique use cases, and get them into the hands of dealers, and we'll learn as we get them into the market what works and what doesn't. But yeah, I think human labor is going to be pretty disrupted — not just in auto. Our engineers at Podium are 50% more productive than they were a year ago, and they're our most expensive resource. And that's as bad as it's ever going to be — a year from now, who knows, they might be ten times more productive.

Yossi: One of the things I wanted to ask about, with onboarding — one of the tools we use here at Car Dealership Guy is something called Granola, an AI notetaker that's very passive, in the background. It's interesting because it can observe lots of conversations. My thought was: is there a world where something like that sits in the background with the top BDC agent or the top salesperson, listening to them, so the training doesn't have to be such a proactive act — it can be reactive, in the background? Just something that came to mind thinking about adoption and getting to the best customer experience as quickly as possible.

Eric: I love that, and that's for sure the future. Am I joining the product team at Podium? We are hiring. No, but that's going to be the case — it already can be, too. Granola's already doing that, honestly, and we could build something tomorrow that could listen in on your best reps, with consent of course, and do it the right way. The interesting thing is, in the early days, because we have all the text messages from all these dealers, we thought that for a dealer, when we set up a dealership, we could take all of their text messages and use them to reinforce the model and make it better. When we tried that, it turned out there were so many bad habits baked in that it actually made the model worse. I'm serious, we tried this — but that was like two years ago. I think AI has come a long way, and I think we can get to a point where you say, "I want you to take all the conversations from this rep and this rep and use that to train Jerry," and it'll get better and better over time — you'll basically be able to keep training it forever.

Yossi: Can you imagine picking from a menu — I sign up with you right now, and it's like, "Okay, whose BDC do you want to operate like — Longo Toyota, or whoever?" It's just incredible, picking from a menu.

What results are Podium's top-performing dealers seeing?

Yossi: Talk to me a little bit about your best-performing dealers — whether you can name drop or not. I'd like to know what you're seeing from a process and actual results perspective from your top-performing dealers.

Eric: Van Horn has been a customer forever, and they've been awesome — they've embraced this stuff, they're one of the earliest adopters, so they're really good. I can't remember all the exact stats, but basically lead response times are about 70% faster when you use AI, it's handling eight times more conversations than the top BDC reps — it can handle almost unlimited — and conversion rates are up to 30% higher. That's crazy.

What should dealers ask when shopping for an AI vendor?

Yossi: I want to peel back the curtain a little more here — this is something somebody specifically asked me to ask you: if you were a dealer shopping for an AI tool, whether chat, text, or voice, what questions would you ask? You're behind the scenes — what would you want to know?

Eric: It's a good one. I think this is really the problem right now — I believe even when we started selling it, our demo was so good, because you only have to be good along a controlled environment. Honestly, you and I could spend 20 minutes right now and build the most amazing voice AI demo to show a dealer, and they'd be blown away. So I think the questions you have to ask are: who's using this, what are they seeing — get really deep, like you're interviewing someone. The way I'd treat it is: you're not buying software, you're not buying features, you are buying a person — an AI employee that's going to talk to your customers every day. So you've got to ask questions like you're interviewing a new BDC rep. You should ask: what does Jerry do, how does it work with inventory, what does Jerry say when there's a vehicle on our website that's actually not on the lot — how do you handle that? A lot of companies have no clue that's a huge problem, but they can show you a really compelling demo that gets you to pull the trigger. So I'd treat these AI buying cycles more like interviewing an employee than buying software — because software is just, "Here's the demo, here's where the buttons are, this is a little easier than what you currently have."

Yossi: Well, speaking of interviewing — if I'm not mistaken, you don't even call Jerry a chatbot. You call it an AI agent, is that right?

Eric: We call it an AI employee. But yes.

Yossi: Does that mean I have to give it benefits and healthcare and all that?

Eric: No. That's the beauty.

Yossi: How far are we from the day some government somewhere tells me I have to pay employment tax on an AI employee?

Eric: I don't know. It's interesting, because you know that's going to happen. There was an HR tech company that came out with this big PR thing saying they were going to create the ability to add your AI employees to their HRIS — their human resources information system — and basically put them on par with human employees, and it was a major backfire. People were like, "This is insane, what are you talking about?" But I don't know where it's all going to go. The way we look at this right now is these agents are super smart. When we met a year ago, we thought it was going to get much better, but I didn't think it was going to get so much better so fast. You're very involved in tech, you're pretty popular on X — every week a new model comes out, and it's cheaper and better, cheaper and better. So, man, I don't know — it's going to be interesting.

Yossi: I don't know if you saw this study that went viral about the effect on the human brain from AI usage — it's super depressing.

Eric: Yeah, because I'm one of those chronic AI users where I can see that I'm thinking less when I'm just writing things.

Yossi: The punchline for our listeners is that it pretty much said the people leveraging AI strategically, as a supplement whenever they really needed it, were the ones benefiting most — it wasn't bankrupting their brain — versus people like me who are completely exploiting it and using it for many different things throughout the day, actually becoming dumber. So I want to apologize to the audience if I get dumber over the next couple of years. Hopefully not, but that was pretty depressing.

Eric: I don't even use Google anymore, dude. I use Perplexity, Grok, and ChatGPT. I had this prep doc for the podcast, and I was reading through it, and I wanted to know what we talked about last time — the stats I shared. The old instinct is to ask my team, and they'll go read through the transcript or listen to the whole episode — it's like 45 minutes — write everything down, and send it to me, and it'll take three hours. Instead, I took the link, put it into Perplexity, and said, "I want you to go through this and tell me all the stats that Eric shared." Ten seconds later, it had all the stats. I was like, boy.

Yossi: I know, it's crazy.

Why is service the next big AI opportunity for dealers?

Yossi: So give us the goods — what did you share then versus now? What's improved? Has anything gotten worse in terms of actual stats?

Eric: Let me bring it up. Back then I shared that Podium works with 7,000 dealers, the platform's being used by over 100,000 businesses across all verticals, and at that time we were serving about a thousand dealerships. Today it's 10% of the OEM dealers in the United States.

Yossi: And the big dealer groups — likely with many stores — are they asking you for any major feature requests? So much has happened in a year. What should we anticipate over the next three, six, nine months?

Eric: We're really focused on making sure we're the best AI BDC agent out there, because that's not something you can move on from. We'll have an engineering team on the AI BDC forever, because it's always going to be getting better, and our competitors aren't moving off it either — or maybe they will, and we'll dominate. But then service AI is huge, and adding voice as a new modality is massive too, and then obviously parts, revenue mining — basically the entire sales process from start to finish. But I'd say it's interesting — I think there's actually more opportunity in service, because once you get somebody in the dealership, it's human to human. Jerry's not going to show up in an Optimus robot and sell you a car, at least not now. On the service side, though, there's tons of opportunity, because it's a very back-and-forth, transactional experience — even scheduling is complex, there are always questions up front about whether you should even schedule an appointment. Then you drop your car off, you leave, and you're asking for updates. And on the technician side, there's so much waste — these technicians bill 185 bucks an hour or whatever, and they spend two hours of their day just staring at software, starting from scratch, like, "Oh yeah, it's a Nissan Xterra," and so on. So there's so much of the service side of this business that's ripe for opportunity.

Yossi: Yeah, it's huge. The one thing I'll caution, and I think it's important to look at, is what capacity dealers have in their service departments. This has been one of the complaints — "I don't have capacity, so even if you get me more." And then the shortage there is technicians, of course, or labor. That's the difference, because if you told me you could bring me another hundred customers a month, I'd absorb a hundred customers a month. Service is a bit of a different beast.

Eric: Yeah, they've got different bottlenecks. We've been digging into this a ton, and we found the same thing. My instinct was that voice AI would be so good for service — somebody calls and wants to schedule, and if you can do that with AI, that's great, and I think that does happen. But as we've been talking to service managers, a lot of the phone calls are questions about things preemptively, before they actually schedule service — and what we've heard from a lot of service managers is they ask all these questions, call, call, call, and then they still go online to schedule. So that's interesting. And then, yes, the big thing is: you've got technicians — how can you make them more efficient, how can they just show up and have everything already ready, so they don't have to worry about filling in all this information? We were talking with a big dealer group, talked with all their service managers, and the big thing is: can my guys spend two hours a day not working on cars, not making money? There's so much opportunity there.

Yossi: Yeah, and by the way, assuming they're flagging hours, the techs aren't happy about that either, because they're losing out on potential income. So the incentives are definitely aligned.

Eric: Well, then the other thing — this is where my tire shop experience kind of relates. When I was working at my dad's tire shop, I worked at the front mostly, I'd sometimes go do tire repairs, but you'd spend half your day getting a phone call: "Hey, my car's in there, can you give me an update?" You'd say, "Okay, let me put you on hold, I'll call you back." Then you'd go in the back and ask the tech, and they'd say, "Oh yeah, it's going to be another two hours," and you'd go back and call them back, and they wouldn't answer because they're busy, so you'd leave a voicemail, and then they'd call you back — that whole thing takes up a technician's time. I know a lot of service centers are doing this proactive stuff now, where the tech takes videos and sends that out, video inspections. But that can all be automated as well. I see a future where every technician has an AirPod in their ear all day, and they're asking it questions, and also giving live updates — "I just finished the brakes on this, looks like this person could use tires" — that's how they should all be working.

Yossi: I'll put a little shameless plug here — I invested in a company called MasterTech.ai, and that's essentially what they're creating — companions you can chat with or interact with all day to help diagnose vehicles. I thought that was interesting, because this is a use case popping up in a lot of different places — keeping that AirPod in, which is really what's happening in many cases, and just literally having it as an advisor. I know some dealers who are literally chatting with ChatGPT, so why can't you do that for more specialized tasks and just have it with you all day? I think it's a very smart thing.

Eric: Yeah, and if it's hooked into your inventory, or your parts inventory and all that stuff, it's going to save these guys tons of time.

Should dealers be early adopters of AI, or wait?

Yossi: So, Eric, before we wrap up — I asked you what you would ask if you were shopping right now for your own product, but what's your take on being an early adopter? I wouldn't even call it early adopter at this point, but in general, what's your take on AI — being an early adopter versus waiting? Is this the kind of thing where if you wait you're completely missing the boat and falling behind, or do you not see much risk in that? How do you think about that balance?

Eric: I think about AI — my buddy Wade at Zapier made this AI fluency table for his company: engineering, finance, sales, a row for each one, going from "not acceptable" up to "expert," with descriptions of how good you should be at AI based on your job. I think that's a big part of this — this stuff's not going away. Is it perfect today? No. You'll get frustrated with any AI solution you buy, because there are edge cases that just aren't great yet. But this is the worst it's ever going to be, and it'll get better.

Yossi: That's a really strong line — "this is the worst it's ever going to be." That might make it into the YouTube title, man.

Eric: But it's true. We tell our engineers that if they don't want to use Cursor or whatever, we say, "Okay, but next week Cursor will be better" — so just get used to using it now. My thing is: this is the future, it's not going away, it's not a fad. It's not like the tablet craze, where everybody was going to have a tablet and now people don't really have one unless they're flying on an airplane. This is not that. It's hanging around, it's here for good, and you may as well figure out how to run your dealership with it, because I don't see a world where that's not the case.

When will voice AI go mainstream in auto?

Yossi: And in terms of voice AI, what's your prediction on when that's going to be ubiquitous across every dealership? You said you're getting into that now, right?

Eric: Yeah, that's tough. In a different vertical — in our home services vertical, we have voice AI live with hundreds of customers. There are voice AI companies in auto too, but it is so much harder than asynchronous text or even live chat. It's insane to me — it's the difference between driving a car and flying an airplane. When you drive a car, if things go wrong, you pull over to the side of the road, you're fine, you can look something up on your phone, call whoever. If your car breaks down, no big deal. But if you're flying a plane and something goes wrong, you don't know what to do — everybody dies. And I think with phone calls, and voice AI in particular, that's it. We're finding that with text AI, if it takes a minute for us to go in the background and do deep thinking, make a bunch of calls, our average number of calls to an LLM is like 30 per turn on a text chat. You can't do that with voice, because it's too long. It's super complicated, and I think that's going to be the next frontier — but it's legitimately ten times harder than text AI.

Yossi: I see your point — you lose that person when they hang up, as opposed to text, where you can keep going. You could call them back, you could text them, but you lose that engaged person. So, in essence, the risk is higher — your margin for error is lower.

Eric: I think we're going to bridge that — there are many companies tackling that issue, and I'm sure the industry is going to leverage it. We believe voice is going to be ten times bigger than text, even though we think text, email, and all these asynchronous channels are great. Think about the number of phone calls across — I actually prefer text, it's quicker, it saves you more time.

Yossi: Me too. We built our whole company on the idea that texting was superior to a phone call — we really did. We started with reviews via text, and then realized businesses didn't really text their customers, and that's this ubiquitous thing for consumers. So that was the next wave Podium did — let's make texting your car dealership normal and good. We built this whole thing. Guess what, though — 80% of conversations across all of our customers still happen over the phone. So eventually, guess what we had to do? We built a phone system.

Yossi: Because if Podium only serves 20% of the conversations happening at your business, how valuable is it? So in a huge number of our customers today, we're powering all of those channels. Oddly enough, even a company that started on the idea that text is far superior to phone calls, we feel like voice is not going away, and with AI, voice interactions are going to become even bigger. I'm sure you use these advanced voice tools, OpenAI's advanced voice and all these things, and I'm finding now — and this is probably to the point of us getting lazier and dumber with AI — I don't want to type out my question to Perplexity or ChatGPT, I want to just say what I want. And because the AI is smart, you can say things in a sloppy way, no prepositions. Something I typed into Perplexity two days ago was, "way e-foil good," and it came back with, "Waydoo is this manufacturer of e-foils," and then, "Here's all the reviews, here's the comparison models." I'm talking like a caveman.

Yossi: I know, oh man. Well, dude, exciting stuff, and I think people really appreciate you sharing some of your insights on the best ways to shop for these tools and how to think about it. It seems like you've been making incredible progress, and I'm excited to hear about your venture into voice — it's going to be a big year for you. I'm sure we'll follow up again, but Eric Rea, thanks for coming on the podcast. It's been great as always.

Eric: Yeah, thanks Yossi. Appreciate it.

Yossi: Thanks, man. All right, hope you enjoyed that episode. Please give the podcast a rating, consider subscribing to the show, and check the show notes for links to what we talked about. Thanks for tuning in — I'll see you guys next time.