AIMinds #043 | Will Bodewes, Co-Founder & CEO at Phonely

About this episode

Will Bodewes is the CEO and co-founder of Phonely, a company using voice AI to transform call centers. With a PhD in applied AI and a background as an ultra-endurance athlete, he focuses on enhancing customer support through innovative AI solutions.

Listen to the episode on Spotify, Apple Podcast, Podcast addicts, Castbox. You can also watch this episode on YouTube.

In this episode of AIMinds, host Demetrios sits down with Will Bodewes, the CEO and co-founder of Phonely, a company revolutionizing call centers through voice AI technology. Will shares his journey from ultra-endurance athlete to entrepreneur, including the lessons learned from his first venture, Spokesound, and the inspiration behind founding Phonely after earning a PhD in applied AI.

The conversation dives into Phonely’s mission to reduce human workload and improve customer interactions using AI that simulates real conversations. Will discusses the potential of AI-to-AI interactions, focusing on the challenges of multilingual support, security, and minimizing hallucinations in AI responses.

Will also reflects on the future of AI in customer support, sharing how Phonely is tackling challenges in sales environments and operational efficiency. He concludes by highlighting the significant impact Phonely is having on enhancing customer satisfaction and revolutionizing call center operations.

Fun Fact: At just 14 years old, Will Bodewes and his family kayaked around Lake Superior. It took them 72 days, and they set a world record for the youngest people to ever achieve this feat.

Show Notes:

00:00 Early experiences shaped my limitless perspective.

04:50 Accepted PhD offer for funded AI research.

07:47 Starting company to maximize potential through dedication.

10:26 Voice AI for efficient call center solutions.

14:59 People may prefer AI over human interactions.

17:45 AI resolves 80-90% of calls, saves costs.

More Quotes from [Speaker]:

Transcript:

Demetrios:

Welcome back to the AI Minds Podcast. This is a podcast where we explore the companies of tomorrow being built AI First. I'm your host, Demetrios. And this episode, like all episodes, is brought to you by Deepgram. The number one speech to text and text to speech API on the Internet today, trusted by the world's top enterprises, conversational AI leaders and startups. Some of them you may have heard of like Spotify, Twilio, NASA and Citibank. Today we are joined by none other than Will, the CEO and founder of Phonely. How you doing, man?

Will Bodewes:

Yeah, good. It's great to be here.

Demetrios:

So you've had quite the journey when it comes to founding Phonely. Let's just start with the non traditional stuff that you did back in the day when you were an ultra endurance athlete. What's that about? I heard you kayaked for a very long time.

Will Bodewes:

Yeah, they've done. I've done quite a few, like, ultra endurance things. When I was, When I was 14, my dad and my two brothers set out on this trip to kayak around Lake Superior, which when you're 14, you don't really realize is like such a huge deal. But we spent 72 days, on the shore of Lake Superior just kayaking through basically every condition and ended up like, setting a world record for the youngest people that ever did.

Demetrios:

And this was during the summer?

Will Bodewes:

Yeah, this was during the summer. So he took an entire summer off and just did that.

Demetrios:

Okay, so you also did some other wild stuff like bike home from university and what gave you this bug of wanting to just push it to the maximum?

Will Bodewes:

Yeah, I think, I think doing a lot of that stuff from a young age really like, changed my perspective on a lot of like, the way that I approach life, which is, you know, like life for me is it feels like an experiment into what is the most that you can possibly do and like, how far can you push yourself to be able to accomplish these things. And so I think that for me it's always been really interesting to see, like, you know, what is, what is the maximum of the potential rather than just being like, okay, am I just like happy or comfortable or stuff like that? And so I think for me it's always just been like, you know, with whether it's starting a company or, you know, doing these like ultra endurance cycling races or, you know, whatever, it's always been trying to find that the maximum limit. And I, I think it just started from a young age.

Demetrios:

You make me want to get off the couch and go to the gym right there. I didn't realize we were going to get so deep so fast, but I like it. Now. Talk to me about the first startup that you created. You decided to go into the hardware space and go D2C. What was it, what were you selling and what were some learnings?

Will Bodewes:

Yeah, so the first company that I started, it was called Spokesound. It was a. Basically, it was a Bluetooth speaker that was built into a canvas photograph. So it was just like, you hung it on the wall, it looked like a picture, but it played music like a speaker. So it was pretty cool. We had some decent engagement. Like, my. I think probably the.

Will Bodewes:

The peak of that was Chance the rapper slid into my DMs once. That was pretty cool. It still follows me on Instagram from it. but. But ultimately, like, learn. It was like, fresh out of college, didn't really know much, but great learning experience of, like, okay, what to do, what not to do when starting a company. But ultimately, for me, it was like, I didn't feel like I was solving a problem in the world. I was creating a nice to have.

Will Bodewes:

And so it created this, like, bug of, like, okay, I wanted to be doing entrepreneurship, but I just wanted to really be solving a problem and know that I was working towards something that was going to help people, you know, be more efficient and, you know, save their time.

Demetrios:

How did you recognize it wasn't a problem?

Will Bodewes:

It was because I was always trying to put a problem statement on top of it. I mean, I think that, like, you can assign a problem statement to what it is to be like, okay, well, speakers are ugly and they need to be pretty and. But I just didn't feel like that was a big enough problem for me to spend the next, like, five to 10 years of my life just, like, working on. So that was what ultimately, like, after a year and a half, I just kind of like, okay, I'm gonna move on to something else.

Demetrios:

All right, so as far as life accomplishments, you have set world records and Chance has slid into your DMs. Which one do you hold in higher esteem?

Will Bodewes:

Uh, yeah, definitely the world record one. I mean, I think, like, I don't know, I. I'm not a big celebrity person, but I thought it was funny, and it's like a random two truths and a lie. That is great.

Demetrios:

So then you decided to take a little hiatus, go to Australia, do your PhD, explain more about that.

Will Bodewes:

Yeah. So after, like, it was basically like, as I was running this company I had, it was middle of COVID So I had a, like, contacts with a PhD advisor over in Australia that I was looking at doing a PhD before starting spoke sound. Uh, he basically just reached out and he was like, hey, like, I'll give you, like, if you come and do this PhD, you know, it's, it's fully paid for, and then I'll give you like a trip to, to Africa to go and do some like, exploratory research. Cause I just got this grant and I was like, well, that seems interesting. And, and so I, and at the time still continued to like, run the company that I was working on remotely. And so I said that, that seems interesting. So I, I said yes, and started doing that. I was doing just like, AI research over there, still a little bit of work in Africa, but then, you know, ended up doing most of it in the University of Melbourne in Australia.

Demetrios:

And what was the PhD?

Will Bodewes:

It was just like applied artificial intelligence research. So I was looking at different, like, I looked at a lot of different things as, as you do in a PhD, but like, ended up. My first couple of research questions were just around like, applying like, AI methods to energy problems in, in Africa.

Demetrios:

Wow. Oh, that's awesome. Yeah. Hence the trip to Africa. Now that makes sense. But what happened?

Will Bodewes:

Yeah, yeah, exactly. Well, as I was going through and doing this, it was like, for my first startup, I knew that I wanted to build a company. Like, I, I knew that I wanted to do that. And I think that I was always really interested in, like, in tech and LLM since, since, you know, OpenAI had come out. And I saw this opportunity to, you know, like, I kind of saw all these different things coming together. One of them was like, you know, like Voice models, like 11 labs are coming out. These, you know, AI models, like 3.5 was coming out that could do function calling. And I kind of saw this, like, a combination of these technologies to be able to provide a ton of value to different businesses.

Will Bodewes:

And the way that I, the way that I thought about the problem was it was like, you know, I thought that with voice and with voice AI, there's. We've now created like a million new knobs and dials to be able to like, turn, to be able to not just create like a chat bot that talks, but to be able to recreate that amazing experience of calling in and like, having a really good customer support agent that has empathy, that can answer the phone, that can respond in the way that you need to, and then also the ability for these AI models to be able to do function Calling so they can actually interact with software just like a person is. Because it feels to me like we have this problem where we have millions of people in the world that are just sitting in a desk, they're working on a computer, and then they're telling somebody about what they're doing on the computer. And that seems, that whole pipeline seems like something that should be automated so that we can let, you know, humans do less robotic work and robots do more robotic work. So it felt like a big problem to solve and so I wanted to work on it.

Demetrios:

Now, did you have that in a moment of inspiration? What brought you there?

Will Bodewes:

Yeah, I don't know if it was a moment of inspiration or as, or as more of like, I knew that I wanted to start a company. I was not super happy with, like, you know, having one company that didn't really like, succeed and kind of like, I think I, you know, from doing this like, ultra endurance stuff as a young age, I was always kind of thinking of myself as somebody who was able to, you know, I was like, okay, like, I'm, I'm top. I'm, I'm, I'm special. You know, if, like, if we go that route and the, the feeling of just like, you know, getting a PhD was, didn't, I didn't feel special. So I said, okay, I'm, I'm doing, I'm, I'm changing this. I'm gonna put, you know, basically from 6am until 10 or 11pm at night, I'm just gonna work on building something and I'm gonna work as, as hard as I possibly can. And I, I set the time for like 30 days where I literally didn't do anything else, didn't eat any sugar, didn't, didn't do like nothing. Like, nothing that was like bad, like by any standard, and said that I'm, I'm just gonna create something and I wanna see the maximum of my potential.

Will Bodewes:

And so that's how Phonely got rolled out. And then six months later we got into YC and, you know, it's been kind of up and to the right since then.

Demetrios:

I love how you preface it with you didn't eat sugar.

Will Bodewes:

That was hard for me. All right.

Demetrios:

As a fellow glutton, basically, I can very much think, like the 30 days I could do, the working from 6am to 10pm I could do. But the no sugar part. Yeah, congratulations. And, and so then this was born out of it. And what was really the. So the. You gave me the problem statement, but what was the inception? How Long did it take you in that 30 day journey to realize I want to go and I want to create something like phonely.

Will Bodewes:

So it was during the beginning it was like spinning around, like, definitely a bunch of different ideas, you know, playing around with a lot of technology that I had been playing around with for a while. But it was like, you know, I think within like the first few days of just like looking at, at the time like, you know, some of the, some of the businesses on the market. Just doing my research and looking at the opportunity size, I said, okay, this one seems like a really big opportunity. Seems like it's going to take a lot of, you know, hard work to solve. And I felt like it was something that I could spend, you know, a long time working on, I mean, like, you know, years and still like not get to the bottom of the technical problem that that is, that needs to be solved. So yeah, I guess that was the, that was kind of the, you know, the time and then it's, it's taken, you know, a good, a good chunk of time to get to where we are today, obviously.

Demetrios:

And where are you today? What does the product look like?

Will Bodewes:

Yeah, so we think of the product is. So we work with a lot with like contact centers and call centers. So if you either run like an in house contact center or an outsourced contact center and you want to add voice AI answering into the offer to basically, you know, significantly reduce the cost of, you know, answering the phone, that's where we work really well. And then what the product does is we think of the product as not having to, you know, spend a bunch of time doing prompt engineering. We handle almost all of that with kind of the way that we've built the platform and the flows that we've built and then also the interaction with software. Because, you know, one of the things that, like I was alluding to earlier, that's really important about these AI models is their ability to be able to actually do things. And so we kind of take like type of a like agentic plus voice approach to be able to actually accomplish real tasks that businesses need done so that you can have resolution rates that are, that are high without having to spend, you know, three to six months building out all these different API integrations to be able to do that. We also focus a lot on like, you know, how do you, like I was saying, how do you recreate the best human level experience and that we find ourselves doing that by, you know, like running, having like a simulation test feature built in the platform where people can go in, they can upload existing call recordings, they can build an agent of Phonely and then they can simulate how those calls would have performed with phonely.

Will Bodewes:

So those are a couple of like our key features.

Demetrios:

Oh wow, wait, keep talking about that simulation. So you just have one side of the conversation and then you expect Phonely to fill in the other side. Or do you just have the start of the conversation and then firmly goes and does what it wants?

Will Bodewes:

Yeah, so, so essentially like what we can think about is phone is the platform for building these AI phone agents. Right. And so, but once you've built a AI phone agent, you either have to call in and like spend like a million hours just like dialing this thing, or you can like run like AI based simulation models. Like we've seen a bunch come out in the YC batch of these, like these companies running simulations on top of these voice AI agents to like solve their problems. But one of the interesting things that we thought was like all of these businesses that we're working with have years and years of call recordings that they've just saved. And so if we build out an agent with phony, it should be able to like, if we, you know, transcribe and diarize the conversation, it should be able to go forth and resimulate that conversation and see how it would have, it would have performed. So we can actually get a ranking of like, okay, how did the AI perform versus how did the human perform? And then the idea is that we can continue to optimize for performance on that over time so that we can go live knowing to the customer that it's going to resolve a certain percentage of their calls.

Demetrios:

Wow, okay. Yeah, that's very cool. Now I had recently just heard a friend ask me about a bit of a thought experiment that I want to propose to you because I'd love to get your take on it, which is in the future when it comes to these voice agents, do you think there is a world that we are going to live in where we have two voice agents talking to each other? Like my voice agent is talking to a call center voice agent. Because. And so I thought, instinctually I thought, no, that's not going to happen because it'll just be on the software level. Right, but, and that's way more efficient. But what if you have only the means to contact the company through a call? And I'm not sure if your call center has agents, but I am sure that I can use my agent, which I've yet to seen fully be something that is very popular where it's like consumer agents, but that's a whole nother story. But my agent calls your call center and you have call centers.

Demetrios:

And so we have voice agents talking to each other. Do you foresee that as like a potential timeline that we could ride on?

Will Bodewes:

Yeah, I think it's definitely going to happen. It feels like a short term thing. You know, it's probably going to be like in Silicon Valley when Dinesh's AI talks to Guilfoyle's AI and then everything shuts off. No, but I think it's definitely going to happen. But I feel like voice is a very efficient means of communication. Right. And so I think that, you know, at some point people are going to become so accustomed to calling in and getting an AI that they're actually going to start preferring that because they know that it can, it can resolve, it can resolve their problems. Right.

Will Bodewes:

And so I think that right now people have this like initial, like, you know, they call and they get an AI and they're like, I don't want to talk to that because they can't do anything. Right. But once people start having good interactions with these AI models calling into businesses, they're probably going to start preferring them because, you know, if you think about like there's, there's definitely a lot of businesses where, you know, humans are actually create an additional variable. Like when you're talking about your finances over the phone, you always feel bad that like you went and bought that fancy new car that you knew you definitely couldn't afford. So you don't tell your financial advisor that you did that. Right. And so an AI model, you would feel better about doing that because you're like, well, it's just an AI. Like, I don't have to feel that bad about it.

Will Bodewes:

Right.

Demetrios:

What? I just want to underscore this. The solution to not feeling guilty about buying a car that you can't afford.

Will Bodewes:

Oh, yeah.

Demetrios:

Is not having to pay the piper when you are talking to the bank. It is just that now it's going to be an AI model and so you don't have to worry about.

Will Bodewes:

Exactly, exactly. That's the solution to all problems.

Demetrios:

Yes. So I do see that though also the sad case is that it creates another security vulner vulnerability. Right. There is a world, I think it's pretty locked down, but there is a world where you have these layers of humans that can take some of this information or they can use it in nefarious ways. So I'm, I'm not saying that AI is counting itself out of that world, but I do think that if you offered an AI agent a couple thousand bucks on the side to give me my identity, it wouldn't be as motivated as if it were maybe a human agent.

Will Bodewes:

Yeah, yeah, I think it's, I think it's interesting. There's definitely, there's two, two sides of every like, new thing that comes out. Right? It's like, you know, the same thing with them as the Internet was coming out. It's like, you know, it's created a lot of good and it's allowed us to be able to like, have this podcast while we're, you know, far apart. But there's also definitely a lot of bad that's going to be used from it. But I think that we see like, I don't know, we've seen with, with generic, like, technical innovation that, you know, good outweighs the bad. Like there always will be a percentage of bad, but the good that it provides to society usually ends up being, being better. And I don't foresee the AI is going to be any different.

Will Bodewes:

Like, there's definitely going to be people that are cloning voices, you know, trying to do scans, you know, all of that stuff that we've seen already happening. But I think that, you know, that's part of where regulation comes in and where, you know, just like the people that are there to help hopefully solve some of these problems can come out and help people.

Demetrios:

So now what are some stats that you've seen when it comes to folks that have adopted phonely?

Will Bodewes:

Yeah, so, so usually like, I mean, right off the bat, it's like if, if you're missing any calls, out of hours or anything like that, like, that goes to zero. And then in terms of resolution rate, you know, it really depends on the customer. But you know, we've seen everywhere from like, from like 80 to 90% of like, calls successfully resolved by AI without having to like, transfer to a human. And that's, I think that that's really exciting to just like walk through these conversations and see how AI can not just like, answer their calls, answer frequently asked questions, but then go ahead and like, perform an action and be able to do anything that like, a person would be able to do. And so usually like, I guess it comes with like, lots of cost savings, usually about 50 to 60% as to what they have so far, 100% of their calls answered. And then not having to deal with things like multilingual agents and just being able to get something up and running that. That actually is able to cut costs significantly.

Demetrios:

Talk a little bit more about the multilingual agents, because that's a fascinating one for me.

Will Bodewes:

Yeah, so a lot of the businesses that we work with, they have, you know, they. They have this problem where, like, people will be English or Spanish or English or French, and they have, like, 50% of their people calling in on both. And having the AI that's able to, like, you know, understand that it's whatever language it is and then switch the language is very valuable to these, like, contact centers and call centers that don't have to hire, like, people with this specific skill set. Especially when, like, the pain point may be really high. So when it may be out of hours or when it may be like, you know, midnight to, you know, that graveyard shift. That's. That's the. That's the exciting part there.

Demetrios:

And where are the agents? Falling flat, I think.

Will Bodewes:

So from what we've seen in our use case, and to be fair, we're not, like, focused on this a ton. Like, the sales use case, we haven't seen that nailed yet. Like, it can. It can kind of go back and forth and it can do some of that stuff. But I think that there's a bit of, like, you know, humanness that can. That can end up doing just, like, a better job than these AI models when it comes to that. I think it's going to be something that will continuously get better, but I think that. I think that that's one of the areas, you know, a lot of these agents we're seeing is, like, prone to hallucination and prone to, like, you know, getting, like, when.

Will Bodewes:

When a hallucination happens, everyone thinks it's like the AI makes up some, like, crazy fact that's, like, trying to end humanity. But it's. It's really not that. It's usually, like, when they're asking to reset a password, you know, it will say, like, oh, you click on this button instead of a. Like, that button just doesn't exist because it crowds from the context. So I think that those are problems that are still being worked on and still being solved. We've got a pretty good, like, optimization method for that, but it's. It's definitely still a challenge.

Demetrios:

Sweet, dude. Well, this has been an awesome conversation. I really appreciate you coming on here and chatting with me.

Will Bodewes:

Yeah, absolutely. It's been a pleasure.