Ep. #1158 - Blending AI and Human Touch

Matt DeCoursey  00:01

And we’re back. Back for another episode of Startup Hustle, Matt DeCoursey. Here to have another conversation, I’m hoping helps your business grow. Alright, so we have run full speed into the era of AI at the end of 2020. To show people we’re using AI for a lot of things. It’s been around, but then chat GPT comes out gets on everyone’s radar. There’s billions of dollars flowing in to AI startups and companies. But what do they need? And what are they missing, it could be the human touch. And that’s what we’re going to talk about today. And before I introduce today’s guests, today’s episode, Startup Hustle is powered by FullScale.io. Hiring software developers is difficult and Full Scale can help you build a software team quickly and affordably and has the platform to help you manage the team. Go to FullScale.io and learn more. There’s a link for that in the show notes. If you’re not aware of that’s my company. We love talking to Startup Hustle listeners. So reach out, we might be able to help you solve some problems. Joining me today is Aaron Lee and Aaron is the Cofounder and CEO of Smith.ai. And you can go to Smith.ai. There’s a link for that in the shownotes, to learn more about what they do. I’m sure we’ll talk all about that and more. Straight out of a little town in the Bay Area that no one including me has heard of that probably should have Los Altos, California. Aaron, Welcome to Startup Hustle.

Aaron Lee  01:28

Hey, thank you, Matt. Excited to be here.

Matt DeCoursey  01:31

Yeah, and I’m interested in talking about this because I think that the blend of, of human and AI is a good mix. Now, before we get into that, how about a little bit about your backstory? Because Aaron, you’ve, you’ve been on some impressive teams.

Aaron Lee  01:44

Yeah. So Smith.ai is my second startup. And before that, I started my company called the Redbeacon, which is a online marketplace to connect the homeowners with their home improvement professionals. And we started, like, I distinctly remember in the month of the financial crisis that was in 2008, September. And back then, all the VCs are like, You know what, there’s no money for you. We said great! That means we could be has done on execution. And by the time we launched in 2009, we took the top prize of TechCrunch, we raise by seven and a half million dollars in the first week, like, the VCs were pitching us, not the other way around. So fast forward, we grew very rapidly. And we launched to nationwide and Home Depot got noticed knock on our door. In fact, they caught our front desk receptionist and tried to reach us and, and the rest was history. And we got acquired in early 2012 became part of the Home Depot team. And today if you go to any of the Home Depot store, you will see my work there. I think they’ve rebranded it as Pro Referral. And before that I was a early Google engineer back in 2004, almost 20 years ago, I was one of the two founding engineers on Google Video. Building the video technology, massive storage. And later on when Google acquired YouTube, I joined the YouTube team build a monetization to search video for essence and all the great products. So, on top of that, I’m also a very active startup investor. Invested in early MongoDB, Facebook, NerdWallet, and a bunch of other startups.

Matt DeCoursey  03:33

So you’ve done a couple things.

Aaron Lee  03:37

A few things. So but really, I mean, like serving SMB, it’s like near and dear to my heart like that was happening both at the Home Depot, and also at Smith.ai.

Matt DeCoursey  03:49

So I don’t have to explain this to someone with your level of experience. But you know, for those of you listening, if you want to build a great business, it needs to solve a problem that people want and need solve that has value. What’s that problem at Smith?

Aaron Lee  04:05

Well, the problem is, if you look at like every day, right, if you need to have a home improvement project, you need to fix some things. You tried to call someone from Yelp or from any other online kind of websites, when you call them they don’t pick up the phone. When you text him, they don’t reply your texts, you may leave a voicemail, they don’t get back to you. And that is the problem because for the 30 million SMB small, medium business owners in United States, they have no time, right, the team is small. Typically, it’s like less than 10 people. And they don’t have a dedicated staff for customer support or customer engagement. And they really don’t have the tech savviness to know how to be with the CRM to build like the fully dedicated customer support functions in order to meet the needs of the demand for the customers. And it was very apparent when I was at Home Depot. Like every day, the pros would call them the pros, they just couldn’t meet the demand. They are the project owner. They are the business owner. They are the one doing the work on your roof at your crawlspace. And but they also are the one doing the bookkeeping and everything else. Like they just don’t have the time and money.

Matt DeCoursey  05:17

Yeah, I ran into that. I’m also the founder of GigaBook.com, which is a scheduling platform. And I built that in a in a similar way. I went to take my dog to the groomer. And I had been calling and I kept calling no one’s answering the phone. I’m, like, Is this place even open now it was opened by a friend of mine. And I figured they were there. So, I went to drive over there anyway. And when I got there, I opened the door and the little bell jangled. And I hear a voice in the back that said, I’m back here. And I went back in the lady was in the back washing this dog that was like my size, and it was covered in bubbles. And so uh, she and I realized why she couldn’t answer the phone. But that offended my sense of business. Because here’s the thing, most people just pick up, they just hang up and they dial the next person on the phone. In the phonebook, which it doesn’t isn’t a real thing. If you’re using the real phonebook, please don’t. But with that, you know, like it says on the on the flip that right here on the header at Smith.ai, responsive businesses when more clients and that is the truth, and I think that so many people, so many small business owners, I’ve just experienced this with GigaBook because if you talk about being able to schedule an appointment, or doing any of that, they’re also like, you look at like a massage therapists, like a massage therapist can’t stop providing services in the middle of massage to go answer the phone. And if they did that, you probably wouldn’t go back. So you got to find a way to do it. And yes, businesses in general are pretty bad about it. I think that also a lot of when you look at like what we’ll call a blue-collar-type-service provider, you know, a carpenter isn’t usually also an expert at embedding a booking widget onto a website or something like that.

Aaron Lee  07:02

So there’s the blue collar, if you think about lawyers, right? I mean, when caught, they cannot legally answer the phone, right? I mean, because it would be very disrespectful to the judge and the jury. So these are many examples. We, we have so many different industries from like, like property managers, real estate agents, like, lawyers, and like general contractors, there’s so many variety of the people that want to deliver that kind of very responsive, when people call you, you pick up the phone, when people text you, whether it’s SMS, or Facebook or web web chat, they want to respond it. And in today’s era, I think people are expecting that type of like instant gratification, that when you ping someone, people will respond.

Matt DeCoursey  07:49

Well, that’s an Amazon effect. And you know, Amazon, in my opinion, has been changing that for years, you know, and, and you know, now I order stuff, it’s there the next morning or even the same day, and that that’s really kind of boosted everyone’s expectation for call and return time. And also, there’s just too many ways to communicate with people to not be effective at it. And yes, you are, in fact losing a ton of business, if you aren’t answering your phone, or just getting some method of record with that. Now, you know, what we’re talking about is blending AI and a human touch. How do you guys do that?

Aaron Lee  08:24

Yes. So let’s go back to like 2015 2016. When we first started the company, we said, look, this is a gigantic market. This is a very underserved market. And the technology which like both me and my Cofounder, Justin Maxwell, we’re technologists, we said, there’s got to be a better way to do it. And we thought back then the AI would be there in a few years. And boy, we were wrong for seven years, until I would say eight months ago, when OpenAI launched the ChatGPT. Right. That is when you see the groundbreaking game changing route, when the large language model actually understand what you’re talking about, they can actually respond in a very human way. That is very different from all the AI technologies that we look at before, like, by the way, we look at many different AI technologies from like the Google dialogflow to Amazon lakhs to IBM Watson. Many of them they’re still in the old stage of AI called the rule based system. You define the intent, you define, like the variations of the different ways that you say things. And you come back with the response. And when you say the response, you actually need to map the response to in your scheduling software, it could be the date could be the time could be the reason for calling is very mundane. It’s not scalable. It’s definitely not like widely deployable to many of the businesses and that was to timely say, Wow, I mean, that was the game changing thing. But if you think about the background, we started with like having like aI being the core of the company. Many, we do have a lot of AI technologies, but not at the scale that they can add the understanding of the human language. But in the past eight months, that has been changed.

Matt DeCoursey  10:13

So when you guys go and do this for let’s just say I’m a carpenter, and I’m on a roof, I’m building the frame, I can’t answer the phone. There’s specific questions that I mean, do you guys go in and help tailor this for the actual is that the true human touch? Like, because I think that an effective business owner that’s done the business for a little bit, you’ll find that there’s always like five questions or 10 questions that everyone seems to ask. That would seem to me to be the logical approach for what we train the AI to be able to answer that was a little unique. You know, Is that Is that how that goes?

Aaron Lee  10:52

Yeah, that’s exactly right. So if you look at like the way you talk to your, your Amazon Echo, or you talk to your Google Assistant, you can talk about anything you can talk about, like, Hey, play me my favorite song, or like, what’s the weather tomorrow? But if you talk about like the B2C, right? Yes, we do have different industry playbook for different, like, verticals, like, legal, real estate, general contractor carpenters. But in summary, in essence, they all fall into this bucket about, okay, are they the new clients? Do they qualify for my business, like, kind of like need, like, meaning? Like, are they a good client? For me? Are they too far away? Are they are they are they close enough? Is that the job that I wanted to do? For example, if you’re a commercial painter, you don’t want to do or you don’t want to take any jobs to residential painting because there’s just not your, your area, right? Or there’s jobs that are too small, you don’t want the waste of time to drive like, you know, two hours. So these are the fundamental questions. In the B2C we call lead qualification, can we qualify the lead so that we save you, as a business owner, that time to qualify the lead, book an appointment and schedule it on your calendar versus with, like, the traditional call center will just like take a message or transfer the call to you. So that is the biggest thing that when we do we call it onboarding? We actually talk to you, we understand your business. Now we have done it like tens of thousands of times. So we know what kind of questions and we already handle millions of conversation. So we know what people are going to ask. And we just tend to run in our Hey, Matt, like, these are the questions that we need you to tell us so that our receptionists can be answering the question right away.

Matt DeCoursey  12:41

So here’s one of the things for those of you listening and I you know, for years of GigaBook appointments and stuff like that, I will tell you that wants for most people that are calling or inquiring, once they get something scheduled, they quit looking.

Aaron Lee  12:58

Absolutely.

Matt DeCoursey  12:59

That’s what you want. That’s what you’re looking for. If you can’t answer the phone, you want them to quit looking and wait for you. And then that’s that’s the hardest part. I mean, and one of the reasons that you know, and so you talk about so we built GigaBook way before its time Calendly had just come out. I remember when people I remember kind of laughing at their model. I was like, how is this gonna succeed? Now it’s a billion dollar company, you know, hey, you know, shit happens. But, but with that, and one of the things so I get offended, I have a sense of, of efficiency that often gets offended. And you also look at some of these tools. So if you’re a small business owner, you’re trying to streamline things, and maybe you can’t afford to pay someone to sit there and do all that stuff manually. You know, you look at how much time is wasted in email threads where people like, okay, so people have gotten a lot better about booking links now. But like three years ago, I would send people I’d be like, hey, click this link, pick a time and sometimes people would reply, they’re like, Yeah, I don’t do this kind of stuff. And I’m like, you just need to pick a time and put your name in, you know, and, and there’s a level of adoption. Now, I almost think now when I communicate with people, if they don’t give me a booking link to to click, I feel that’s kind of rude. But you look at all the time that goes back. You’re like, well, I’m available next Wednesday at 1 pm. Oh, well, that doesn’t work for me. What else do you have and like, and there’s this time arbitrage that occurs in and around your business that gets really expensive. And then the missed opportunity. Now, for so many businesses, there’s businesses that provide maybe a one-off service and you know, even more of a reason to get those people on the calendar. But it’s so hard, I think, for a lot of business owners to really know and understand what’s the opportunity cost that you’re creating by not being responsive and, you know, for businesses that have have ongoing services and keep clients and don’t want to turn on, that’s high. I mean, at Full Scale, you know, our average clients spending about $20,000 a month. Missing one of those opportunities for just an average sized client. If we miss four of those in a year, that’s a million dollars a year in revenue.

Aaron Lee  13:05

That’s exactly right. If you look at about, like, what is our ICP, the ideal customer profile? are the ones that high opportunity cost or high hourly rate lawyers right, like, time is money, like, they recognize that.

Matt DeCoursey  15:26

Also with an attorney, though, like there’s certain things that’s what I was saying, like the stickiness like most people don’t want to switch accountants or attorneys or whatever. So, yeah.

Aaron Lee  15:36

Yeah, I think what you’re hitting about is not that people don’t think that they’re important. They just don’t recognize that is one of the most important things that you could set up your business for success. If we’d had them. We onboard, so many different business, the first thing we asked was like, Do you have a booking link? Do you have a scheduling link? And often they say, Oh, we don’t we, we just write down on a piece of paper or something like that. We say, Look, we will be happy to set you up on any free online calendaring system. Calendly is one of them, Acuity who got acquired by Squarespace is another one. They’re free, right? I mean, if you just have a few people with the basic features, but at least get you connect your calendar, and most people say, Oh, I would do it later. No, no, don’t do it later. Do it now.

Matt DeCoursey  16:22

Right? Because there is no later there’s

Aaron Lee  16:25

Right? I mean, the gratification for your potential customers to know that they’re secure spot on your calendar, they are not going to look elsewhere. Yeah, that’s the beauty.

Matt DeCoursey  16:36

Well, some of this to you, like, you look at customer service in general. So let’s just say it doesn’t even matter what you do. someone fills out a form. If they get a response that isn’t just a robot, right? It isn’t just hey, we received your farm, someone will contact you someday. But they get an actual response that says, hey, this is now I just looked through your ticket. We’re a little backed up right now. And but I’ll get a better response in 36 to 48 hours, people are gonna usually be okay with that, because they got an actual response to government acknowledgement. If you don’t have that step in there. They get they get angry in a hurry.

Aaron Lee  17:15

That is where the human touch comes in. It’s a lot of them to say, Yeah, I need someone to come by my house like yesterday. Yeah. But what they really want is I want someone to come to my place as soon as possible. Now, if you talk to the AI and say, Look, we don’t have any appointment today or tomorrow, but the AI doesn’t have that emotional intelligence, or the compassion to say, Well, Matt, wow, that’s, that sounds really bad. Like your water, water Heater just broke and you’re flooding your garage. Yeah, let me get someone to you ASAP. Like, today is going to be tough. Are you available tomorrow, like after evening like because like our technicians are fully booked up, but your case is really important to us? Like we will try to squeeze you in and get you an appointment tomorrow at 6 pm. I think that, I call closing, is really important. AI, it’s good for FAQ, what are your opening hours? Do you open on July 4th? But when it comes down to like, converting the customer, or or like, like, caller to customers, that emotional intelligence is really important.

Matt DeCoursey  18:18

How far away? Do you think we are from a world where there is a little I understand that the nature of the machine doesn’t give it emotions? Now, you know, well, let me back up for a second. So I’ve been using AI and AI everything from like, Okay, so my first two books that came out, and that was years ago at this point, that’s 2017. I use people are like, so I wrote two books at the same time. And people like how’d you do that? Well, I was using transcription technology at the time, which by the way, was frickin terrible. I mean, it was really bad in 2017, like, you know, but so I that I sped up things. So I would sit there and I would talk on the different cars, you know, and I wanted to keep a voice in the writing. And, you know, some I’ve had a couple of others say, well, is that cheating? No, it was still my thoughts. It was just my fate. I can speak a lot faster than I can type. And, and it was and it was bad. In a lot of cases, we had to correct a lot of stuff. And then you see it kind of move forward. I think the problem that people had with a lot of AI and a lot of AI tools was it could get it 95% of the way right. But no one looked at the 95% they looked at the 5% where it was wrong. And they’re like, this is the worst thing I’ve ever seen. It’s terrible. It does this What’s intelligent about this and, you know, like, for anyone that builds software, like when people have talked to me about GigaBook, they’re like, it only does about 90% of what I need. I’m like, that’s your wedding, dude. That’s amazing. And they’re like, What do you mean? I’m like, that’s like huge. So you kind of got to focus on what you got now with with GPT And I know that’s driven by OpenAI, but it is definitely at a different level because it does sound very intelligent. Now you translate. Now, I don’t know what your driver is, if it’s your own AI or whatever, but that also is verbalized, like, does it speak?

Aaron Lee  20:17

Yeah, it does. So let me tell you, what are the biggest limitations of AI today? So the two big problems number one, if you go to ChatGPT, you type of questions like, you know, tell me more about the founders of Full Scale IO, or tell me more about the founders?

Matt DeCoursey  20:33

If you do that, but it’ll give you the wrong answer.

Aaron Lee  20:35

Right, exactly. That’s the number one is AI hallucination. Like, they don’t know what they don’t know. But they will probably speak in a way that is so confident that sometimes you believe the information that is provided to you. Number two, when you see that AI or the ChatGPT are actually spitting out the words in, like, almost like they’re thinking about, they’re thinking about the next word, that is actually not JavaScript, or any kind of animations. It is because they are a gigantic statistical machine that subject the next probable word. So they don’t really have an understanding. They just know, like, the next word is going to be the most likely one, and they will spit it out. So that means if you want to deploy in a voice application, there will be a latency between five to six seconds for the complete response, right before you can actually convert from text to speech. Now, we all know that five to six seconds is like eternity. People say, wait, are you still there? Right? And when it’s not there, people start asking the question again, now you’re talking over each other. And that makes the problem even worse. Now, if you believe in Moore’s Law, that means every year we’re going to speed it up by 100%, you go from six seconds to three seconds to one and a half. And that means in a few years, you can get to sub second kind of latency, which is great, because now it’s becoming interactive. Instead of, like, you ask a question, you wait for five seconds, and you get back a response. Now for the chat bot, you can kind of get away because you can have these like in a blinking light, or like typing indicator that kind of like smooth it out. But for the voice, I think that is difficult. Now on top of that human has a lot of like contextual knowledge that the AI doesn’t. So unless the AI can train on the content is specific to your business, then it can only answer the questions based on the billions of texts or articles or books that they read. So that’s another problem that they need to overcome. And you can start seeing like startups, like, building technologies, like the vector database, and some of the stuff that can help business to do that. Now even with that, like, you in the high tech area, right, deploying that in the production scale, is challenging, it’s not easy. So there needs to be some layer, the people like Smith.ai, that can democratize the technology so that if you are a small business owner, then you can just use it, you don’t have to understand you don’t have to learn about how to configure all the systems.

Matt DeCoursey  23:10

Yeah, and that’s important. Now, I want to talk a little bit about the responsive nature of that. Now, before I do that, I want to remind everyone that finding experts, software developers doesn’t have to be difficult, especially when you go to FullScale.io, where you can build a software team quickly and affordably. You can use FullScale.io platform to define your technical needs, and then see what available developers testers and leaders are ready to join your team. FullScale.io to learn more, there’s a link for that in the show notes. Click that link, link to Smith.ai as well you can learn more about Aaron’s company, what they’re doing. Now, hy don’t wanted to go back to you know, a few years ago, there was this this clip that kind of went viral. And it was Google’s, it was a Google AI. Do you remember this where it was caught, it was calling, it like called and ordered some Chinese food and that in another place it like scheduled an appointment. And I remember this was a two or three years ago, and I just remember going wow, like I couldn’t tell, do you? Do you remember? Do you know what I’m talking about? So? Well, with that it sounded very much like a real person. Because you know, like, they’re calling and saying, Oh, can I um, I’d like to order or you know, have uhms and buts and stuff like that. And I don’t think I could have been that response was instant. And that sounded very conversational. I wouldn’t have known the difference.

Aaron Lee  24:32

Yeah, I think that was a very, very nicely curated demo. Like they they launched that in the Google IO. And I think like shortly after that they launched a product called CallJoy. And unfortunately, that was like actually designed for SMB to convert the callers to customers by booking an appointment for them. And unfortunately CallJoy kind of folded like after two years because it actually didn’t work. Now, if you read the fine print, they will say, okay, the technology works 80 to 90% of the times. What you don’t recognize is, if not 80, or 90% of the call is within the call, they can only understand 80 to 90%. That means every single call, there’s an error rate of 10 to 20%.

Matt DeCoursey  25:21

And that’s pretty hard for that.

Aaron Lee  25:23

That is very, very high. Because if you’re a business, you get the 100 calls per week, you’re losing 10 to 20 calls, it almost feels like that system is making a mistake, every single day. And I think that was a gap in understanding that. Now, if you can fine tune the model for one single business, I’m sure they can get to 90 or 95%. But the technology doesn’t scale that way, unfortunately.

Matt DeCoursey  25:48

You know, what? You mentioning that made me think. So I have two Teslas. They’re both self-driving. And, you know, that’s not as common here as it might be in the Bay Area. But I’ve had so many people ask me questions about it. And they’re like, well, does that seem weird? Well, it does for like, the first two days because you have to learn to trust it. And I feel like with any AI, especially like, with what you do, or as a business owner, like, let’s keep in mind, like, it’s that it’s that like that old example, that a happy customer might tell a couple people and upset one tells dozens. So you know, people are going to hear business owners are going to hear about the horror stories that they had, rather than the good ones. And yeah, and so once you realize that the Tesla is going to stop at the stoplight, and then in some cases, it may, it’s kind of weird to because it actually stops short of the stop sign, like, the law says you should. And at first you’re like, what is this thing doing? It’s not even all the way to the stop sign. And you’re like, oh, yeah, you’re supposed to stop in front of it. But I think that that’s I would imagine that that is an important part of this kind of stuff with anyone’s business. Because if you don’t trust the technology that you’re using, or you don’t trust the systems that are in place, or you find yourself explaining them to people a lot, or cleaning up the mess that they create, it becomes a negative or a zero sum kind of thing. And if you’re not picking up efficiency business, bookings. It’s, like, one of the things I learned with GigaBook as well as the scheduling component itself is worth a little bit. But if you could actually find the business and then schedule it? Oh, man, you go from like $15 a month to hundreds of dollars a month. And then I also unfortunately, learned that not enough business owners understand that saving money is making money. There’ll be more focused on the revenue growth in some cases, but they’ve got this massive inefficiency. And that could be with anything from customer service and missed opportunities. So the whole nine yards. So they’re overspending on the backside. So sometimes, a couple things. And then maybe the best part of it is just a little peace of mind. Like most of all business owners, especially when it comes to appointment booking and scheduling. And this is where whether it’s GigaBook or Acuity or Calendly, get something going people because people don’t want to talk to you. I’ve had people that are like, oh, they like my personal touch. No, they don’t they want to schedule an appointment and go back to doing whatever they’re doing. I think it’s possible once at once they do at one time, they’ll understand it forever.

Aaron Lee  28:22

Yeah, I think what you’re seeing is a lot of the business are having a very large leaky bucket. Right. So a lot of marketing agencies, they always complain to us and say, Look, we’re doing good work for them. But the business owner is not happy. So why stat because every time we send a click or call to the business, they don’t pick up the phone. We don’t respond to the click. And so like they said, Oh, the marketing campaign wasn’t effective. But when we work with the agencies, we can say look, you don’t have to ask your client to staff a 24/7, like, always on receptionist or like inside salesperson. We can combine with your campaigns so that every single call or every single click to your website, we can catch them, we can convert them, we can qualify them. That’s one thing but it’s funny because I have two Teslas and that we’re still in the autopilot stage meaning you don’t feel comfortable. You want to put your wheels, hands on your whales because you’re worried that they may not do

Matt DeCoursey  29:22

I don’t want to put my hand on the wheel. I hate that. I wish they take that off if I had to. So you know if it’s self driving, you have to like nudge the wheel. Yeah, exactly. He gets mad at you. Yeah. And I get in trouble for not looking because it has a little camera above your rearview mirror that makes sure you’re actually watching the road. It’s not fully so it’s not self-driving yet. It’s attempting to.

Matt DeCoursey  29:42

Yeah, 92% wouldn’t get me real excited. You know, there’s, there’s, it’s such an interesting world. Now, let’s just talk about AI in general because I know that’s such a hot topic right now. You know, it’s interesting because I saw this graph the other day, that was, you know, overall, right now for venture capital and private equity that’s pouring in to startups. It is about 20 to one in favor of AI stuff. And as opposed to the rest of the field. How sustainable is that?

Aaron Lee  29:42

Yeah, I think today that’s why you see a lot of the AI solutions to claim that can be 100% fully automated. Like, you still need, like, I call it a human oversight, your supervision or you want to make sure because like you said, if you miss qualify or you lose a call a lead for For Full Scale, it could be talking about $20,000 per month. That is, if you lose four of them in a month that you lose a million bucks. You really want to make sure the system is like 99.9% instead of 92%.

Aaron Lee  30:49

I think this is go back to like, whipping around the blocks, right? I mean, like, early 2000. It was like internet web, right? And then you

Matt DeCoursey  30:58

It feels bubbly is why I asked.

Aaron Lee  30:59

It feels bubbly, right? You get to do the mobile, right? When when when Apple launched the first iPhone in 2007, they get their social media, right. And now we’re at the AI stage. I think there’s a lot of companies that will get funded just because they have the AI in the name or AI in pitch deck, but they probably don’t really leverage the scale. But for us, that is the core of what we do. Because the more AI and the technology that we can deploy, the less that we need the human, right? So we’re the AI first company, but we also say, Look, we need a human in the loop so that we can make sure that like 99.9% of the conversations are done right. So if you think about the venture kind of landscape, that is the time where you will have a lot of investments. And after a year or two of you will say okay, now that everyone has the AI in the pitch deck, what are we gonna do, right?

Matt DeCoursey  31:54

That makes me think so we were out in your neighborhood and 2019 for TechCrunch. And our joke, by the end of the first day, we were just attending, but we went around and you know, they have a main exhibitor hall. And there’s a bunch of people talking about their startups. And by the end of the first day, we were joking because everyone was, like, our machine learning algorithm. And I’m sitting there talking, I’m like, there is no way there is a real machine learning anything. And this. So you know, we’re out at dinner that night. And they’re asking what, you know, what salad dressing, and we’re like, my machine learning algorithm says, I should probably go with blue cheese here. But it was like it was in every, I mean, it was nine out of ten places that we talked to you were pitching some and that was pretty clear that that was a trend that went with it. Because you know, like some of it. I don’t know, it was just no way.

Aaron Lee  32:43

Yeah, I believe that certain companies that will benefit greatly from the AI, right? If you take away the AI, you say is this problem is still worth solving? Are people, or the business people or the business owners paying you for solving the problem? Will adding AI significantly move the needle? So that you can deliver faster, cheaper and more effectively? That is the thing that we need to ask instead of just saying, Oh, I have this AI copy generator, which is like a nice skin on top of ChatGPT.

Matt DeCoursey  33:14

Yeah. Well, there’s a big difference between a lot of types of AI and generative AI and the what Aaron was alluding to earlier, is, you know, when you look at ChatGPT it’s not as, quote, intelligent, as it’s not really thinking about anything other than what’s the most prot, what’s the highest probability word that I should put next. And that through, you know, reading the internet, basically, it’s able to assume that which is also why it usually doesn’t require grammar, corrections and do a lot of stuff. It’s pretty smart in that regard. That’s the thing I like about it actually, is that it gives you you know, it’s not it’s not a goofy response, but it’s just predicting what the next word should be. It’s not truly thinking about it. It’s not like the models that are sitting there playing chess all day or go or something like that, that are actually thinking and having forms of strategy and understand a win as compared to a loss. And you know, that’s a completely I mean, can you validate that that is, in fact, a different approach with AI like ChatGPT is just predicting words.

Aaron Lee  34:18

Yeah, I think that’s why we’re still in the very first inning of AI like that is like kind of the foundation. And I’m sure that people at Google and Microsoft are still building things that will say, can we actually understand what people are talking about? Actually be intelligent, instead of just predicting the best or most likely we’re after this work?

Matt DeCoursey  34:40

Yeah, yeah. It’s been very interesting to watch the news feed over the years like I remember a man it’s been five or six years. But I remember when Twitter was experimenting with something, they put it up and they asked people to, to converse with it through tweets and because people are mean and have nothing better to do it turned it into like a racist hate bot in like two hours, you know, and they turned it off. And then they had another one where they had two AI that were speaking to each other and invented its own language amongst itself to turn it back. And remember, they the engineers kind of freaked out and just basically unplugged the thing. They didn’t know what to do with it. And you know, we’ll we’ll see where all that goes. It’s definitely a very interesting horizon, as far as AI in general goes, Do you think that that, that, that like, as mentioned, the generative piece? Like, I mean, how influential is that going to be for startups in general, in the next 10 years?

Aaron Lee  35:39

I think it’s a game changing technology, you will see that like, you start looking at the content, we call like, garbage in, garbage out, right? If gorgeous, in gorgeous out. If you can fit the language model with good content. Now, even in any companies, I’m sure, like, we use Notion and you may use other kinds of like knowledge base kind of like management system, a lot of the times their contents that are outdated, a lot of the times that are contents that are contradictory. Now, if you fit that into the system, no matter how good your generative AI is, it will generate a confusing response, or it will give you the wrong content or the wrong answer. So that is the part, I think, like you need to combine, like good content or good knowledge base with the language model so that it can converse with your team or your support agents or wherever that needs to consume the content. But in the next slide, I will say five to ten years, I think there will be technology that can say, wait a minute, like we found these two articles that are telling contradictory information? Can you go and clean that up? Or we find that like, these are the two pieces of content that are like 80% similar? Which one is more updated? Can you combine them? Can you merge them? Can you consolidate them?

Matt DeCoursey  36:55

You mentioned the garbage in, garbage out? I think that’s the perfect buzz line for the human interaction. I’ve talked about this a lot on the show and other episodes that, you know, AI models, they, in many cases need, especially early, need to know what is right and what is not. Because they can get real haywire. I can’t remember the episode but I was talking to one guy about in some cases, if that doesn’t occur, you literally have to reset the whole model and kind of start over from scratch because it learned the wrong way. It’s the same way as a child that grows up in odd environment, you know, and it’s like, I don’t know, I mean, there’s people out there that have a lot of interesting beliefs about a lot of interesting things. And they’re entitled to certainly do that. But why do their kids often have that same belief? It’s because that’s the way they were programmed from when they were young. So

Aaron Lee  37:48

That’s exactly right. So if you think about that, on top of that, it makes it even more challenging if you have business, because now you know, you need to learn how to ask the right question that they caught the prom engineering, how can you write good questions, a good pronoun is going to be the hard part that is going to be a hot pot, because knowledge might be there, the answer might be there. But if you did not ask in the correct way, you may not be able to get the answer.

Matt DeCoursey  38:16

Well, there’s a you know if you call it so right before recording this, I actually booked a call with an electrician to come out to a property I’ve recently purchased. And you know that the there that the lady was great at answering questions, but was not very good at selling stuff. And that’s a completely different approach. I think a lot of times people are good at answering like you mentioned the knowledge questions. But remember, you want to do business. And some of these are simple things like do you want to book an appointment? When can we come out and give you an estimate? You know, like, are just simple things. And I mentioned it’s probably a good time to remind everyone that today’s episode Startup Hustle is brought to you by FullScale.io. We do a whole lot of stuff, folks, we can help you with some AI probably not as much as as some places because it is still emerging technology. Once again with me today is Aaron Lee, Cofounder and CEO of Smith.ai. Go to Smith.ai. Link in the show notes for that. Aaron, here we are. I told you the I told you before we hit record that these things go kind of fast. So we’re here. We’re here at the end of the episode, this episode was not created with AI. I’m still looking for a solution to replace myself on the show. If that’ll be the case, I’m willing to give it a shot. But you know, it’s time for the founders freestyle, which is my guest’s opportunity to recap some of the things that we talked about or maybe things we forgot to talk about. I mean, what’s what from this conversation or what I mean what what stood out what are the important things that matter for people wanting to get into this space, whether it’s with Smith.ai or just in general?

Aaron Lee  39:58

Yeah, I think if you’re running I’m building a startup, one of the most important things that you need to understand is, like, that early customers matter. You need to capture the voice. You need to understand the pain points and every interaction that they make with you. Because you’re operating at a smaller scale, it’s important. So at Smith.ai, we have inbound, which is like answering services. We also have outbound that we can actually call to our customers. I’ll give you an example. So I think last year, we were doing an annual campaign, which means, like, if you pay or prepay a year ahead, we’ll give you a discount. So we call our customers using our team. I think on the last day of the year, probably the worst time to call people, right, 31st December. And we closed more than a million dollars for the commitment that people wanted to pay. Because it shows that if you are willing to get the human touch to the people, you’re willing to explain why this is beneficial for them for tax purposes or cost savings, and whatnot, like, people will respond. And they will respond quickly because when you talk to someone on the phone, they’re giving you their undivided attention. So, things like that, don’t forget about the human touch, even though I know every day you open your LinkedIn or Twitter. It’s like 99% of them are AI or AI-related stuff. The human touch is really important. So don’t forget about that. Number two, AI is rapidly evolving, whatever things that cannot be done today, it may be viable tomorrow, next year, right? So I would say keep in touch with the technology, look at how other people are using the technology, how other companies are leveraging the AI, like, both on the people side, the process side and also the, like, sometimes, like even brainstorming, I think that is important.

Matt DeCoursey  41:52

For my freestyle, I got a couple things here. First off, missed opportunities for me are the ones that always feel the most painful. And sometimes you don’t even know you’re missing them. You know, I think that a lot of people still say things like I want to talk to a person or I don’t want a robot, maybe try the robot, figure it out. You know, for those of you that that run a business and you say, well, we’re not really a technology kind of place. Why not? You learn how to be a carpenter, you learn how to do something else, like you might have two things that are worth doing and worth pursuing require a little bit of effort. The good part about tools like this in your business is usually when you get it right. It’s right. And the thing that I love, you know, people ask you have asked me, they said, well, I’m trying to sell my business as a service company or something, and they’re like, I get this shitty multiple. Why the software gets so much software shows up to work every day, it’ll work 24/7 for you. It’ll answer the phone at 3am. If it’s done well, it doesn’t have sick days, it doesn’t have a lot of problems. So therefore it is very scalable. And if you want to grow your business, like keep in mind, like so I think that there is there is a value ratio that it equals an infinite, infinite sum. And that’s what is your peace of mind worth. And I think that that’s the biggest problem that I think a lot of business owners know in the back of their head that they need to be doing something better or different. They either just don’t want to do it, because it required, it’s not their favorite thing to do. You can solve a lot of this with modern technology. I think AI in general has never been more accessible and reachable for folks alike. I love what you’re doing, especially for the group of people that you’re doing it for, because those are also jobs that are very difficult to fill with the right people. You know, and that’s, that’s the thing, too. And what I love about these kinds of solutions is when you talk about scalability or whatever, let’s talk about uniformity. You get to answer the same question again and again and again and again and again. And that’s a big thing. So, man, I really liked what you’re doing. Keep it up, keep it.

Aaron Lee  44:10

Yeah, thanks, Matt. I appreciate that.

Matt DeCoursey  44:12

Yeah, if you want to learn more, go to Smith.ai. A lot of great solutions if you want someone to help you build a software solution, FullScale.io we’re there for you to. Aaron, I’m gonna I’m gonna check up with you down the road and we’ll see what kind of advances and solutions we’ve come up with compared to what we’re at today.

Aaron Lee  44:30

Awesome. I would love that.

Matt DeCoursey  44:31

Yep. Thanks for joining me.

Aaron Lee  44:32

All right. Thank you so much, Matt. Take care.