Ben Parr discusses the future of AI and its impact on startups

Ben Parr of The AI Analyst - Podcast Transcript

Healy:	Hello and welcome to the Kruze Consulting Founders and Friends podcast. I’m your host, Healy Jones. I’m joined today by Ben Parr, a well-known AI founder, investor, writer. Dive into a lot of the cool trends that are shaping AI and we’re going to learn from him what it takes to run a profitable AI company, which I would say amazing. Not too many of those, but first, a quick word from our sponsor. Hey, this is Healy Jones, VP of Financial Strategy here at Kruze Consulting, andI want to say thanks to our podcast sponsor, ARC. At Kruze, we’ve got a number of clients successfully using ARC to manage their deposits, payments, access financing, all in one place. One of the things that ARC provides that’s really great is over a quarter of a million dollars in FDIC coverage. Their insurance program goes beyond the standard limit and it secures up to five and a quarter million dollars. So, startups that have even more cash than that can go and access treasury solutions to provide yield and safety. If you’re a startup looking for a secure financial solution that can help you scale, please check out our sponsor ARC at ARC.tech. . And we’re back. We’re joined by Ben Parr. Ben was the co-editor of Mashable. He’s the co-founder of an AI company that I guess we can maybe still call it a startup, even though it’s a profitable business now. He is an active investor in AI. He writes a column on AI for the information, and he has his own newsletter called The AI Analyst. Ben, how are you doing?
Ben:	I am great. I’m now thinking about what the actual definition of a startup is. Is it based off of the valuation? Is it based off of the number of years it’s been around? Is it based off of both? So, what is or isn’t a startup? I feel like I’ve had that question asked to me over and over again 1,000 times and I still don’t have a framework for picking it, but if you’re a $1 billion company and you did it in two years, are you a startup? I don’t know.
Healy:	That’s a great question. I think about it in terms of where are you taking risk when you found a business? So obviously, someone who starts a dry cleaner, it could be a startup business, but they’re not taking business model risk in the same way that someone that starts like your company that you started where no one was using AI for anything back in 2016 really. So, you took different levels of risk. That’s how I like to think about it.
Ben:	So then that gets into another layer, which is like I for sure consider anyone who’s like, you’re starting a dry cleaning business, you’re opening a restaurant, you are an entrepreneur, period, end of discussion, but when what they’re starting also be a startup or does it not count or is startup exclusive to the domain of venture capital tech? Here we got into the philosophy. This is not even what we’re supposed to talk about, but I like going on tangents.
Healy:	All right. Yeah, nice. Well, it’s a good tangent. I’m going to dial it back and let’s talk about your startup Octane, which again, is it a startup or not? We’ll call it a startup for the purpose of this conversation, but you started it back in 2016 and it was an AI business in 2016. What were you thinking?
Ben:	We’re thinking that the future of communication between consumers and brands was going to be not through websites but through conversation and through chat, and the one way you facilitate that is through AI. My co-founder Matt Schlicht, who had a product at Ustream and was an Y Combinator alum in the early days alongside in the same class as Coinbase and Instacart, he was in China years back, and he realized that back then 2015, that almost all the conversation and almost all the business happened over messaging apps and WeChat and realized that would eventually come stateside. We started building a business, especially when Facebook first launched the Messenger platform, which we built some of the earliest chatbot technology. We have multiple patents in our name on chatbots, interestingly enough, long before there were large language models to make them actually good, and we had to hack together stuff with neural networks and NLP and all sorts of stuff, and it was serviceable and lots of celebrities use our software like Lindsay Lohan, Kiss, Aerosmith, Jason Derulo, Rick Ross, I have stories to the moon.
Healy:	Your user conferences must be pretty fun then.
Ben:	Back in the day.
Healy:	I imagine the bar bill gets a little out of hand.
Ben:	We never got to the point with that product of having user conferences. That would’ve been fantastic. The problem was the technology wasn’t quite there and celebrities don’t pay, but e-commerce, our software was working really well for them in printing money, and so we focused on building a business and we have built over the course of nearly a decade of pain and suffering, a profitable AI business, which I don’t know how many of those you know, Healy, but I can’t think of too many off the top of my head.
Healy:	I can think of one and I’m talking to him right now, so there’s not a lot at all. What does it take to have a profitable AI business? What does that even mean? I understand the metrics there, but how do you do it? How do you do it?
Ben:	Well, okay, a couple of things. One, overnight success of a decade, the rule of number one is you don’t die. You have to not die. Now, if you can’t find any product market fit, that is one thing you need to be really lean and experiment on product market fit before you try to raise more dollars or you try to do more things. You have to be like, does this really have product market fit constantly? And then once you do find that, it’s a lot of can you… The AI part only matters so much. In the end, customers are looking for something that makes their lives better and generally, my thesis and my co-founder’s thesis on AI deals for example, is that AI will allow us to build businesses, everyone to build businesses that are better than their non-AI counterparts. I really o think that there will almost certainly be an AI powered business, whether it’s an existing one or a new one that hasn’t been started yet that will power almost everything in almost every vertical and AI being a very broad term, and that’s a whole different tangent, but we found a product that our customers really love being able to build quizzes and do product recommendations leveraging the AI that we built for we’re 3,500, 4,000 e-commerce brands. We’re growing and these brands love our software and use our software and it makes them money. It just also so happens to leverage AI to do the product recommendations and to write the questions for the quizzes and that sort of thing. And so, the AI, the best kind of companies are going to be ones where you interact with it, but you don’t even really think about it being an AI company. It’s just better than the alternative and we’re going to get there long-term with you don’t think of Snapchat and Uber as mobile app companies, but really they are. That’s like they don’t exist without the Apple and the iPhone. It is going to be the same thing with AI in a few years, but right now I think we are starting to see the great experimentation era of AI and there’s still a lot of businesses to be built and it’s long-term too. You could still build a huge business on top of the iPhone. Same thing with AI. It’s going to just be a core fabric same way that almost every company you start now is a internet company.
Healy:	I’m a founder starting an AI company. I aspire to having a business that’s leanly financed like Octane was, that’s operating and generating profits. Other than some of the advice you just gave, what other advice would you have for founders in terms of how they can achieve a profitable AI startup?
Ben:	Hiring is hire as little as you can. Especially a few years ago, the metric of success or the vanity metric was like how many people are on your team? And I still get that question and you want for your ego for something else to be like, I’ve got 150 people, and I can say words and they’re all like, ooh and awe because as a founder, you see that. You see that. Now it’s not a really useful metric, especially now. It just burns money if you hire a lot of people and you don’t, it just burns money. You could do things a lot leaner, especially leveraging AI, which is probably the second one, which is like only hire when you absolutely have to and leverage AI internally. AI for sales, AI for marketing, build your own internal AI tools, AI for ops, there’s new tools being built and tools that you can build internally that will make your business a lot more efficient. And if you could resist the idea that you need to hire a lot of people to be popular to build a business, then you can build something really great. It’s just in the end when people raise money, they end up hiring a lot of people more than they actually need, and then it’s really hard to scale back after that and instead you could keep more of that money, scale really thoughtfully and build a bigger business over time. Because the number one rule again of startups is if you’re not dead, you have a chance. Think Pinterest, think Roblox. They were doing nothing for years and then they went boom, overnight success, but it was overnight success over the course of eight years for each of them. Something ridiculous like that.
Healy:	I strongly agree. We’re seeing a lot of our companies that are maybe not AI companies but traditional software service businesses trying to stay lean and mean here. So they’re raising real amounts of money, but they’re not ramping up headcount like crazy. And the other thing that we’re noticing is that a big slug of our clients, over 60% of our clients are paying OpenAI for something and they’re also paying GitHub copilot and the people are purchasing these tools now. What are the tips for helping yourself and helping your employees understand where they could try to use AI instead of just going and hiring another person?
Ben:	So we teach our entire team how to leverage AI. In fact, my co-founder and I teach a course on AI called Coworking with AI. It’s on maven.com and we teach this course to teach executives how to leverage AI for greater productivity. We teach things like optimized prompting and so the biggest thing for getting a team to be more efficient with AI is to have them use it every day. A lot of the problem is just habit formation. If you just use AI, you have a tab open with your ChatGPT every day, you’re going to use it, you’re going to get better and if you rely on it a little bit more, you are going to end up with a better and you’re going to end up doing more and being better at it just like anything else. We taught our team how to leverage AI. We taughtour team how to do optimized prompts. We ask them what they used AI for, and the end result is they think in terms of AI, they think about how they could leverage us to write content or to make things. Some of them make GPTs. A lot of it is really just getting into the mindset and getting them to use it and asking them on your calls, what did you use AI for? Even if it’s for personal stuff, the habit formation really matters. And then once you get to that point, then the team starts thinking more about how could I build something or a specialized use case or find a specific tool? Because there’s lots of amazing tools beyond a ChatGPT or a Midjourney or a DALL-E 3 that can make your life a lot easier if you just look for a few minutes.
Healy:	Very cool. So that leads me to a question. As someone who’s an analyst of AI, we see there are these different competing large language models. I’m hearing a lot of people ask questions around where’s the value chain going to lie? Is the large language model going to be the place where people make a lot of money or is it something… Where are the real business opportunities here versus some underlying infrastructure that gets competed away in terms of margin and potential?
Ben:	Everywhere. I hate the discussion, not hate, the discussion of which part of the layer, the answers both. Apple, again.
Healy:	Nice.
Ben:	Accrued a lot in Google, a lot of value on the app store layer. That’s just two companies. There’ll be more companies than that for the AI layer. And then there’s a bunch of stuff built for analytics for mobile apps and things like that I consider part of the infrastructure layer and then there’s thousands of apps on top of the iPhone that a bunch of whom have become multimillion dollar, multibillion dollar companies. It’s the same with AI. There’s definitely opportunity on the infrastructure layer. It always starts with that. That’s why OpenAI and Anthropic and these other LLM companies are getting such nice valuations. They are infrastructure for doing this, but there’s a lot of interesting applications in different places, legal and in accounting and in [inaudible 00:13:21]. I’ve seen everything and they’re real businesses and they have technology that they’re building on top of it. Same way here you build on top of AWS, you’re not going to build a server, but you’re going to go and build a lot of other technology on top. It’s the same thing. You might have a unique data set and then build a bunch of AI on top of it, and it’s not just generative AI, which is the topic of the day right now. AI is such a big world and I have slides in some of my other decks where it shows the AI, it’s in actually the trends deck that I put out. AI is such a huge ecosystem. Generative AI is a tiny portion of deep learning, which is a small portion of machine learning, which is a small portion of AI overall.
Healy:	So you’re basically saying you think there’s enough opportunity there. This is so transformational that there’s a lot of good places to benefit, including you’re optimistic that some of these big models are going to get something done. I think this is the chart you’re talking about in terms of the just overall market map and the size, and we’ve got transformer models which are inside of generative AI inside of deep learning and neural networks inside of machine learning inside of… It’s…
Ben:	Yeah, exactly.
Healy:	Pretty amazing. Yeah. Awesome. Cool. Well, so I know that you like to invest some in AI businesses. I’d like to learn a little bit from your perch what you are seeing are the most interesting places to consider investing in right now in AI.
Ben:	So I’ll say first in the end, it’s always about the fundamentals and it doesn’t matter whether it’s AI or it’s Web3 or whatever or space or whatever it is, the fundamentals matter. Do the founders fit the problem that they’re tackling? Are they extraordinary people? Are they founders that can adapt really quickly? Those are a couple of the things in addition to are they tackling a big enough market? How quickly, what is the traction? There’s a reason why they’re standard questions because they really do matter when it comes to building a business. Just because you put .ai in the name of your business doesn’t mean that you have built a successful business. The competition might even be higher and then when you go at, those are the couple of things that I think a lot about and then gets down to what is this being applied to and is AI the main way that you could apply it, one of the best or most unique ways you could apply it? Because there’re lots of businesses I see where they’re growing at an okay rate, they’re scrapping by and they’ll build a real business. A lot of people can, but is it a venture backable business? And they’re very different things. A venture backable business has to be able to grow really quickly, get to the multibillion-dollar valuations, has to make hundreds of millions in revenue. It’s a very different thing. So that’s like I think a lot about those kinds of things. So, I’m really looking for AI companies that are disrupting historically non-tech industries. I think AI being applied to medicine or real estate or legal or all sorts of areas that you wouldn’t think. Waste management I saw recently. All sorts of areas where you can build, I think real businesses and then I’m thinking a lot about how autonomous agents can transform the world and I’m thinking longer term about AI is definitely going to make it so that you can build a bigger business with less people. I expect. I made a prediction that got a lot of flak. I thought there would be a three- or five-person company worth $1 billion in the near future, and they were using AI for everything else. I think Sam Altman tried to one up me and be like, we’re going to be a one-person company worth $1 billion, which I also believe is… It’s not even that far of a leap. There was like 12, 13, 15 people at Instagram when it got acquired for a billion. That’s over a decade ago. So, there’s just so much to do.
Healy:	I definitely agree. We are seeing clients come to us or that are successful raising funding that are applying AI to different verticals and it’s actually amazing the verticals that we’re seeing. It’s not just the standard, even the easy ones that you would think of like legal or accounting or something like that. There’re all sorts of other verticals that it’s being applied to.
Ben:	Anything where there’s a lot of data or there’s a lot of repetitive processes can be greatly enhanced by AI.
Healy:	So any particular companies that you’ve been advising or investing in recently that you find are super interesting you want to talk about?
Ben:	A few, I can’t talk about. One. A couple of years ago, I became advisor to a company called Augie. augxlabs.com and the founder was the head of product over at Warner Media and I’ve known him for a long time, so I was going to always want to help him. But he’s built one of the key generative storytelling platforms. You put in a prompt or maybe a couple of ideas and it can make an entire video based off of audio or a prompt or all sorts of things. They raised an oversubscribed seed round recently. He’s been really heads down executing. I think I see a lot of super interesting companies and I hate when I have to tell a whole bunch though because I’m trying to be, I’m very picky. I’m really laser focused on the set of AI companies I think can get to venture scale, but it’s really hard.
Healy:	Angel investments, that’s what you have to do for sure.
Ben:	Right.
Healy:	Yeah, totally. But Augie sounds pretty cool, and I buy the idea that AI can help you story-tell. I’ve got kids and they like me to tell them stories and I have to admit occasionally I’ve been using AI to do it for me, and it’s done a pretty good job. Although ChatGPT is getting a little redundant with some of the storylines, but it works. You can use it to produce stuff.
Ben:	This is where great entrepreneurs are making it so that it creates new kinds of outputs and it’s going to only get crazier with GPT 5s and new models and all sorts of things. And there’s new ways to process this data too. It’s going to get cheaper. I talked about in came out with 110-page deck on the state of AI investing. I’m expecting things like Mamba and liquid neural networks to make everything cheaper for entrepreneurs to build businesses on top of the AI stack because it’ll only get… It’s just the law of everything. It’ll only get cheaper as you can imagine, to build on top of these large language models. The Moore’s law just always applies to new technologies. It’s cheaper than ever. You don’t have to pay for thousands of servers like you used to, to start a business. You don’t have to pay tens of thousands of dollars to host stuff. It just keeps on going down. Same way.
Healy:	Yeah. So right now, our average client that’s spending money with OpenAI is spending $1,500 a month, which is actually a big number, but the median is much lower. It’s about $60. So, there are some API users who are pretty intensely using it. And then there’s a lot of folks that have 2, 3, 4 seat licenses that are just busting it out to make individuals more productive either in coding or marketing or whatever tasks that they’re in. So, it is amazing though how broad even one tool like that can go, that’s ridiculous. Are you seeing a lot of tools that can go that broad? It’s just amazing to me that ChatGPT can be used by your financial analyst who is doing stuff in Excel and somebody who’s coding and a marketing person. It is just how is one tool applicable to all of them? That’s amazing.
Ben:	This is an exception to the rule is the honest, it is very difficult to build a business like that and OpenAI is truly one of a kind. I don’t think it’s advisable to most founders to try to build a business that applies to everybody. Sam Altman’s been around the block, he was the head of YC, he’s done this a few times. I’ve known him for a long time, and he wasn’t necessarily trying to do that. He was just putting on an experiment that went viral and then everything happened. So sometimes this stuff happens by accident, but that one’s a rare exception to the rule and it’s a combination of timing and interface and technology and luck and I have seen almost no other kind of technology that can do that. Large language models are very unique in that front. I think we’ll see more like AI assistants and autonomous agents and things, but they’re leveraging a lot of the same stuff that has made ChatGPT so popular.
Healy:	Very cool. All right, so now I have a fun question here. We’ve got ChatGPT, there’s Claude, there’s Bard is now Gemini, there’s Perplexity, there’s even other ones. What’s your favorite one to use right now?
Ben:	I try them all on a semi-regular basis. I will probably do a newsletter benparr.com where I compare all of them again. I did it last year. Just to see how I feel. ChatGPT is still my default right now, but I’ve tried new things with Claude and Anthropic and a couple of and Perplexity and others and they do certain things that others don’t. So, I’m going to run a more methodical experiment and then I’ll be able to give you a specific answer because I feel like that’s what I have to do and I’m just trying to find a little bit of time to do it.
Healy:	Well, you better hurry up before they release something new because you’ll have to do it again.
Ben:	Well, you know what, then I will have incorporated that into my next one. I’m just going to be a regular thing I do every six months to a year just to test everything. At least once a year at the worst, but maybe every six months given everything moves so quickly.
Healy:	Awesome. So, let’s dive into that AI trends report that you published recently. And I’m going to share a screen here. These are your forecasting AI trends in 2024. And I’m not going to read through this, but let’s go through it here. What are you forecasting for 2024?
Ben:	So I’m going to talk quickly about 2023, which is there was definitely an increase in AI deals, although nowhere still near the size of 2021 just because 2021 was the greatest anomaly ever in terms of venture deals. It’s hard. For those who haven’t been through multiple cycles, it was such an anomaly in terms of what the broader landscape looked like. But it’s picking up in AI.
Healy:	That was crazy talk.
Ben:	Right and one out of four deals I think last year was AI. I think that number might even increase this year. And AI deals definitely were at higher valuations than their non-AI counterparts. 20%ish if you were seed and pre-seed, which is not so bad of a bump versus series B, which was like a 50 to 60% bump, which is really hard if you are a late-stage investor trying to get a deal, like a steal of a deal. But I’m expecting this year in 2024 for those who are listening in 2024, we’ll see how many of these are right. I don’t think there are anything crazy, but there’ll be more AI unicorns in 2024 than last year and there were a decent number. Deal making will increase slightly, especially as interest rates fall, which I expect to happen later this year. There’ll definitely be more AI exits. I think wat we’re going to start to see is people start AI companies, they build some interesting stuff, but they maybe aren’t having an easy time raising the next round. They sell to a bunch of other companies who want to have that AI talent. I think that trend will really kick off in 2025 when people start running out of money. But I think you’ll see more of that in 2024. There’ll be a lot more focus on verticalized solutions this year, which is I’m already seeing. AI for very specific verticals versus last year, which was more broader marketing. Those areas are usually taken up pretty early in the first round of building and then there’s other stuff like I don’t expect copyright to be an issue for this year. It’s going to take a long time to figure out the legality of certain things in AI, but the long-term is like the cat is out of the bag. There are models that are not… You could download them and put them on your computer, on your phone, you’re going to see more open-source models and it’s going to make it so that people could go build this stuff even if OpenAI got sued into oblivion, which I don’t expect they will be. And I think lastly, AI startups, one big trend I think I’m seeing, and I expect to continue, less money raised overall, especially in the later stages. Founders still need money in those early stages, seed, pre-seed, series A, I’m seeing less need to raise more money after that. You can get profitable after that, especially if you found some product market fit. People are hiring less. Now the brag is how few people do you have on your team, how many AIs are you using? And I think that means that early-stage investments will be really great deals for those investors who are investing at pre-seed. That’s why I’m doing it because I think you’re going to have huge returns on that front if you’re investing at the early stage because I think they’re going to be able to build big businesses. They don’t need a lot of capital. There’s a lot of upstream money for them and there’s lots of opportunity to build big businesses with less people.
Healy:	That’s awesome. That’s good because we don’t want to repeat 2021 where people just blew and wasted money. So that’s a good thing.
Ben:	I am happy to have a 2021 again for myself. That’d be great. But for the industry as a whole, it wasn’t healthy for sure. I think we were still in an unhealthy. VCs are holding onto their money in part because LPs are holding onto their money. The people who invest in VCs. And I think it’ll open up a hair as interest rates go down and why interest rates matter for those who don’t know, it’s because the people who give money to VCs can decide, okay, I need to make a certain return every year on my investment. And right now, putting money into bonds or other safe [inaudible 00:27:49] makes you much more money than it did two years ago because of the interest rates. It’ll make less money theoretically later this year and especially next year if the interest rates do go down, but they haven’t gone down yet. They won’t go down immediately, and they certainly won’t go back down to the 0% that we basically had at the end of the 2010s. Which that’s why we partly, we had such a bull run in VC, if you are only making one or 2% on your cash and VC offers, even with some risk, the opportunity to make 8, 10, 12, 20% or more on your cash, you would do that. But you don’t need to do that if you’re going to make 8% and you’re going to get 6% on your just keeping your cash in a savings account essentially. This stuff can sound boring, but it is actually really… I always tell founders one other thing to really think about the mindset of VCs when they are fundraising, because VCs are trying to raise their own money from other investors, and so the way you position yourself is going to matter to them because some of them need to get markups. Some of them are looking for specific types of deals. They need to sell you to their LPs or to their prospective LPs. Having that knowledge will help you a lot with understanding what’s going to happen in the market and why investors are saying the things they’re saying.
Healy:	For sure. In fact, not to pitch Kruze too much, but we do have a whole bunch of information on our website around how venture capitalists make money, how they think, how they evaluate investments. It behooves us to have our clients understand that stuff so they can be more successful in their fundraising. So, Ben, where can people learn more about you? Where can they find you online?
Ben:	So I publish a lot of stuff on benparr.com. B-E-N-P-A-R-R .com. That’s my newsletter, The AI Analyst. And I publish a lot of stuff on all my socials at Ben Parr on literally every social network. I do a comm with the information on AI. I post those things on my social regularly, so I do something once a month or two. I’m trying to get to even more. Octane AI is octaneai.com. But stay tuned. I got some cool stuff coming out. Announcing stuff. It’s going to be a crazy year for me, and I think for the AI industry as a whole. So, it’ll all be in benparr.com is what I will say.
Healy:	Amazing. Well, and we’ll have you back on here in a few months as well. It sounds like you have some neat stuff you’re cooking up that you can’t talk about now, but we will have you back on as well. Ben, thank you so much. It is always fun to catch up with you and let’s do it again.
Ben:	Let’s do it.

Ben Parr discusses the future of AI and its impact on startups

Explore podcasts from these experts