Artwork

Content provided by thoughtbot. All podcast content including episodes, graphics, and podcast descriptions are uploaded and provided directly by thoughtbot or their podcast platform partner. If you believe someone is using your copyrighted work without your permission, you can follow the process outlined here https://player.fm/legal.
Player FM - Podcast App
Go offline with the Player FM app!

534: How Toma is Transforming Car Dealerships with Voice AI Technology

44:37
 
Share
 

Manage episode 429438301 series 1401629
Content provided by thoughtbot. All podcast content including episodes, graphics, and podcast descriptions are uploaded and provided directly by thoughtbot or their podcast platform partner. If you believe someone is using your copyrighted work without your permission, you can follow the process outlined here https://player.fm/legal.

Host Chad Pytel interviews Monik Pamecha, the Co-Founder and CEO of Toma, a company specializing in AI for the automotive industry. Monik discusses how Toma automates phone calls for car dealerships, enhancing customer service and streamlining interactions. Despite advancements in digital communication, phone calls remain crucial in the automotive sector, and Toma leverages AI to improve these experiences significantly.

Monik shares his journey in the tech industry, detailing Toma's evolution from experimenting with different AI applications to focusing on voice AI. He explains the challenges and successes faced along the way, highlighting how AI technology has matured since his early work with chatbots in 2016. The conversation reveals how Toma's voice AI quickly gained user traction, validating their focus on this innovative technology.

The episode also delves into the practical implementation of Toma's AI solutions in the automotive industry. Monik emphasizes the importance of integrating AI with existing dealership software and the gradual rollout process to ensure effectiveness. He discusses the need for clear communication about AI's role in customer interactions, reflecting diverse responses across different demographics. Monik's insights provide a compelling look at the future of AI in automotive customer service.

Transcript:

 CHAD: This is the Giant Robots Smashing Into Other Giant Robots podcast, where we explore the design, development, and business of great products. I'm your host, Chad Pytel. And I'm joined today by Monik Pamecha, Co-Founder and CEO of Toma, which provides AI for the automotive industry. Monik, thanks for joining me.

MONIK: Hey, Chad, thanks for having me.

CHAD: Obviously, in fact, as evidenced by the guest list that we've had over the last few months, a lot of companies are either integrating AI into their products or starting new companies. And you've been around doing AI for quite a while now. Tell us about Toma.

MONIK: Yeah. So, Toma automates phone calls for the automotive industry right now, and we build a lot of different AI products as well. It's an interesting market, but one of the leading users of phone calls for doing business. So, a lot of the business, which is buying cars, you know, the first touch happens over the phone, you know, people bringing in their cars for service, getting updates, and all that, like, mostly happens over the phone, even though you have had websites and apps and all of these around.

And to give you, like, an idea of scale, like, there are 290 million cars in the U.S. alone, which is, like, about 90% of the population has at least one car. So, scale is massive, and Toma is making that experience of getting service and just dealing with anything related to automotive, like, 100 times better.

CHAD: So, I would encourage people to go to the website and check it out, because, I have to admit, I was a little skeptical, at first, about how good the phone call could actually be. And I was impressed by how natural it was, how it was able to respond in the video demos that I saw. So, how did you know that this was going to be possible?

MONIK: I think a lot of it comes from our own experiences, I mean, not with automotive, but with technology. So, I've been in tech for a long time. I mean, I started writing code when I was, I think, 11 or 12, a similar story for my co-founders as well. But I've been doing machine learning research as well in the past. In fact, this was in 2016 when I wrote a paper on this as well, and we built a chatbot that was based on generative models. And, at the time, in 2016, it was really funky. Like, Google had come up with something called Sequence to Sequence, and we were using that to train it on a little bit of data that we had, and we had something that kind of worked.

And, at the time, I was thinking that, I mean, when you were working with that, you'd see it, like, go off the rails and, like, do something really stupid. It couldn't even get grammar right. And, at the time, I saw all the holes that I was like, if somebody plugged these, like, you know, this would be phenomenal. Like, this is what it takes for it to work, you know, these are the places, more from a practical experience, right? Like, if you had to take it to production, like, what would you need to fix?

And then, six years later, I see that things actually started picking up, right, like, they actually fixed all those holes. And it came back to me, and I was like, all right, this is the time. You know, those were the issues. They're all fixed. Now you can go ahead and build. So, I think a lot of it came from experience as to like, all right, this is what we should build, or this is why we should build, like, in terms of the technology. But we didn't really arrive at this idea, so to speak, you know, at least as a founding team. It was a lot of pivots.

CHAD: What was the original sort of idea that you said, okay, we're going to start a company to do this?

MONIK: Well, that was very different from this very, very different. So, my co-founder and I, like, most founders, like, the first thing they do is they try to look for problems that they themselves have. And we're like, huh, looks like...what do we have in common? We have some chronic conditions, and we've used diet to, you know, manage them. So, maybe let's build a tool to help, you know, patients manage conditions with, you know, diet recommendations. And we spent six months on that, and it went absolutely nowhere.

Also, I think consumer products are just different. They require a different kind of thinking. But, you know, we were just trying to throw something on the wall and pray it sticks. And, you know, it was honestly pretty miserable because we got banned on a bunch of communities on Reddit and Facebook trying to promote it. And, like, all the people who tried our product they just never came back again, and, you know, things like that when you have something that people don't want, right? So, we see that side.

I mean, and after that, we go ahead and we try to do...I think it was during it, I don't know, like, when you get a sense of, like, something's not going to work, where, like, then we realized maybe we should stick to what we know best, which is, you know, we're both technical. So, maybe we should do something that's, you know, more relying on those skills than something entirely different, which I think Y Combinator calls it founder-market fit. I think that's also very true. Of course, you can, like, build something for a business you know nothing about, but there must be some compatibility.

So, yeah, we started with that, kept on experimenting. And then, I think, at some point, we were so annoyed that we made, you know, a list on Google Sheets. And we're like, all right, let's just, you know, vomit out 10 ideas that we have had, and let's write them all, and let's go after them one by one. Let's spend two weeks until we hit something that, you know, maybe we think has legs. And the third idea or the fourth idea on that was building something with voice AI. And, at the time, even that was, like, just a horizontal platform. That was it. All right, so we go ahead and commit to that.

So, if we go through the list, try the first two or three ideas, I think the first thing was...then we went on the other extreme where we're like, all right, let's do something we do all day long, which is, as engineers, we are on call. So, you know, even at, I don't know, 2:00 a.m. in the night if your system is down, you get a pager on your phone saying, "Get on the computer and fix it." And we're like, how could we make that better? That was the first thing on that list. We spent some time trying to do it and, again, we kind of get that feeling.

I think the more you fail, I guess, the better you get at detecting failure. I don't know about success but failure for sure it works like that. I think the third or the fourth idea was voice AI, and then we go ahead. We hack a prototype over a weekend and then put it out on...again, the communities that we know how to market were, like, Facebook groups and Reddit. And it picks up. Like, within, like, 3 days, we had 200 demo calls set up. And that just blew our minds because having been on the other side, we're like, oh, this is what it feels like when people kind of want something, what you have. I mean, it's still not clear, right? But --

CHAD: And you put it out there as automated voice assistant for businesses?

MONIK: No, actually what we did [laughs], I mean, nobody will want to click that if you put it like that [laughter]. You know, just out of curiosity, I was like, "Hey, you know, I've built this thing. It does this, you know, what do you think? Do you think this has any use for you?" And that's it. Like, people are, you know, messaging me nonstop, like, DMing me that, "Can you please share it with me? You know, I run this business. This might be helpful."

It was, like, more genuine. Like, I was just exploring, but, you know, that was a question that I posed. And that had, like, so many people show up, and they're like, "How about you just give us that and we can make money off of this?" And then, we started, you know, digging deeper, and we're like, oh, okay, it looks like you have so many manual processes and across industries. So, we had, like, some people from healthcare, some people from, like, you know, MLMs, multi-level marketing, so many different industries, optometrists, some in construction.

Anyway, so we're, like, thinking at that point, huh, okay. Maybe there is something here. Again, no mention of automotive, no mention of dealerships, nothing. We had a single dealership then. And I would say this was, like, about six months. I don't even know how many months ago, but, like, a couple of months ago. I think, at the time, is when we applied to Y Combinator as well.

CHAD: So, you applied to Y Combinator with the voice idea.

MONIK: Right. And we put something out there. I forget if it was the healthcare idea or the voice idea, but it's probably one of the two. I mean, that's also the other thing about Y Combinator. I think they don't really focus on ideas as much as they just focus on teams, which I think is probably the best practice. You know, we pivoted again [laughs].

But yeah, so we did voice AI, and then spent some time just trying to do everything, right? Trying to build a horizontal layer for voice, where building assistance for all kinds of businesses. And, you know, one of the businesses, at the time, was a dealership. I always like to think of this as an arranged marriage, where, you know, we have the customer. We kind of work through it --

CHAD: So, you had an actual dealership that you were partnered with as sort of a expert in the industry?

MONIK: Our first customer, right. And they were very progressive dealers, so they're always trying new things. And, at the time, we were working with a bank. We were working with some healthcare locations as well. We had some construction industry... whatnot. And we were going crazy trying to build something because everybody had these different requirements. And then, in practice, like, if you push AI out in the wild, to make it work, you need a lot of things, like deep domain expertise being one of them.

So, that realization is happening, you know, where we're coming to terms with that. And, at the same time, it works really well for the dealership, and they bring another customer. And they're like, "Oh, they also want to use it," and we're like, "Okay, sure. We'll turn it on." Then we do it. And then, it works again. And then, they bring more customers. And then, we're like, wait a minute, you know, like, we're not doing any outreach. We're not pushing out anything, and it seems like customers want it. And then, there are these other places where we're struggling so much.

Like, even with healthcare, you know, the regulations in banking and healthcare they slow down, you know, any sort of, like, AI implementations. So, even that world was very different with the automotive space. And you kind of do more of it and then you're like, oh, okay, looks like there is something here, and then we just decide to double down on that space. And then, we go further deep in and you realize, oh --

CHAD: Did Y Combinator...that can be a difficult decision for founders to make. So, did being in Y Combinator help you sort of give you the push to make that kind of bold decision?

MONIK: At least from our experience, it's always been that they're, like, get to the truth very quickly, whatever it is, and then make a decision. Do not delay it. I think we were, in fact, slow to do that. I think they were probably pushing people to do it more because we saw companies pivot in our batch, like, two times, three times right before demo day, which is the end of the program. Like, two weeks before, they just completely changed the company, and that's completely okay.

CHAD: So, how does an implementation actually happen? How does it roll out to a new customer?

MONIK: I mean, this is also very new, right? I think as you come across new customers, you have to adapt the process. But the essence of it is that you first have some data to start, which is, for example, for us, we work with a bunch of call recordings because a lot of our customers are already recording a lot of their calls. So, that gives you, like, some data as to what the experience is like today. Then the next thing is you get an idea of what you know your customers want the experience to be like. And then, you're basically now figuring out the delta between the two.

And then, you're configuring the AI agents, making sure, testing it. And then, you have, like, a period of, like, a week or so where you get through all of that. Then you work through integrations with existing softwares. That also, by the way, is another, like, I would say automotive is a sleeping giant. Like, an auto dealership, on average, like, per month will be spending $50, 000 on software.

CHAD: Wow.

MONIK: Because the whole business runs on software—everything starting from sales, inventory, parts, service, everything, repair orders—all of it comes through that.

CHAD: Now, is there a single common platform that a lot of dealers are using?

MONIK: Unfortunately, no. There are some major players; CDK Global is one of them, which actually was hacked recently. And it's in, like, over 15, 000 dealerships, and all of them shut down.

CHAD: Wow.

MONIK: Like, they just couldn't do any business, and they had to come up with creative workarounds. So, it was pretty painful, kind of, like, a COVID, you know, COVID moment for them. And then, yeah, we've been trying to help our dealers, whoever used that software, to, you know, again, come out with workarounds, where the AI is actually capturing all the information. And, you know, instead of dumping it into that system, it's, you know, finding workarounds on how to get it to our dealers. But yeah, so you integrate with them. That is, like, another major step. And they're, you know, they're not the most tech-forward companies so, you know, that can be a little challenging.

CHAD: Right. So, they use a lot of software, but they're not necessarily tech for...they probably don't have big IT departments and that kind of thing. And then, the users are probably non-technical.

MONIK: Correct. Yeah. The thing about dealers, I think, is that they're so plugged into the business, like, they know everything that is happening in the business. Everybody knows what the bottom line looks, what really will move the needle, what is a good customer experience. They may not be technical, but I don't think it even matters. That's the thing.

But yeah, we were talking about, like, the process, so it takes a couple of weeks So, you do, you know, you get all the information from them as to what needs to be done. You integrate into the systems. And then, the next thing that you do is you start slowly, where, for example, when we start taking phone calls for them, we initially start with off hours and overflow. So, when nobody's able to pick up a phone call, we get the phone calls. And that's how you get, like, some training data at the beginning in a safe manner.

And then, as the volume increases, you know, you get more confident, and you roll it out to a larger audience. But I think the key thing here is it has to be a gradual process because, even for the customer, it's something so new, like, to have a full-fledged conversation. Like, you can have a phone tree where it's, like, press one; press two. You're used to that kind of stuff, and that's been around for 15 years, and it still is, you know. And it is not the most natural thing, but it continues to exist. So, this is the next, you know, natural evolution of that interface where it's more free-flowing and, you know, less annoying.

CHAD: Do all of your customers...when the AI agent answers, does it say it's an AI agent?

MONIK: I mean, our recommendation is to always say that. I mean, it's up to the customer, eventually, if they want to say that or not. And, in fact, it is pretty interesting. Demographics make a huge difference. Like, we're live in, you know, all the states, not all but I would say, like, all the major states. And the way people behave with AI agents is so different, you know, Florida, and Michigan, and California, like, we see the call quality. We see the metrics. We see how annoyed people are or how happy they are and things like that. The way they talk is so different.

And one of the parameters in that is actually, you know, letting them know that it is an AI or not as well. So, we tweak that based on, you know, where we are. But for the most part, we always say that because we want to set the expectations, right? Because, initially, when we didn't, the most popular question on the call was like, "Are you a human?"

CHAD: [laughs]

MONIK: And the fact that people are asking it was also pretty insane, right?

CHAD: Right. They could sort of tell that something was different.

MONIK: Yeah, if you have, like, a long enough conversation because, obviously, it's not human, right? And then, you go, like, five turns into the conversation, and then you realize, okay, it sounds like a human, like, you know, it's speaking pretty quickly. It's giving me the kind of answers I want. But, like, this thing is strange because, you know, humans have a personality now.

Like, with AI, like, a lot of the systems, I mean, you can build personality into it, but it still doesn't have a personality, like, the truth is still that. And it does show up, you know, in interesting ways. And, of course, there can be, you know, some sort of mechanistic issues, you know, like, whatever, right, like, what the customer is really used to and then what you actually say. I think the best practice is to almost always declare, like, it's an AI. And that has improved call quality significantly.

MID-ROLL AD:

As life moves online, bricks-and-mortar businesses are having to adapt to survive. With over 18 years of experience building reliable web products and services, thoughtbot is the technology partner you can trust.

We provide the technical expertise to enable your business to adapt and thrive in a changing environment. We start by understanding what’s important to your customers to help you transition to intuitive digital services your customers will trust.

We take the time to understand what makes your business great and work fast yet thoroughly to build, test, and validate ideas, helping you discover new customers.

Take your business online with design‑driven digital acceleration. Find out more at tbot.io/acceleration or click the link in the show notes for this episode.

CHAD: Did you train your own models?

MONIK: Yeah, we collected enough data to be able to do that, and we have trained a lot of different components and different models. So, when you think of it, there's not, like, one model that does the whole thing. You have, like, a lot of these small, medium, and large models that do different parts. So, the voice and speech are, let's say, two components.

I think the brain of the agent is really the thing that needs the most amount of training because, you know, text to speech and voice, I mean, they have, like, you know, some limits and then, some, you know, business return. Like, after some point, like, there's not really much value to be gained there because if you can transcribe everything, you know, to a certain level of accuracy, all the regions change, you know, accents change. You can always improve. Then it's just expanding scope.

But really, with the brain of the agent, you have multiple different models that actually interact with each other, and they're not just LLMs or generative models. You have a lot of different types of things that are going. You know, you're looking up information. You are, you know, validating something. You're making sure if, you know, this is compliant with what, you know, your company's tone is, all of these happening at the same time. And then, these are the different things that you actually really need to train because that's so specific to, you know, the type of business that is happening.

CHAD: So, are you also doing your own hosting of the models, or are you using a cloud provider for that?

MONIK: Yeah, we use cloud providers. I think having a small team it's insane. I mean, you can host custom models on a lot of these providers now. And then, a lot of them even offer services for you to, like, train and, you know, they take care of the infrastructure as well. I think it's a good thing to rely on it if you're lean and small. There's only so much a few people [chuckles] can do and focus on.

CHAD: What are you most focused on right now, either from a business or a product perspective, or, you know, where's your area of biggest risk?

MONIK: Of course, there is always, you know, risk of competition. And I guess the real question is, like, where...a lot of popular AI companies get asked this as well, right? Like, what is your moat, right? And then, I mean, I think that is the most obvious risk, right? Where, like, what is stopping anybody else from doing what you're doing, right? And there are certain parts of it, which, you know, you can de-risk. Like, having data and having proprietary data is, like, one of the biggest factors in this, right, of de-risking this.

I mean, there's also like, you know, the risk of, let's say, especially in our industry, is, like, taking technology to an industry where, you know, consumers are not pro-technology. You know, they don't want to jump at the best thing that there is, right, especially customer service generally suffers from that, right? Like, people anytime they hear a bot or something and they're like, "Ah, representative agent." So, there's, like, some underlying risk in, you know, human tendencies as to what they want.

But, again, to think about it, like, you know, IVRs, which are these interactive voice response systems, they've been around for so long. Nothing about it is natural. It is completely alien to how we interact, but they've been around for so long. So, a lot of times, like, this innovation is actually pushed down from the business to the consumer and not the other way around. It may not be the best experience. It's getting there.

So, it's that battle between the two, which I think can delay implementations. You know, some people, like, one of our customers, at one point, we went through the entire deal. And I think the owner of the dealership group just said, "I don't believe in AI." Now, you know, like, it's pretty hard, right? Like, you have the metrics. You have the numbers. You're generating value, but the belief is strongly held. And then, at that point, you know, there's nothing you can do.

CHAD: Do you know why they were saying that? Was it like, I don't believe it as in I don't think it can do it, or I don't believe in it as in, like, it's against my ethics, or something like that?

MONIK: I feel like it's probably because they've been burned by past experiences of AI. Like, I think chatbots have been around for so long, and a lot of people in the automotive industry have used them. Now, they used to suck. And as I remember in 2016 as well, you know, over the few years that came, like, it was still pretty terrible. So, I think it's some muscle memory from that.

And then, also, I think AI has been hyped a lot, and I feel like people just generally discount anything that is hyped. And the opinion is that let's just wait for the dust to settle, and then we'll just pick the winners. So, it's also possible, right? On the adoption curve, like, there's you just hit some people who are probably not on the early or even the mid, right? Maybe on the tail end, which I guess is completely fine and true for any tech adoption cycle.

CHAD: And it's true for any product that you're...this is not just an AI company problem. I think it's a startup thing, you know, to find the early adopters and then to move on from there. But you need those early adopters, those champions who are willing to do something new before other people.

MONIK: Exactly. And, I mean, yeah, it's just surprising to me how many early adopters, even in...like, there are almost early adopters in every industry. Every business has people who just want to see something, you know, they're just excited about it, like, they're willing to take the risk. And sometimes I'm not even sure why. But, you know, there's just that element of thrill, and then also, you know, beating the market to it and things like that.

And once you feel it, you understand the adoption curve initially. Because when you see customers, you see, ah, I see every dealership. Everybody should use us. I mean, as a naive founder, I think that that's what I used to think initially. And then, you know, over time, you get a sense of like, all right, you know, these are the types of customers that you should go after. These are the people who you should talk to first. And you build that kind of muscle as a founder and, yeah, new learnings.

CHAD: So, you started with the voice assistant. But are you moving into providing other AI-driven solutions for the automotive industry?

MONIK: Right. So, as we work with more dealers, we found out, you know, more areas that can be improved and, you know, gaps in, for example, communication. I think a lot of, like, quality of service really comes from how you can, you know, communicate with your customer. And it's not just about...you could do a good job and, you know, you could just completely destroy your, you know, quality scores because you didn't communicate well enough.

And you could do a bad service and still have a great, you know, service, you know, experience by communicating well. So, I think a lot of it is key to communication, and that's our focus: using AI to make it better. Voice is one channel. There are other channels as well. And there are a lot of, you know, communication gaps within, you know, our customers, you know, business set up as well. So, we try to bridge that gap.

CHAD: So, since you're focused on communication then, you're probably still leveraging generative AI solutions.

MONIK: Oh, a lot of it is, you know, improved by that technology. Like, so I always think, like, great products usually bring in two things, right? One is a necessary evil. Let's say, you know, something that has to be done like a phone system, for example, like, you need people to call in. You need to set up all the numbers, phone trees, whatever, routing. And then, there is AI, which makes that whole process easier.

So, I think good products usually have these two things combined, where it lets you do one nasty thing, which, you know, obviously, everybody else can do in a different way. And then, there's this one exceptional thing that you can do. And then, [inaudible 27:04] together, and it makes, like, a great offering for the business. I think that's what we're working towards.

CHAD: What do you think about the way things are right now, in general? I do think that there are some companies that are saying, "Well, that great thing is the AI," but they're not necessarily solving a problem that needs to be solved in that way.

MONIK: Yeah. I mean, I think to that part, right, the hype is real. Even in my mind, I just discount, like, 40% of the things that people say about AI now. Like, I mean, I would say it's more true than not, like, 60%, sure, but, like, a rough number in my mind is just 40%, and people, like, exaggerate. But, I mean, that's not because, I guess, they're lying. It's because they're, you know, hopeful, right? Because nobody knows, like, in practice. Like, I mean, now that we've done, you know, hundreds and thousands of minutes of AI phone calls, like, that has, like, you know, added to my judgment. And I kind of know, like, you know, what is possible and what is not with even the most cutting-edge stuff there is.

I think a lot is possible. But it's unfair to say that, oh yeah, it's as good as a human, for example, right? Like, in certain use cases, that just is not true. It's a different paradigm. It's just a different design interaction that has never existed before. There's nothing human about it. You can try to force it to be as human as you want but then it is forced human. Like, it is still not natural because it just isn't.

CHAD: So, I'm getting the sense then that that might not be your north star. That might not be what you're shooting for.

MONIK: Yeah, not at all, no. At the end of the day, a tool should drive business outcomes, right? And then, to drive business outcomes, you got to understand what your customer and their users want, for example. You know, I can imagine a world where people will say, you know, when a human picks up the phone, and they're like, "No, I don't want to talk to you. Can you transfer me to a, you know, the virtual agent?" Like, it will happen, right? And it won't be natural. Like, I do not think it will be natural, and it will be different.

Because imagine, like, a human having access to all the information at the same time. Like, how would they behave, right? Like, humans behave in a serial manner, and then there is, like, some simplicity to some interactions and some complexity to others. That's not the case with, you know, all the information you have. Like, I already know, for example, if you call me, right, and I'm an AI agent for your business, I know so much about you already, right?

Like, I'm not going to act the way, you know, an agent would act who's, like, now pulling up something on the screen, and they're like, "Give me a moment," and then they're reading through your stuff. Like, I already know all the issues you've had, all the conversations you had in the past. So, now I know what's exactly wrong, and, in fact, I'll give you the answer straight up, right? Because I can kind of get ahead and figure out what you really want.

CHAD: There's an example in one of the example videos I watched or, actually, I was trying to think, would I ever want to not talk to a person, right? There's an example in one of your videos where you can see the person does exactly what I do on a call. They say their email address is their first name dot last name at gmail.com. And that's not exactly what mine is, but it's like that.

And I say that, and most of the people that I talk to on the phone when I say that they...and I think it might be because they're not on a screen that has my first name and my last name on it anymore. Oftentimes, they don't remember my name, or can't see it, or can't understand what I'm saying. But the AI has all the information, and it understood what you were saying, and it just gets it instantly.

MONIK: Exactly. Another example of that, right? Let's say, like, you called six times, right? I mean, usually what happens in call centers often is that, like, you get thrown around, different agents pick up, and then maybe the data comes there. I've heard that, like, on existing recordings of, like, humans, where it's like, "I'm calling for the sixth time. Like, do you need me to repeat the same thing again?"

And then, they go through the same flow again because that is the policy, for example. And then, they're just so annoyed. Like, with AI, there's no such thing. It's just, you know, one model that's consistent, that's listening to everything. And it's like, even before you say...like, "I see you've called for the sixth time. You know, I'm really sorry that this is happening," and, you know, whatever. Just simple things like that.

CHAD: I'm just thinking about those experiences that I've had with customer service that have been that. And, yeah, that's why I think that this is really, you know, has a lot of potential. So, how do you sort of, you know, critics of AI will often point to, like, putting people out of work, right? How do you think about that?

MONIK: Yeah. I always, like, to pin it down to, like, evidence, and, I think, at this point, I have enough to talk about this. I think what we've seen with our technology is that a lot of it leads to repurposing of existing talent. So, for example, there are, you know, business development companies that dealerships rely on for inbound and outbound calling.

Now, when we free their time up from inbound, like, that's what we focus on right now, and take off all the mundane tasks, like, the agents that they have are now free to do a lot more outbound, which actually drives more sales or, you know, gives a better experience because, you know, people are checking up on them and saying, "Okay, how was your service a week later," right? And the person feels really good. And if there's any problem, they address it, whatever, right?

So, I think there is more stuff to do than humans will ever be able to do, and our desires have no end. We will continue to pursue that. So, as you free up something...it's like a race which has no finish line. You get a little bit of lead, but that doesn't mean anything because now you still got to keep going and keep going, and that's what we've seen. So, you know, with service advisors, for example, who would get phone calls in the service department of the dealership all the time, now they don't get calls anymore, right? But they're able to spend more time with people in the store. So, they're actually able to upsell more.

So, this kind of efficiencies that you drive, like, they take off the stuff that, you know, you don't want to do all the time and is repeatable to some extent, and then you free them to do things that they couldn't have done before. So, it really is, you know, realizing that there is this endless amount of work that always needs to be done. And here, I took this off your plate, but you still have all this work to do. So, it's just repurposing of, like, talent that's been happening again and again.

And, I mean, there is, of course, that's not to say that there is not going to be a loss of job opportunities, things like that, because, you know, it's just part of creative destruction as it is called, right? Where anything new will create some sort of disruption and then, you know, destroy certain things, but then it creates more, you know, on a net basis. That's happening, yeah.

I mean, if you think about it, like, I mean, I remember I grew up in India. This was, like, 15 years ago, I don't know, maybe 10 years ago, too. There was somebody who would sit in the elevator, and their only job was to press a button. If you think about it, right? Like, I mean, is that job like, you know, what value is it driving? Of course, like, to some extent, right? And then, they came up with these elevators where you could punch in the numbers, you know, exact floors right at the beginning, and you just walk into the right elevator, and that's done.

So, I mean, that job obviously does not exist anymore or does not need to exist anymore, right? But, I mean, I don't know if anybody else in the world has an opinion on that job existing, for example. Like, it's just, over time, when we look back, it just seems, like, obvious that, you know, why were we doing that? We should be doing this other thing. So, I think it's just movement.

CHAD: Right. Yeah. I think it is uncomfortable in the moment but, you know, there is a certain trend to the world, aside from AI, aside even from technology, specifically of progress. And, you know, over time, positive comes from that, but that doesn't mean that there's not pain in the meantime.

MONIK: There is, yeah. Definitely, there is pain. And I think the real reason why people feel this a lot is sometimes, like, even I make that mistake myself of viewing yourself as, like, stationary in terms of, like, skills and learning. It's like, you are everything you've learned up till now. And, okay, if what I knew up until now is not going to be relevant tomorrow, then what am I going to do?

But the thing is that everybody has the capability to learn and improve, and, in fact, even that gets easier and easier with time because technology makes that easier. And then, people are able to do more things than they could do before, learn faster, for example. And it's important to not forget that we have that ability. You know, we can always change and improve, and, in fact, knowing so much makes us even better at knowing more. And that's why we've been able to adapt to every change in history so far; we always have.

You know, so that fear is natural. But I think, over time, when we all look back and we're like, oh my God, why were we doing that? You know, like, and we will all be doing different kinds of things. Like, that is guaranteed. That is going to happen. But that fear still exists, and I think that is what causes the pain. It's the anxiety of it. Like, really? Of course, you have to change, you know, tracks. That is very real, but it's not as painful as fear makes it.

CHAD: Well, speaking of growing, and changing, and improving, you mentioned that, you know, you and your co-founder are both technical. How have your roles changed as the company has progressed, and what have you learned [chuckles], and have you settled into any sort of roles?

MONIK: Yeah. So, I'm the CEO. My co-founder is a CTO. We both used to write code at the beginning. Now only one of us writes code.

CHAD: And I'm guessing it's not you [laughs].

MONIK: Yeah, it's not me. Although I do miss it sometimes, but, actually, to be honest, I don't.

CHAD: [laughs]

MONIK: I feel so happy to have found that, I mean, to have realized that. But yeah, I think, basically, I think now that I understand at least B2B business to some extent, I think you always need to have a clear split of, like, build and sell. And then, of course, there's all this additional stuff that you need to do, but I think these two distinctions need to be absolutely crystal clear because both are full-time jobs.

And more than that, you need, like, owners of those spaces, and it's very hard to jump between the two. I mean, if they are solo founders, I mean, it's incredible how they do it if they're able to do both, or they rely on AI, or they rely on, you know, consultants, or contractors, or whatever. But I think those are the two roles.

And it came, I mean, I think it was just a natural progression. It's, like, when there's more work than people, I mean, that's usually a good place to be, I think, and that's how you know something is going well. You automatically assume natural roles. Because it's in that moment of, oh no, my list is, like, growing, like, quicker than you know, something, and then I need to jump in, and then you pick up the most natural and important things to you.

And then, even if you're not good at, like, you don't have an option. You have to get better at it. For example, selling, like, I never did any sales, ever. Like, I was doing machine learning and distributed systems and whatnot. But I've come to now realize that, okay, that is something that I enjoy that I think I can learn and get better at. And everything I did before, actually, even the engineering mindset helps with sales because it's just a process. So, we kind of assumed our roles when we just had too much on our plate, and we're like, "All right, you'll do this. I'll do this. Okay, fine." And then, we just talk about it, and then, all right, we keep doing it, and then now it just becomes a routine.

CHAD: Yeah, how has your team grown?

MONIK: Well, we've been two of us now, and we're hiring for a software engineer right now. So, we've been very lean. And I think...and this is also to something that Sam Altman said. I think it was him or I don't know who said that but, you know, you'll probably see the first one-person billion-dollar company.

CHAD: Yeah, I think it was, yeah.

MONIK: And I think there is some truth to it. Like, I don't know about billion, but, like, maybe a couple of million, a couple of hundred million, like, that might happen sooner. Because we've always tried to stay very, very lean, and I think we've relied on using technology wherever possible. But yeah, that's not to say...we still need people to build. And we are looking for a software engineer because, at some point, there's only so much we can do.

CHAD: Yeah. So, what would make someone a good fit for your team in that software engineering position? What are they coming to the table with?

MONIK: Everything, I wish. No, I mean, there are some exceptional people. And I think that's exactly what we're looking for is engineers with a founder's mindset because a founder's mindset is always like, you know, give us all the information. We'll make the decisions and figure out like, you know, what needs to be done. And someone who's, of course, exceptionally skilled at technology, at writing code, at building software, but also at understanding like, you know, what to build. I think that is, like, a killer combination and that is what differentiates, like, a great extraordinary engineer from anybody else.

CHAD: Well, especially since it's going to be the second developer [chuckles], you're going to have high needs and expectations for that.

MONIK: And with startups, right? It's always a little bit of chaos, and it's people who thrive in that chaos. And that's the thing, right? I've worked in, like, a bunch of startups that actually went on to become unicorns. I worked at Turing.com, which I think is...$4 billion, something like that now. But when I joined, there were, like, you know, just 10 people. And every company has problems, you know. And there's always this chaos that ensues, you know, at every stage. But there are some type of people who, like, thrive in that. They just love it. And there are some people who complain about it.

I think the ones who complain about it lose that opportunity to grow, and they don't have the mindset to see opportunity in it. And I think those are the people who are absolute amazing, you know, future founders, you know, or even, like, great founding engineers are employees like that because they like that challenge, you know. It's like, this is wrong. Let me go fix it.

CHAD: Right. And the reverse is also true. There's a point in a company's life cycle where they need a different kind of person that is more, like, stable [laughs].

MONIK: Well, I don't know. I think I disagree with that. I think, I mean, that's when the company, you know, plateaus. Like, if you bring in more people like that, you...really, like, what is a company? It's just a collection of really smart people. The fact that OpenAI is able to do what Google cannot over the span of six years is because they just do not hire people, you know, of certain caliber, certain mindset. They just keep them out. Again, that's their policy, or even some larger companies. I think the idea is to keep that mindset going and going. It is tiring, right? But it is what drives innovation. Like, that's just the nature of it.

CHAD: Well, if what you're describing sounds like someone who's listening, or if someone's in the automotive industry and wants to learn more, where can they do that?

MONIK: Yeah, they can reach out to me, you know, my email is monik@toma.com. So, they could get to us.

CHAD: And you can subscribe to the show and find notes along with a complete transcript for this episode at giantrobots.fm. If you have questions or comments, email us at hosts@giantrobots.fm. You can find me on Mastodon at cpytel@thoughtbot.social.

Monik, thank you so much for joining me and sharing the story with me.

MONIK: Thanks, Chad. This was great.

CHAD: This podcast is brought to you by thoughtbot and produced and edited by Mandy Moore.

Thanks so much for listening and see you next time.

AD:

Did you know thoughtbot has a referral program? If you introduce us to someone looking for a design or development partner, we will compensate you if they decide to work with us.

More info on our website at: tbot.io/referral. Or you can email us at: referrals@thoughtbot.com with any questions.

Sponsored By:

Support Giant Robots Smashing Into Other Giant Robots

  continue reading

552 episodes

Artwork
iconShare
 
Manage episode 429438301 series 1401629
Content provided by thoughtbot. All podcast content including episodes, graphics, and podcast descriptions are uploaded and provided directly by thoughtbot or their podcast platform partner. If you believe someone is using your copyrighted work without your permission, you can follow the process outlined here https://player.fm/legal.

Host Chad Pytel interviews Monik Pamecha, the Co-Founder and CEO of Toma, a company specializing in AI for the automotive industry. Monik discusses how Toma automates phone calls for car dealerships, enhancing customer service and streamlining interactions. Despite advancements in digital communication, phone calls remain crucial in the automotive sector, and Toma leverages AI to improve these experiences significantly.

Monik shares his journey in the tech industry, detailing Toma's evolution from experimenting with different AI applications to focusing on voice AI. He explains the challenges and successes faced along the way, highlighting how AI technology has matured since his early work with chatbots in 2016. The conversation reveals how Toma's voice AI quickly gained user traction, validating their focus on this innovative technology.

The episode also delves into the practical implementation of Toma's AI solutions in the automotive industry. Monik emphasizes the importance of integrating AI with existing dealership software and the gradual rollout process to ensure effectiveness. He discusses the need for clear communication about AI's role in customer interactions, reflecting diverse responses across different demographics. Monik's insights provide a compelling look at the future of AI in automotive customer service.

Transcript:

 CHAD: This is the Giant Robots Smashing Into Other Giant Robots podcast, where we explore the design, development, and business of great products. I'm your host, Chad Pytel. And I'm joined today by Monik Pamecha, Co-Founder and CEO of Toma, which provides AI for the automotive industry. Monik, thanks for joining me.

MONIK: Hey, Chad, thanks for having me.

CHAD: Obviously, in fact, as evidenced by the guest list that we've had over the last few months, a lot of companies are either integrating AI into their products or starting new companies. And you've been around doing AI for quite a while now. Tell us about Toma.

MONIK: Yeah. So, Toma automates phone calls for the automotive industry right now, and we build a lot of different AI products as well. It's an interesting market, but one of the leading users of phone calls for doing business. So, a lot of the business, which is buying cars, you know, the first touch happens over the phone, you know, people bringing in their cars for service, getting updates, and all that, like, mostly happens over the phone, even though you have had websites and apps and all of these around.

And to give you, like, an idea of scale, like, there are 290 million cars in the U.S. alone, which is, like, about 90% of the population has at least one car. So, scale is massive, and Toma is making that experience of getting service and just dealing with anything related to automotive, like, 100 times better.

CHAD: So, I would encourage people to go to the website and check it out, because, I have to admit, I was a little skeptical, at first, about how good the phone call could actually be. And I was impressed by how natural it was, how it was able to respond in the video demos that I saw. So, how did you know that this was going to be possible?

MONIK: I think a lot of it comes from our own experiences, I mean, not with automotive, but with technology. So, I've been in tech for a long time. I mean, I started writing code when I was, I think, 11 or 12, a similar story for my co-founders as well. But I've been doing machine learning research as well in the past. In fact, this was in 2016 when I wrote a paper on this as well, and we built a chatbot that was based on generative models. And, at the time, in 2016, it was really funky. Like, Google had come up with something called Sequence to Sequence, and we were using that to train it on a little bit of data that we had, and we had something that kind of worked.

And, at the time, I was thinking that, I mean, when you were working with that, you'd see it, like, go off the rails and, like, do something really stupid. It couldn't even get grammar right. And, at the time, I saw all the holes that I was like, if somebody plugged these, like, you know, this would be phenomenal. Like, this is what it takes for it to work, you know, these are the places, more from a practical experience, right? Like, if you had to take it to production, like, what would you need to fix?

And then, six years later, I see that things actually started picking up, right, like, they actually fixed all those holes. And it came back to me, and I was like, all right, this is the time. You know, those were the issues. They're all fixed. Now you can go ahead and build. So, I think a lot of it came from experience as to like, all right, this is what we should build, or this is why we should build, like, in terms of the technology. But we didn't really arrive at this idea, so to speak, you know, at least as a founding team. It was a lot of pivots.

CHAD: What was the original sort of idea that you said, okay, we're going to start a company to do this?

MONIK: Well, that was very different from this very, very different. So, my co-founder and I, like, most founders, like, the first thing they do is they try to look for problems that they themselves have. And we're like, huh, looks like...what do we have in common? We have some chronic conditions, and we've used diet to, you know, manage them. So, maybe let's build a tool to help, you know, patients manage conditions with, you know, diet recommendations. And we spent six months on that, and it went absolutely nowhere.

Also, I think consumer products are just different. They require a different kind of thinking. But, you know, we were just trying to throw something on the wall and pray it sticks. And, you know, it was honestly pretty miserable because we got banned on a bunch of communities on Reddit and Facebook trying to promote it. And, like, all the people who tried our product they just never came back again, and, you know, things like that when you have something that people don't want, right? So, we see that side.

I mean, and after that, we go ahead and we try to do...I think it was during it, I don't know, like, when you get a sense of, like, something's not going to work, where, like, then we realized maybe we should stick to what we know best, which is, you know, we're both technical. So, maybe we should do something that's, you know, more relying on those skills than something entirely different, which I think Y Combinator calls it founder-market fit. I think that's also very true. Of course, you can, like, build something for a business you know nothing about, but there must be some compatibility.

So, yeah, we started with that, kept on experimenting. And then, I think, at some point, we were so annoyed that we made, you know, a list on Google Sheets. And we're like, all right, let's just, you know, vomit out 10 ideas that we have had, and let's write them all, and let's go after them one by one. Let's spend two weeks until we hit something that, you know, maybe we think has legs. And the third idea or the fourth idea on that was building something with voice AI. And, at the time, even that was, like, just a horizontal platform. That was it. All right, so we go ahead and commit to that.

So, if we go through the list, try the first two or three ideas, I think the first thing was...then we went on the other extreme where we're like, all right, let's do something we do all day long, which is, as engineers, we are on call. So, you know, even at, I don't know, 2:00 a.m. in the night if your system is down, you get a pager on your phone saying, "Get on the computer and fix it." And we're like, how could we make that better? That was the first thing on that list. We spent some time trying to do it and, again, we kind of get that feeling.

I think the more you fail, I guess, the better you get at detecting failure. I don't know about success but failure for sure it works like that. I think the third or the fourth idea was voice AI, and then we go ahead. We hack a prototype over a weekend and then put it out on...again, the communities that we know how to market were, like, Facebook groups and Reddit. And it picks up. Like, within, like, 3 days, we had 200 demo calls set up. And that just blew our minds because having been on the other side, we're like, oh, this is what it feels like when people kind of want something, what you have. I mean, it's still not clear, right? But --

CHAD: And you put it out there as automated voice assistant for businesses?

MONIK: No, actually what we did [laughs], I mean, nobody will want to click that if you put it like that [laughter]. You know, just out of curiosity, I was like, "Hey, you know, I've built this thing. It does this, you know, what do you think? Do you think this has any use for you?" And that's it. Like, people are, you know, messaging me nonstop, like, DMing me that, "Can you please share it with me? You know, I run this business. This might be helpful."

It was, like, more genuine. Like, I was just exploring, but, you know, that was a question that I posed. And that had, like, so many people show up, and they're like, "How about you just give us that and we can make money off of this?" And then, we started, you know, digging deeper, and we're like, oh, okay, it looks like you have so many manual processes and across industries. So, we had, like, some people from healthcare, some people from, like, you know, MLMs, multi-level marketing, so many different industries, optometrists, some in construction.

Anyway, so we're, like, thinking at that point, huh, okay. Maybe there is something here. Again, no mention of automotive, no mention of dealerships, nothing. We had a single dealership then. And I would say this was, like, about six months. I don't even know how many months ago, but, like, a couple of months ago. I think, at the time, is when we applied to Y Combinator as well.

CHAD: So, you applied to Y Combinator with the voice idea.

MONIK: Right. And we put something out there. I forget if it was the healthcare idea or the voice idea, but it's probably one of the two. I mean, that's also the other thing about Y Combinator. I think they don't really focus on ideas as much as they just focus on teams, which I think is probably the best practice. You know, we pivoted again [laughs].

But yeah, so we did voice AI, and then spent some time just trying to do everything, right? Trying to build a horizontal layer for voice, where building assistance for all kinds of businesses. And, you know, one of the businesses, at the time, was a dealership. I always like to think of this as an arranged marriage, where, you know, we have the customer. We kind of work through it --

CHAD: So, you had an actual dealership that you were partnered with as sort of a expert in the industry?

MONIK: Our first customer, right. And they were very progressive dealers, so they're always trying new things. And, at the time, we were working with a bank. We were working with some healthcare locations as well. We had some construction industry... whatnot. And we were going crazy trying to build something because everybody had these different requirements. And then, in practice, like, if you push AI out in the wild, to make it work, you need a lot of things, like deep domain expertise being one of them.

So, that realization is happening, you know, where we're coming to terms with that. And, at the same time, it works really well for the dealership, and they bring another customer. And they're like, "Oh, they also want to use it," and we're like, "Okay, sure. We'll turn it on." Then we do it. And then, it works again. And then, they bring more customers. And then, we're like, wait a minute, you know, like, we're not doing any outreach. We're not pushing out anything, and it seems like customers want it. And then, there are these other places where we're struggling so much.

Like, even with healthcare, you know, the regulations in banking and healthcare they slow down, you know, any sort of, like, AI implementations. So, even that world was very different with the automotive space. And you kind of do more of it and then you're like, oh, okay, looks like there is something here, and then we just decide to double down on that space. And then, we go further deep in and you realize, oh --

CHAD: Did Y Combinator...that can be a difficult decision for founders to make. So, did being in Y Combinator help you sort of give you the push to make that kind of bold decision?

MONIK: At least from our experience, it's always been that they're, like, get to the truth very quickly, whatever it is, and then make a decision. Do not delay it. I think we were, in fact, slow to do that. I think they were probably pushing people to do it more because we saw companies pivot in our batch, like, two times, three times right before demo day, which is the end of the program. Like, two weeks before, they just completely changed the company, and that's completely okay.

CHAD: So, how does an implementation actually happen? How does it roll out to a new customer?

MONIK: I mean, this is also very new, right? I think as you come across new customers, you have to adapt the process. But the essence of it is that you first have some data to start, which is, for example, for us, we work with a bunch of call recordings because a lot of our customers are already recording a lot of their calls. So, that gives you, like, some data as to what the experience is like today. Then the next thing is you get an idea of what you know your customers want the experience to be like. And then, you're basically now figuring out the delta between the two.

And then, you're configuring the AI agents, making sure, testing it. And then, you have, like, a period of, like, a week or so where you get through all of that. Then you work through integrations with existing softwares. That also, by the way, is another, like, I would say automotive is a sleeping giant. Like, an auto dealership, on average, like, per month will be spending $50, 000 on software.

CHAD: Wow.

MONIK: Because the whole business runs on software—everything starting from sales, inventory, parts, service, everything, repair orders—all of it comes through that.

CHAD: Now, is there a single common platform that a lot of dealers are using?

MONIK: Unfortunately, no. There are some major players; CDK Global is one of them, which actually was hacked recently. And it's in, like, over 15, 000 dealerships, and all of them shut down.

CHAD: Wow.

MONIK: Like, they just couldn't do any business, and they had to come up with creative workarounds. So, it was pretty painful, kind of, like, a COVID, you know, COVID moment for them. And then, yeah, we've been trying to help our dealers, whoever used that software, to, you know, again, come out with workarounds, where the AI is actually capturing all the information. And, you know, instead of dumping it into that system, it's, you know, finding workarounds on how to get it to our dealers. But yeah, so you integrate with them. That is, like, another major step. And they're, you know, they're not the most tech-forward companies so, you know, that can be a little challenging.

CHAD: Right. So, they use a lot of software, but they're not necessarily tech for...they probably don't have big IT departments and that kind of thing. And then, the users are probably non-technical.

MONIK: Correct. Yeah. The thing about dealers, I think, is that they're so plugged into the business, like, they know everything that is happening in the business. Everybody knows what the bottom line looks, what really will move the needle, what is a good customer experience. They may not be technical, but I don't think it even matters. That's the thing.

But yeah, we were talking about, like, the process, so it takes a couple of weeks So, you do, you know, you get all the information from them as to what needs to be done. You integrate into the systems. And then, the next thing that you do is you start slowly, where, for example, when we start taking phone calls for them, we initially start with off hours and overflow. So, when nobody's able to pick up a phone call, we get the phone calls. And that's how you get, like, some training data at the beginning in a safe manner.

And then, as the volume increases, you know, you get more confident, and you roll it out to a larger audience. But I think the key thing here is it has to be a gradual process because, even for the customer, it's something so new, like, to have a full-fledged conversation. Like, you can have a phone tree where it's, like, press one; press two. You're used to that kind of stuff, and that's been around for 15 years, and it still is, you know. And it is not the most natural thing, but it continues to exist. So, this is the next, you know, natural evolution of that interface where it's more free-flowing and, you know, less annoying.

CHAD: Do all of your customers...when the AI agent answers, does it say it's an AI agent?

MONIK: I mean, our recommendation is to always say that. I mean, it's up to the customer, eventually, if they want to say that or not. And, in fact, it is pretty interesting. Demographics make a huge difference. Like, we're live in, you know, all the states, not all but I would say, like, all the major states. And the way people behave with AI agents is so different, you know, Florida, and Michigan, and California, like, we see the call quality. We see the metrics. We see how annoyed people are or how happy they are and things like that. The way they talk is so different.

And one of the parameters in that is actually, you know, letting them know that it is an AI or not as well. So, we tweak that based on, you know, where we are. But for the most part, we always say that because we want to set the expectations, right? Because, initially, when we didn't, the most popular question on the call was like, "Are you a human?"

CHAD: [laughs]

MONIK: And the fact that people are asking it was also pretty insane, right?

CHAD: Right. They could sort of tell that something was different.

MONIK: Yeah, if you have, like, a long enough conversation because, obviously, it's not human, right? And then, you go, like, five turns into the conversation, and then you realize, okay, it sounds like a human, like, you know, it's speaking pretty quickly. It's giving me the kind of answers I want. But, like, this thing is strange because, you know, humans have a personality now.

Like, with AI, like, a lot of the systems, I mean, you can build personality into it, but it still doesn't have a personality, like, the truth is still that. And it does show up, you know, in interesting ways. And, of course, there can be, you know, some sort of mechanistic issues, you know, like, whatever, right, like, what the customer is really used to and then what you actually say. I think the best practice is to almost always declare, like, it's an AI. And that has improved call quality significantly.

MID-ROLL AD:

As life moves online, bricks-and-mortar businesses are having to adapt to survive. With over 18 years of experience building reliable web products and services, thoughtbot is the technology partner you can trust.

We provide the technical expertise to enable your business to adapt and thrive in a changing environment. We start by understanding what’s important to your customers to help you transition to intuitive digital services your customers will trust.

We take the time to understand what makes your business great and work fast yet thoroughly to build, test, and validate ideas, helping you discover new customers.

Take your business online with design‑driven digital acceleration. Find out more at tbot.io/acceleration or click the link in the show notes for this episode.

CHAD: Did you train your own models?

MONIK: Yeah, we collected enough data to be able to do that, and we have trained a lot of different components and different models. So, when you think of it, there's not, like, one model that does the whole thing. You have, like, a lot of these small, medium, and large models that do different parts. So, the voice and speech are, let's say, two components.

I think the brain of the agent is really the thing that needs the most amount of training because, you know, text to speech and voice, I mean, they have, like, you know, some limits and then, some, you know, business return. Like, after some point, like, there's not really much value to be gained there because if you can transcribe everything, you know, to a certain level of accuracy, all the regions change, you know, accents change. You can always improve. Then it's just expanding scope.

But really, with the brain of the agent, you have multiple different models that actually interact with each other, and they're not just LLMs or generative models. You have a lot of different types of things that are going. You know, you're looking up information. You are, you know, validating something. You're making sure if, you know, this is compliant with what, you know, your company's tone is, all of these happening at the same time. And then, these are the different things that you actually really need to train because that's so specific to, you know, the type of business that is happening.

CHAD: So, are you also doing your own hosting of the models, or are you using a cloud provider for that?

MONIK: Yeah, we use cloud providers. I think having a small team it's insane. I mean, you can host custom models on a lot of these providers now. And then, a lot of them even offer services for you to, like, train and, you know, they take care of the infrastructure as well. I think it's a good thing to rely on it if you're lean and small. There's only so much a few people [chuckles] can do and focus on.

CHAD: What are you most focused on right now, either from a business or a product perspective, or, you know, where's your area of biggest risk?

MONIK: Of course, there is always, you know, risk of competition. And I guess the real question is, like, where...a lot of popular AI companies get asked this as well, right? Like, what is your moat, right? And then, I mean, I think that is the most obvious risk, right? Where, like, what is stopping anybody else from doing what you're doing, right? And there are certain parts of it, which, you know, you can de-risk. Like, having data and having proprietary data is, like, one of the biggest factors in this, right, of de-risking this.

I mean, there's also like, you know, the risk of, let's say, especially in our industry, is, like, taking technology to an industry where, you know, consumers are not pro-technology. You know, they don't want to jump at the best thing that there is, right, especially customer service generally suffers from that, right? Like, people anytime they hear a bot or something and they're like, "Ah, representative agent." So, there's, like, some underlying risk in, you know, human tendencies as to what they want.

But, again, to think about it, like, you know, IVRs, which are these interactive voice response systems, they've been around for so long. Nothing about it is natural. It is completely alien to how we interact, but they've been around for so long. So, a lot of times, like, this innovation is actually pushed down from the business to the consumer and not the other way around. It may not be the best experience. It's getting there.

So, it's that battle between the two, which I think can delay implementations. You know, some people, like, one of our customers, at one point, we went through the entire deal. And I think the owner of the dealership group just said, "I don't believe in AI." Now, you know, like, it's pretty hard, right? Like, you have the metrics. You have the numbers. You're generating value, but the belief is strongly held. And then, at that point, you know, there's nothing you can do.

CHAD: Do you know why they were saying that? Was it like, I don't believe it as in I don't think it can do it, or I don't believe in it as in, like, it's against my ethics, or something like that?

MONIK: I feel like it's probably because they've been burned by past experiences of AI. Like, I think chatbots have been around for so long, and a lot of people in the automotive industry have used them. Now, they used to suck. And as I remember in 2016 as well, you know, over the few years that came, like, it was still pretty terrible. So, I think it's some muscle memory from that.

And then, also, I think AI has been hyped a lot, and I feel like people just generally discount anything that is hyped. And the opinion is that let's just wait for the dust to settle, and then we'll just pick the winners. So, it's also possible, right? On the adoption curve, like, there's you just hit some people who are probably not on the early or even the mid, right? Maybe on the tail end, which I guess is completely fine and true for any tech adoption cycle.

CHAD: And it's true for any product that you're...this is not just an AI company problem. I think it's a startup thing, you know, to find the early adopters and then to move on from there. But you need those early adopters, those champions who are willing to do something new before other people.

MONIK: Exactly. And, I mean, yeah, it's just surprising to me how many early adopters, even in...like, there are almost early adopters in every industry. Every business has people who just want to see something, you know, they're just excited about it, like, they're willing to take the risk. And sometimes I'm not even sure why. But, you know, there's just that element of thrill, and then also, you know, beating the market to it and things like that.

And once you feel it, you understand the adoption curve initially. Because when you see customers, you see, ah, I see every dealership. Everybody should use us. I mean, as a naive founder, I think that that's what I used to think initially. And then, you know, over time, you get a sense of like, all right, you know, these are the types of customers that you should go after. These are the people who you should talk to first. And you build that kind of muscle as a founder and, yeah, new learnings.

CHAD: So, you started with the voice assistant. But are you moving into providing other AI-driven solutions for the automotive industry?

MONIK: Right. So, as we work with more dealers, we found out, you know, more areas that can be improved and, you know, gaps in, for example, communication. I think a lot of, like, quality of service really comes from how you can, you know, communicate with your customer. And it's not just about...you could do a good job and, you know, you could just completely destroy your, you know, quality scores because you didn't communicate well enough.

And you could do a bad service and still have a great, you know, service, you know, experience by communicating well. So, I think a lot of it is key to communication, and that's our focus: using AI to make it better. Voice is one channel. There are other channels as well. And there are a lot of, you know, communication gaps within, you know, our customers, you know, business set up as well. So, we try to bridge that gap.

CHAD: So, since you're focused on communication then, you're probably still leveraging generative AI solutions.

MONIK: Oh, a lot of it is, you know, improved by that technology. Like, so I always think, like, great products usually bring in two things, right? One is a necessary evil. Let's say, you know, something that has to be done like a phone system, for example, like, you need people to call in. You need to set up all the numbers, phone trees, whatever, routing. And then, there is AI, which makes that whole process easier.

So, I think good products usually have these two things combined, where it lets you do one nasty thing, which, you know, obviously, everybody else can do in a different way. And then, there's this one exceptional thing that you can do. And then, [inaudible 27:04] together, and it makes, like, a great offering for the business. I think that's what we're working towards.

CHAD: What do you think about the way things are right now, in general? I do think that there are some companies that are saying, "Well, that great thing is the AI," but they're not necessarily solving a problem that needs to be solved in that way.

MONIK: Yeah. I mean, I think to that part, right, the hype is real. Even in my mind, I just discount, like, 40% of the things that people say about AI now. Like, I mean, I would say it's more true than not, like, 60%, sure, but, like, a rough number in my mind is just 40%, and people, like, exaggerate. But, I mean, that's not because, I guess, they're lying. It's because they're, you know, hopeful, right? Because nobody knows, like, in practice. Like, I mean, now that we've done, you know, hundreds and thousands of minutes of AI phone calls, like, that has, like, you know, added to my judgment. And I kind of know, like, you know, what is possible and what is not with even the most cutting-edge stuff there is.

I think a lot is possible. But it's unfair to say that, oh yeah, it's as good as a human, for example, right? Like, in certain use cases, that just is not true. It's a different paradigm. It's just a different design interaction that has never existed before. There's nothing human about it. You can try to force it to be as human as you want but then it is forced human. Like, it is still not natural because it just isn't.

CHAD: So, I'm getting the sense then that that might not be your north star. That might not be what you're shooting for.

MONIK: Yeah, not at all, no. At the end of the day, a tool should drive business outcomes, right? And then, to drive business outcomes, you got to understand what your customer and their users want, for example. You know, I can imagine a world where people will say, you know, when a human picks up the phone, and they're like, "No, I don't want to talk to you. Can you transfer me to a, you know, the virtual agent?" Like, it will happen, right? And it won't be natural. Like, I do not think it will be natural, and it will be different.

Because imagine, like, a human having access to all the information at the same time. Like, how would they behave, right? Like, humans behave in a serial manner, and then there is, like, some simplicity to some interactions and some complexity to others. That's not the case with, you know, all the information you have. Like, I already know, for example, if you call me, right, and I'm an AI agent for your business, I know so much about you already, right?

Like, I'm not going to act the way, you know, an agent would act who's, like, now pulling up something on the screen, and they're like, "Give me a moment," and then they're reading through your stuff. Like, I already know all the issues you've had, all the conversations you had in the past. So, now I know what's exactly wrong, and, in fact, I'll give you the answer straight up, right? Because I can kind of get ahead and figure out what you really want.

CHAD: There's an example in one of the example videos I watched or, actually, I was trying to think, would I ever want to not talk to a person, right? There's an example in one of your videos where you can see the person does exactly what I do on a call. They say their email address is their first name dot last name at gmail.com. And that's not exactly what mine is, but it's like that.

And I say that, and most of the people that I talk to on the phone when I say that they...and I think it might be because they're not on a screen that has my first name and my last name on it anymore. Oftentimes, they don't remember my name, or can't see it, or can't understand what I'm saying. But the AI has all the information, and it understood what you were saying, and it just gets it instantly.

MONIK: Exactly. Another example of that, right? Let's say, like, you called six times, right? I mean, usually what happens in call centers often is that, like, you get thrown around, different agents pick up, and then maybe the data comes there. I've heard that, like, on existing recordings of, like, humans, where it's like, "I'm calling for the sixth time. Like, do you need me to repeat the same thing again?"

And then, they go through the same flow again because that is the policy, for example. And then, they're just so annoyed. Like, with AI, there's no such thing. It's just, you know, one model that's consistent, that's listening to everything. And it's like, even before you say...like, "I see you've called for the sixth time. You know, I'm really sorry that this is happening," and, you know, whatever. Just simple things like that.

CHAD: I'm just thinking about those experiences that I've had with customer service that have been that. And, yeah, that's why I think that this is really, you know, has a lot of potential. So, how do you sort of, you know, critics of AI will often point to, like, putting people out of work, right? How do you think about that?

MONIK: Yeah. I always, like, to pin it down to, like, evidence, and, I think, at this point, I have enough to talk about this. I think what we've seen with our technology is that a lot of it leads to repurposing of existing talent. So, for example, there are, you know, business development companies that dealerships rely on for inbound and outbound calling.

Now, when we free their time up from inbound, like, that's what we focus on right now, and take off all the mundane tasks, like, the agents that they have are now free to do a lot more outbound, which actually drives more sales or, you know, gives a better experience because, you know, people are checking up on them and saying, "Okay, how was your service a week later," right? And the person feels really good. And if there's any problem, they address it, whatever, right?

So, I think there is more stuff to do than humans will ever be able to do, and our desires have no end. We will continue to pursue that. So, as you free up something...it's like a race which has no finish line. You get a little bit of lead, but that doesn't mean anything because now you still got to keep going and keep going, and that's what we've seen. So, you know, with service advisors, for example, who would get phone calls in the service department of the dealership all the time, now they don't get calls anymore, right? But they're able to spend more time with people in the store. So, they're actually able to upsell more.

So, this kind of efficiencies that you drive, like, they take off the stuff that, you know, you don't want to do all the time and is repeatable to some extent, and then you free them to do things that they couldn't have done before. So, it really is, you know, realizing that there is this endless amount of work that always needs to be done. And here, I took this off your plate, but you still have all this work to do. So, it's just repurposing of, like, talent that's been happening again and again.

And, I mean, there is, of course, that's not to say that there is not going to be a loss of job opportunities, things like that, because, you know, it's just part of creative destruction as it is called, right? Where anything new will create some sort of disruption and then, you know, destroy certain things, but then it creates more, you know, on a net basis. That's happening, yeah.

I mean, if you think about it, like, I mean, I remember I grew up in India. This was, like, 15 years ago, I don't know, maybe 10 years ago, too. There was somebody who would sit in the elevator, and their only job was to press a button. If you think about it, right? Like, I mean, is that job like, you know, what value is it driving? Of course, like, to some extent, right? And then, they came up with these elevators where you could punch in the numbers, you know, exact floors right at the beginning, and you just walk into the right elevator, and that's done.

So, I mean, that job obviously does not exist anymore or does not need to exist anymore, right? But, I mean, I don't know if anybody else in the world has an opinion on that job existing, for example. Like, it's just, over time, when we look back, it just seems, like, obvious that, you know, why were we doing that? We should be doing this other thing. So, I think it's just movement.

CHAD: Right. Yeah. I think it is uncomfortable in the moment but, you know, there is a certain trend to the world, aside from AI, aside even from technology, specifically of progress. And, you know, over time, positive comes from that, but that doesn't mean that there's not pain in the meantime.

MONIK: There is, yeah. Definitely, there is pain. And I think the real reason why people feel this a lot is sometimes, like, even I make that mistake myself of viewing yourself as, like, stationary in terms of, like, skills and learning. It's like, you are everything you've learned up till now. And, okay, if what I knew up until now is not going to be relevant tomorrow, then what am I going to do?

But the thing is that everybody has the capability to learn and improve, and, in fact, even that gets easier and easier with time because technology makes that easier. And then, people are able to do more things than they could do before, learn faster, for example. And it's important to not forget that we have that ability. You know, we can always change and improve, and, in fact, knowing so much makes us even better at knowing more. And that's why we've been able to adapt to every change in history so far; we always have.

You know, so that fear is natural. But I think, over time, when we all look back and we're like, oh my God, why were we doing that? You know, like, and we will all be doing different kinds of things. Like, that is guaranteed. That is going to happen. But that fear still exists, and I think that is what causes the pain. It's the anxiety of it. Like, really? Of course, you have to change, you know, tracks. That is very real, but it's not as painful as fear makes it.

CHAD: Well, speaking of growing, and changing, and improving, you mentioned that, you know, you and your co-founder are both technical. How have your roles changed as the company has progressed, and what have you learned [chuckles], and have you settled into any sort of roles?

MONIK: Yeah. So, I'm the CEO. My co-founder is a CTO. We both used to write code at the beginning. Now only one of us writes code.

CHAD: And I'm guessing it's not you [laughs].

MONIK: Yeah, it's not me. Although I do miss it sometimes, but, actually, to be honest, I don't.

CHAD: [laughs]

MONIK: I feel so happy to have found that, I mean, to have realized that. But yeah, I think, basically, I think now that I understand at least B2B business to some extent, I think you always need to have a clear split of, like, build and sell. And then, of course, there's all this additional stuff that you need to do, but I think these two distinctions need to be absolutely crystal clear because both are full-time jobs.

And more than that, you need, like, owners of those spaces, and it's very hard to jump between the two. I mean, if they are solo founders, I mean, it's incredible how they do it if they're able to do both, or they rely on AI, or they rely on, you know, consultants, or contractors, or whatever. But I think those are the two roles.

And it came, I mean, I think it was just a natural progression. It's, like, when there's more work than people, I mean, that's usually a good place to be, I think, and that's how you know something is going well. You automatically assume natural roles. Because it's in that moment of, oh no, my list is, like, growing, like, quicker than you know, something, and then I need to jump in, and then you pick up the most natural and important things to you.

And then, even if you're not good at, like, you don't have an option. You have to get better at it. For example, selling, like, I never did any sales, ever. Like, I was doing machine learning and distributed systems and whatnot. But I've come to now realize that, okay, that is something that I enjoy that I think I can learn and get better at. And everything I did before, actually, even the engineering mindset helps with sales because it's just a process. So, we kind of assumed our roles when we just had too much on our plate, and we're like, "All right, you'll do this. I'll do this. Okay, fine." And then, we just talk about it, and then, all right, we keep doing it, and then now it just becomes a routine.

CHAD: Yeah, how has your team grown?

MONIK: Well, we've been two of us now, and we're hiring for a software engineer right now. So, we've been very lean. And I think...and this is also to something that Sam Altman said. I think it was him or I don't know who said that but, you know, you'll probably see the first one-person billion-dollar company.

CHAD: Yeah, I think it was, yeah.

MONIK: And I think there is some truth to it. Like, I don't know about billion, but, like, maybe a couple of million, a couple of hundred million, like, that might happen sooner. Because we've always tried to stay very, very lean, and I think we've relied on using technology wherever possible. But yeah, that's not to say...we still need people to build. And we are looking for a software engineer because, at some point, there's only so much we can do.

CHAD: Yeah. So, what would make someone a good fit for your team in that software engineering position? What are they coming to the table with?

MONIK: Everything, I wish. No, I mean, there are some exceptional people. And I think that's exactly what we're looking for is engineers with a founder's mindset because a founder's mindset is always like, you know, give us all the information. We'll make the decisions and figure out like, you know, what needs to be done. And someone who's, of course, exceptionally skilled at technology, at writing code, at building software, but also at understanding like, you know, what to build. I think that is, like, a killer combination and that is what differentiates, like, a great extraordinary engineer from anybody else.

CHAD: Well, especially since it's going to be the second developer [chuckles], you're going to have high needs and expectations for that.

MONIK: And with startups, right? It's always a little bit of chaos, and it's people who thrive in that chaos. And that's the thing, right? I've worked in, like, a bunch of startups that actually went on to become unicorns. I worked at Turing.com, which I think is...$4 billion, something like that now. But when I joined, there were, like, you know, just 10 people. And every company has problems, you know. And there's always this chaos that ensues, you know, at every stage. But there are some type of people who, like, thrive in that. They just love it. And there are some people who complain about it.

I think the ones who complain about it lose that opportunity to grow, and they don't have the mindset to see opportunity in it. And I think those are the people who are absolute amazing, you know, future founders, you know, or even, like, great founding engineers are employees like that because they like that challenge, you know. It's like, this is wrong. Let me go fix it.

CHAD: Right. And the reverse is also true. There's a point in a company's life cycle where they need a different kind of person that is more, like, stable [laughs].

MONIK: Well, I don't know. I think I disagree with that. I think, I mean, that's when the company, you know, plateaus. Like, if you bring in more people like that, you...really, like, what is a company? It's just a collection of really smart people. The fact that OpenAI is able to do what Google cannot over the span of six years is because they just do not hire people, you know, of certain caliber, certain mindset. They just keep them out. Again, that's their policy, or even some larger companies. I think the idea is to keep that mindset going and going. It is tiring, right? But it is what drives innovation. Like, that's just the nature of it.

CHAD: Well, if what you're describing sounds like someone who's listening, or if someone's in the automotive industry and wants to learn more, where can they do that?

MONIK: Yeah, they can reach out to me, you know, my email is monik@toma.com. So, they could get to us.

CHAD: And you can subscribe to the show and find notes along with a complete transcript for this episode at giantrobots.fm. If you have questions or comments, email us at hosts@giantrobots.fm. You can find me on Mastodon at cpytel@thoughtbot.social.

Monik, thank you so much for joining me and sharing the story with me.

MONIK: Thanks, Chad. This was great.

CHAD: This podcast is brought to you by thoughtbot and produced and edited by Mandy Moore.

Thanks so much for listening and see you next time.

AD:

Did you know thoughtbot has a referral program? If you introduce us to someone looking for a design or development partner, we will compensate you if they decide to work with us.

More info on our website at: tbot.io/referral. Or you can email us at: referrals@thoughtbot.com with any questions.

Sponsored By:

Support Giant Robots Smashing Into Other Giant Robots

  continue reading

552 episodes

All episodes

×
 
Loading …

Welcome to Player FM!

Player FM is scanning the web for high-quality podcasts for you to enjoy right now. It's the best podcast app and works on Android, iPhone, and the web. Signup to sync subscriptions across devices.

 

Quick Reference Guide