In today's episode, Michael and Ben alongside our guest Alex Levin dive deep into the evolving landscape of AI development and its broader implications on business and society. You'll hear Ben emphasize reducing the cost and time of AI development by leveraging open-source models, while Alex draws parallels between the AI industry and flat-screen TVs, advocating for AI as a public good.

The conversation traverses through the importance of compelling AI services, revenue-generating strategies, and the disruption AI brings—both in job creation and efficiency improvement. From personal anecdotes in semiconductor fabs to the pitfalls of the YC funding model, we explore various facets of success in the tech world. Alex brings a unique perspective from his background in psychology and entrepreneurship, touching on the importance of market timing, embracing uncertainty, and the significant role of mentorship.

Whether you're a startup enthusiast or a seasoned tech veteran, this episode will provide invaluable insights on navigating the complexities of AI development, operational challenges for founders, and the essential balance between innovation and business strategy. So tune in, and let's get started on this journey through the cutting edge of technology with our insightful guests on Top End Devs!

Socials

Michael Berk [00:00:05]:
Welcome back to another episode of Adventures in Machine Learning. I'm 1 of your hosts, hosts, Michael Burke, and I do data engineering and machine learning at Databricks. And I'm joined by my cohost,

Ben Wilson [00:00:14]:
Ben Wilson. I write release notes and release versions of MLflow at Databricks.

Michael Berk [00:00:20]:
Today, we are joined by Alex. He studied psychology at Harvard and since then has taken a variety of professional roles. For instance, he was SVP at Handy, an assembly and installation partner for large retail companies, and that was acquired by Angie, the company behind Angie's List. After that, he worked as a VC partner at BlueTrail Partners, specifically focusing on series a, and most recently, he founded Regal, a data driven outreach tool to help companies reach customers. So, Alex, I saw you studied under the legendary Steven Pinker. How was that experience, and what was your thesis on me?

Alex Levin [00:00:55]:
Hey. I look. I had a great time, at school. You know, at Harvard, typically, they like you to take different you know, they call them something separate majors, basically. Instead, I did sort of, unusual thing and basically, dictated a course of study that was across a number of different departments. So my course of study was largely around consciousness and the underpinnings of consciousness. And Pinker, who's a linguist by trade, has sort of migrated to that area of, you know, how do humans become conscious? How do you know if you're conscious? What is consciousness? Like, what does it mean? Like, you know, what is human nature? And so it was fun to be able to take classes in philosophy, neuroscience, biology, you know, whatever as long as it, like, stayed to that tact. So, yeah, I had a great time.

Alex Levin [00:01:38]:
You know, I definitely considered becoming an academic and realized I did not wanna stay by myself in a room for the next 100 years studying just to get to the edge of what was known about human mind and philosophy. You know, it's much more exciting to go into technology where, you know, a lot of us is so new. You can study for a month and be at the edge of what's known and then go be breaking new ground. And so for me, that was much more attractive.

Michael Berk [00:02:03]:
Got it. And have you been able to fulfill that goal of being cutting edge in the professional world?

Alex Levin [00:02:08]:
Yeah. Like I say, you know, you you I'm sure you could choose areas that are, harder to know, but a lot of this is also new. Like, you know, if I think about the area we're in within voice, when we started, we've always played around with the voice agents, the, you know, the AI voice agents. It was terrible 4 years ago. You know, in the last 6 months, like, we've made enough improvements. Not the LLMs, but actually how we use those have made enough improvements that, you know, we're gonna start rolling out voice agents to our customers now that are really indistinguishable from humans in a lot of use cases. And it's just, like, a great opportunity because what we as a company are very good at is staying, like, at the very edge of what's possible. Right? You know, when you're in a business, it's kinda like an arms race if you're in a consumer business.

Alex Levin [00:02:55]:
And that if you can use the latest marketing techniques, the latest customer service techniques, the latest, you know, in this in how you engage the customer, you have an advantage over your competition. And so, you know, for us as a vendor, it's great if we can stay really at the cutting edge of what's possible.

Michael Berk [00:03:09]:
Ben, question for you as a TL. So currently, I'm working with a customer that invented their own lang chain because lang chain is too unstable. And so I was wondering what your thoughts are are when to adopt a new product. You probably don't wanna be the first person. You definitely don't wanna be the last person. But is there a point in the distribution that's optimal?

Ben Wilson [00:03:30]:
It it depends on what the DNA of your culture is at your company, I think. And that was a question I was gonna ask you, Alex. Was

Alex Levin [00:03:39]:
Yeah. Yeah. I see what you're doing. You're spinning around and putting it on me now.

Ben Wilson [00:03:43]:
But I'm curious from your perspective because you're dealing with probably cutting edge startups as well as well established large enterprise companies. And I see it from from my side where we're building tools for, basically, r and d to use to build products.

Alex Levin [00:04:02]:
Yeah.

Ben Wilson [00:04:02]:
You're offering products for customers to say, hey. Just just use this thing. It works really well. We built it for you or customize it for you, and and it'll work and solve all these problems. But I'm wondering if you see the same thing that I see on the r and d side with certain companies just being completely unwilling or completely incapable, not not due to, like, a talent shortage, although sometimes that does happen, but more they have such crippling tech debt. Like, they've they've hired consultants over decades that have come in that are sort of bargain bidders, and they've built an abomination that their entire stack runs on. And you just look at that, and you're like, well, we can't integrate anything with this because it's so old, it's so fragile, or it's just so big, and there's too many pieces to integrate with. I'm wondering if you see that as well in your field to be like, hey.

Ben Wilson [00:05:07]:
We need to kinda do this off to the side. We can't integrate with your full tech stack because nobody can integrate with it.

Alex Levin [00:05:13]:
Yeah. Well, I'll give you a slightly different perspective, to start from. So, you know, let's just take a very big company, pick your industry. Like, I worked for a long time for a company called Thomson Reuters, which has a number of businesses, but 1 of their big businesses is in financial services. When there was very little change in the industry, when things weren't changing, they had a massive advantage as the incumbent because they could continue to keep customers using the things they were using. They already had those relationships. Customers weren't trying to get the new thing. They, as Thomson Reuters, didn't need to invest in making sure their technology r and d processes were the best because the things weren't changing.

Alex Levin [00:05:55]:
It didn't matter if their product improved. In periods where all of a sudden in financial services, there are massive changes, they had a structural disadvantage because they weren't set up to do r and d, weren't set up to do innovation, weren't set up to offer customers an a weird pricing deal for some new product. And, you know, startups have that advantage. So, you know, I always come back to that sort of famous expression, you know, who wins is dictated by whether, who what happens first, basically. Whether a incumbent gets innovation or, a startup gets distribution. So that's always, like, the battle. It's just 2 very different strategies to approach it. So if if as we were talking about before, I think about ourselves in the AR market.

Alex Levin [00:06:36]:
If what happens is the state the state of virtual agents today plateaus, you know, our advantage in that area is gonna very quickly go away. And, you know, even if it's slow, at some point, the big incumbents like 59, NICE, Genesys, and the contact center space will have that tool, and we'll lose that. Because even though we're the innovative startup, there's nothing new to innovate on. If on the other hand, you believe Altman and, the rate of improvement in AI is gonna continue to be exponential, well, guess who has an advantage? We do. Because we're really good at staying ahead of the curve, and we pride ourselves through pump and demand, creating an organization that's good at that and making it possible for our customers to have the the edge of what's possible, and the incumbents are gonna really struggle. Because even if we don't have the distribution today, everybody's gonna be clamoring for what we have because it's new in a way that's not just different, but new in a way that actually provides a better outcome.

Michael Berk [00:07:28]:
As a leader, how do you emotionally handle knowing that a large company could come in and swoop and just destroy your business?

Alex Levin [00:07:37]:
Well, I mean, again, having worked at a very big business, like, I know, like, I, today, could literally walk up to the CEO of Tom's and Bernie's, a company with, you know, 100, 000 employees and 1, 000, 000, 000 of dollars in revenue, tell him how I was gonna take his business down, draw a diagram, have a conversation for a month, and he couldn't do anything about it. It it is just not a fear that I have. And, like, I think, you know, people have this irrational fear that somebody's gonna go and destroy their business. It's not how it works. Sure. There are certain industries where there are network effects, and it's, you know, people have basically built a monopoly and it's it it is hard if an incoming comes in. In most businesses, that's not the case. Like, there are network effects such that it makes a monopoly, and anybody can go and compete.

Alex Levin [00:08:18]:
It's just a question of, you know, like I said, which strategy is better? Like, sometimes it is better to be the incumbent. Sometimes it is better to be the innovator, you know, and and having the perseverance to stick by the strategy if you think it's right.

Ben Wilson [00:08:32]:
Are you ever terrified as an aside to that? It's something I've always wondered. Why don't more of the established incumbent companies that have majority market share in the in certain industries adopt some of the sort of development philosophy of I'm not saying, like, become like a startup because that's not possible at an established company. It's there's no way to do that, but some big companies do establish sort of Skunk Works projects. Like, we were talking before we started recording. Intel did that. We created a group. It's like, hey. Go figure out this LLM stuff and make it really good, and then they split off and started their own company.

Alex Levin [00:09:17]:
Yeah.

Ben Wilson [00:09:18]:
Mosaic. But there are lots of big tech companies that do stuff like that. They have these r and d departments that are allowed to just play jazz and try to build products. Do you think in in specific industries that you're involved in now that there's a there's a there would be a fear if companies did something like that, like like, it exists in the high-tech market?

Alex Levin [00:09:41]:
Would would that cause me fear? Would that mean that I felt like they were more able to go and compete with us?

Ben Wilson [00:09:47]:
Would you have never even done what you've done and instead have done it at Thomson Reuters?

Alex Levin [00:09:53]:
Oh, I see. So I've lived through this, and I'm trying to think about how much I'm allowed to say, like, publicly. So I'll give you, like, a for instance, there was a big company that I worked at where, they so, you know, Thompson Reuters, they were going through a big transition in financial services, where, historically, they had provided all the data in the big financial services companies, but they were being crushed by Bloomberg who also provided a terminal. TomServe had bought in and, you know, was not able to sell that financial services business because they were getting so beat up by Bloomberg. And so instead of being able to sell the business, which is their preferred way of doing it, they got caught with their pants down during an innovation cycle that we were talking about. And they said, shoot. We're gonna have to build our own product, which is an unusual motion for that company. So they did 2 things.

Alex Levin [00:10:40]:
1st, they took the internal team that was built using the old product and tried to get them to build the new 1. And second, without telling that team, they hired a bunch of very startup y fancy people to build it from scratch using the same data sources without any of the infrastructure requirements. And it's fascinating to watch. The teams didn't know about each other, but I did because I worked for the head of strategy. And, like, at some point, like, this this sort of, like, strategy kind of, you know, both failed, basically. Like, the the incumbent failed because they weren't changing it enough. The new people failed because they were trying to do too much newfangled stuff, which, like, their big financial services clients wouldn't have ever agreed to do, and they collapsed the 2 teams into 1 to find kind of a happy medium. So I'm being a bit sort of broad in the strokes, but you get the idea.

Alex Levin [00:11:26]:
Like, I I don't know that that even if a big company could do that innovative thing, it's gonna work for their customer base because, you know, their customers expect certain things. So a very simple example is a release cycle. You know, you have this innovative team at Thompson Reuters that wants to release every day, but their financial services clients expect the once a year release with 6 months notice of a sort of a cut in advance that they can start playing with. Fine. That it's you know, if that's what the customers need, it's okay. But you're not gonna be able to do the, you know, fast innovative thing if you're expected to have 1 instance that's released a year. So, you know, it's just structurally difficult for them. Even even in our scale, you know, another way of looking at it, we think about, hey.

Alex Levin [00:12:08]:
If we're gonna do new projects, we want it within the core infrastructure we built, where there are even in our scale rules of how you deploy, how you do QA, and how you do certain things versus do we just want it, like, off to the side where somebody can go very quickly build something that works, you know, at a smaller scale? And so even in our scale, you know, we want that to happen off to the side, and it's, you know, tricky to figure out how it integrates. And a big company, yeah, the it's it's tricky to figure out how it integrates. At a big company, yeah, the it's not easy to do. So no is the answer in the end. Like, I still don't have fear of big companies. Broadly, like, the thing that makes people successful at start ups is finding a, you know, AAA market that you where you really understand that the market is growing, even if it's at a small place. And, ideally, it's something that no 1 else sees, so they sometimes call this an earned secret, like, you know something that other people don't know. And when you come in, like, it turns out you're right.

Alex Levin [00:13:01]:
So, like, that special quadrant of, you know, it's something that's that other people don't see but is growing, And, like, you're right that, like, how you're gonna serve it is where you really, you know, generate these massive businesses. So take, like, Airbnb as an example. Everyone thought it was the stupidest idea if I'm allowed to say that, like, at the beginning. Why would anybody stay in another home? That's the worst business. You should never do that. It turns out, though, that customers wanted that, and there was an explosion in the amount of people that are willing to rent their houses. And so it's the perfect storm for building a very big business without much competition.

Michael Berk [00:13:33]:
How much of start up success is market timing?

Alex Levin [00:13:37]:
Less than you think, honestly. Like, definitely, like, there's there there is such a thing as being too early. So, you know, you think about, like, Webvan or something where people were trying to do delivery before some of the, preconditions were met. And so, like, for sure, there's things about being too early. But, like, as you know, I talk to you sometimes where we're like, should I start now? Should I start in 6 like, that doesn't matter. Like, if you believe that you're right over the next couple years, like, just get started. Because there are gonna be a lot of things that you can figure out and solve that, you know, will move you closer to your goal even if the market isn't quite ready. So, yes, if you're 20 years off, fine.

Alex Levin [00:14:16]:
Like, the underlying technologies are not available to do a voice agent and you're trying to do it a voice agent in the 19 eighties, it's probably gonna be bad. But, you know, as long as you're within a couple years, don't worry about it. Just start.

Michael Berk [00:14:26]:
Got it. So, yeah, I was referring more to larger market trends. And to me, it seems like timing these large trends is the key. Like, if you're in the Gen AI space, you could have a piece of shit company, and it would probably be getting some funding.

Alex Levin [00:14:40]:
Yeah. And that's not success. Let's be really clear. Funding is not success. And, like, we need to end the YC version of the world where the You know, being able to get some money from y c, some money from, you know, You know, being able to get some money from YC, some money from a, you know, seed investor and then sell your company for $20, 000, 000 and have it be an acquire, not success.

Michael Berk [00:15:08]:
What is the minimum benchmark for success in your opinion?

Alex Levin [00:15:11]:
Well, look. I'm not gonna define another person's success, like, from from, like, what somebody else wants from me.

Michael Berk [00:15:17]:
Yourself. Yeah.

Alex Levin [00:15:18]:
You know, my co founder and I are ambitious. Like, you know, we really have fun working with our customers, helping them. We ultimately work for b to c brands, helping them figure out how to better engage with their end customer. Right? Using data to drive, you know, when they should reach out with what customer, with what message, on what channel. And so, like, we like it. We enjoy it. I think, you know, what we measure is, in our case, is how much revenue our customers are driving using our platform. So at this point, it's more than $3, 000, 000, 000 have been driven in revenue by our customers using our platform.

Alex Levin [00:15:49]:
And that's exciting because it means that, actually, like, it's having an impact. It's having a difference in how triple a and Angie and SoFi and Roman are engaging with their users. What's, you know, what's the target for that number? I don't have a target per se, but, you know, that is the north star. We do watch, like, how much we're able to have an impact on their business.

Michael Berk [00:16:10]:
Interesting.

Ben Wilson [00:16:11]:
But I

Alex Levin [00:16:11]:
don't look at number of employees. That's my pet peeve, actually, more than VC funding. Is when 1 founder goes to another, goes, how many employees do you have? As if that's like the measuring contest. Like, I'd I'm much more impressed if you have a company doing a 100, 000, 000 in revenue with 20 employees than I am if you have a company with a 1000 employees. So employees is not the measuring stick. In fact, I'm you know, personally, we try as much as possible to have fewer employees doing more valuable work and have more vendors. There's a big advantage to outsourcing things that you're are not gonna be your core structural advantage.

Michael Berk [00:16:44]:
Like what?

Alex Levin [00:16:45]:
So, you know, take machine learning. Like, we're not gonna go and build the best training models. So, So, like, let's not let's not try. Let's go find whoever's gonna be best at that and work with them and push them. Right? Give them feedback. Right? Tell them what features matter to us. Drive a lot of revenue to them so that we're an important part of their road map. And, you know, then we can twist their arm when we need something.

Alex Levin [00:17:05]:
That's much better than having employees who are, you know, doing that. Or the classic example in custom in, b to c businesses is customer service. You know, if you're not gonna have a path to promotion and to internal success for customer service agents, outsource it. Put it somewhere where someone else has managed that organization, where that person can have a fulfilling career within a separate organization, and you're not treating them like second class citizens within yours. Make sure the people that are your employees you're investing in and you're not treating like a a different group of employees, and you're providing them a pathway to be promoted in the world.

Michael Berk [00:17:40]:
Okay. Follow-up to that. Just rapid firing. Ben was alluding to earlier that a lot of consultant built solutions are overblown, overengineered, especially if you're billing by the hour. There's a lot of inverse incentives. How do you know that your results from these external firms are good?

Alex Levin [00:17:59]:
I mean, they're I'll give you, like, a a small example, then we can have I can give you a long answer. When we used to outsource customer service at the last company I was at in the home services space, we always have a challenger model. So we'd have 2 outside organizations, like, typically 75% with 1, 25% with other, and then we always kept a few agents in house. Not allowed us to just keep them honest and know, like, hey, what it would look like when we had in house agents versus the sort of, like, the 2 challengers. So there are ways of setting it up where you can keep people honest. The the that's a very tactical example. The the longer example is, you know, the the beautiful part about technology companies is when you invest deep, you know, research or building in a very hard problem. And then at scale, right, the the margin for distributing that thing you've built are very high, and there are there's less level of investment you need to make.

Alex Levin [00:18:53]:
So we can argue after about whether AI is even a good sort of software model because it's unclear that the long term margins are good, and there's a lot of upfront investment. But leave that aside. My point would be is that you're better off picking 1 hard problem and and really, really spending your money and your time getting very good at that and having big margin than trying to solve lots of easier problems and spreading out your sort of, thinking power on many different things you're not gonna be as good at.

Michael Berk [00:19:22]:
Yeah. That that reminds me of a lesson I learned super early, which was the the if you look at the distribution of returns in whatever you're doing, typically, that long tail is where everybody dies, and there's a couple things that are really, really successful and powerful. And so, getting a bunch of signal on a wide range of topics early is helpful, but then you it's really advantageous to sink a lot of energy and effort into a few things or maybe even 1 thing. Yeah. So what

Alex Levin [00:19:49]:
You know, I'll pick on AI to the conversation I was starting. Like, it's it's not clear that the long term margin on AI is gonna be good. You know, yes, it's clear that there's gonna be a lot of investment dollars upfront. But, you know, if I'm selling AI and it's either commodity, in which case, like, we all have to bring down pricing, or I keep having to do lots of things that look effectively like a like a like service or labor to make it work. Meaning, I need lots of training, which is, like, heavy cost or lots of variable cost to, you know, do the compute, the margins aren't gonna be that good. And so the return profile isn't gonna be as good. I like, I believe the business is built on top of the AI infrastructures are gonna, you know, use that AI to drive value for their businesses. But are AI businesses themselves gonna be very profitable and good investments? I think TBD.

Michael Berk [00:20:40]:
Ben, what's your opinion?

Ben Wilson [00:20:44]:
I think it entirely depends on how disruptive the free offerings become. And, you know, that's something that we're working on at Databricks is like, DBRx, that announcement a couple months ago. The things that are being worked on now, we're not in the business of, oh, well, let's let's retrain the next big 1, and it'll be more powerful. We are doing that, but the reason for doing all of this stuff for offering these these open source models that are, you know, generally well trained on a bunch of topics is to do the research of figuring out how can we make it so that we can take 1 of these foundation models and come up with a way to retrain it where it's exceptionally good, like, better than humans at this 1 task or this 1 discreet area. And it's garbage at everything else, but you put guardrails on it. Say, hey. You you can only answer questions about this, or only interface about this, or only generate images that are in this this style, or only generate audio in this style, whatever it may be. And that research train that's going on is yielding shocking results where it's all about how do you get the infrastructure cost time down in both that that variable cost, like, GPUs are expensive, like, and also the time.

Ben Wilson [00:22:06]:
And that's way more expensive when you really think about it. How do you shorten that so you can get something that's usable that you can build into your product and integrate? And not just give that into the hands of end users who are, you know, b to c, you know, because you need tech investment. You need to hire staff that can do that. But if if there's models and infrastructure around that that's open sourced, you know, companies like yours, Alex, can can take all of that that tooling and be like, we can iterate a 1000 times faster and cheaper because of this. And I think that's the goal right now with a a number of companies is trying to get that to be effectively seen as a commodity. I think it should be a commodity and should be cheap, but that's to like,

Alex Levin [00:23:04]:
it's wholeheartedly. Like, the metaphor I often use with people these days is, flat screen TVs. You might say, like, the AI industry is gonna be a lot like flat screen TVs. So what do I mean by that? Basically, you know, at the beginning, if you wanted a big flat screen TV, it was, like, bulky and heavy and hard to get in the house, and it cost a fortune, and only the, like, billionaires would buy the big flat screen TVs. But to you know, what's happened is that the cost of all the parts in the flat screen TV has just plummeted, and, you know, the margins on that business has plummeted. And now everybody has access to big screen TV. Turns out, it's not it's not good to be in the flat screen TV business. Like, all those businesses have really done very badly.

Alex Levin [00:23:43]:
However, to your point, it's created the video game industry, and it's created the movie industry, and it's created all these people that I'm able to now use that TV to do incredible things that are very profitable businesses. So, you know, I do believe some of this AI should be a public good and should be, you know, available more cheaply, and it will require a lot of investment. I guess my concern is that I do invest a little bit is, hey. Would I put a lot of money into into the into the flat screen TV business knowing what I know now? Probably not. Like, yes. It needs to happen. But if I'm, you know, if I'm doing it for public good, fine. If I'm doing it for a return, probably not a good idea.

Ben Wilson [00:24:23]:
So the key for AI businesses who are doing this research is to provide a compelling service that people can build on top of. That that's, like, our goal for on what we're doing

Alex Levin [00:24:35]:
is Or go and be the next, you know, platform for building video games or be the next video game. You know? Those are all gonna be fantastic businesses because they're not they don't have the enormous cost of research and r and d.

Michael Berk [00:24:49]:
What's your take on chips and the ROI there?

Alex Levin [00:24:52]:
As a as a business model, like, should you be should people be building businesses for new chips? Honestly, I know very little about the hardware business, so I'm not like, my opinion is pretty much worthless on it. But, structurally, like, my belief is the same as the comments I was making on TV, which is we have demonstrated an ability to use innovation to drive down the cost and up the processing power of, computers. Now as I understand it, there are physical limitations to silicon chips, but we're now researching other kinds of chips. And there could be quantum computing 1 day and other things. So all the limits that we hit that provide, you know, create real cost, I believe there's a good chance that we push through those and continue to, you know, create more compute power at less cost, basically. So, you know, if if I'm a betting man, I don't bet that, like, you know, it doesn't change. I bet that it continues to get cheaper, basically.

Michael Berk [00:25:44]:
And then in the AI space, what do you think are the high ROI fields or just things? What are the or even, like, underlying market trends or tenants?

Alex Levin [00:25:56]:
Yeah. Like, you know, so a couple different ways of looking at it. 1 is, III know everybody's been quoting it, but Altman's comment about, hey. Like, you know, if you're building a business on top of OpenAI, like, you're probably pretty silly to bet that it's OpenAI is not gonna keep getting better. Like, do something that bets that OpenAI is gonna keep getting better and, you know, use that to your advantage. So, yeah, I think all these people that have built, like, little basic things, like, what do they call it these days, like, prop stuffing or whatever on top of OpenAI, like, that's probably not gonna be a great business. But if your business, as as their loan gets better, your business gets better, like, then that probably would be better. So I think be careful about what bet you're taking, you know, as this technology changes.

Alex Levin [00:26:38]:
And then the other comment I make is back to what I was saying before. Look for areas where you have an earned secret. Right? Look for areas where you believe something to be true that no 1 else does. So today, the most common use, I think, or 1 of the most common uses of, generative AI is in inbound text customer service. Everybody knows that inbound text customer service is an area ripe for, disintermediation, if we wanna use that word, for, you know, generative AI to replace a human being. And so there's a 1000000 people working on it. And, you know, if today, you can charge 500 or let's just say $5, 000 a month for that generative AI agent, Well, in 10 years, it's gonna be 500 a month. Right? The cost is gonna fall astronomically fast.

Alex Levin [00:27:18]:
Instead, go find an area where maybe it's a smaller market today, but fewer people know about it and there's more growth opportunity without as much competition, you'll be able to build a better business. So look a little further than the obvious things, I would say.

Michael Berk [00:27:32]:
Okay.

Alex Levin [00:27:33]:
So I know I'm not making specific recommendations. Like, these are more sort of general themes that I would look for if I was starting a business on top of, AI today.

Michael Berk [00:27:42]:
And what's your take on the AI will disrupt and take everyone's job and then we won't have jobs opinion.

Alex Levin [00:27:49]:
I if you look at the history of technology innovation, you know, the washing machine, the Internet, cell phones, whatever, email. Like, it's it's fascinating what happened. You know, even automation factories is a great example. Like, it's fascinating what happens each time where it destroys a huge number of jobs, but then it creates a huge number of new jobs. And, overall, like, I'm much happier with the way in which we work today than how my father worked. Right? My father came in at 21 and started working, you know, and it was AAA professional job, and he needed to know something. He, like, took the afternoon off and went to the library and started reading books. Right? If he needed to, like, call people, he had to go figure out their number somewhere and, like, you know, you know, rotary phone start, like, you know, and often have to physically get on a plane to go and see them in China.

Alex Levin [00:28:36]:
Like, just the amount of time it took to do things was very, very significant. I can do that same thing in a morning by using the Internet and by getting on Zoom and talking to the person. So massive efficiencies, but it's not like we've destroyed jobs. If anything, we've created new opportunities for jobs. So, structurally, I am a believer in the deflationary force of AI and that using AI will bring down the cost of products. Those will be fewer humans involved in the chain. But, 1, that means you don't need to make as much money, and I believe also it'll create a lot of new jobs that allow people to find new careers.

Ben Wilson [00:29:09]:
Yeah. There's a an interesting anecdotal story that I tell some people who know that I used to work in a semiconductor fab. So I I worked at the biggest 1 in in, North America for 3 and a half years or something. And the job that I was doing was analyzing yield performance. Like, hey. We have defects on the chips. We're making the chips that go into cell phones, back at the time. And they're very complex, lots of things on them, and just 1 little tiny speck of dust or something that's embedded in there in that wafer, on that chip.

Ben Wilson [00:29:46]:
It's trashed, and my team is tasked with figuring out trends and analyzing, like, what is this failure mode? What does it actually do when it's in this area of the chip? And, you know, how can we see that on test or detect it earlier? And I remember talking to 1 of the metrology engineers, where we had this this 1 incident, and they're using scanning electron microscopes to determine morphology of these defects. And I was standing over the guy's shoulder, and I was like, hey, Dan. Like, can you walk me through the process that you use for, you know, detecting this thing and and telling, like, what it's made out of and and, you know, where could it have possibly come? And he remember he just turned to me. He's like, man, I hate this part of my job. Like, I hate it. So it gets so monotonous. He's like, oh, he's like, if you have 4 hours, you know, sit down next to me. I'll show you what I do.

Ben Wilson [00:30:42]:
And at the end of it, I was almost falling asleep. I was so bored, with how many screens and how many, you know, different manual tasks that he had to go through to to look at this thing and and then submit it for additional testing and then walk over the lab and do all this stuff. And that was in the 1st 3 months of me working there, so I was trying to learn how all this stuff worked. By the time I left, there was no human involved in that. It was built into the scanning machines. They could determine all of that stuff. And it was using machine learning embedded in in the chips in this thing, not what everybody now is thinking about AI and stuff, but, you know, true, you know, traditional, machine learning. And I remember going up to him before I quit that place.

Ben Wilson [00:31:26]:
I was like,

Michael Berk [00:31:27]:
what do you do with your time

Ben Wilson [00:31:28]:
now that you're not wasting it like that? He's like, I do my actual job, you know, the interesting stuff. Is, like, you you know that incident that happened 2 weeks ago where we lost $1, 000, 000, 000 worth of product? I'm like, no. What are you talking about? He's like, exactly. He's like, I'm able to focus on what I should be focusing on instead of doing something that a machine can do.

Alex Levin [00:31:47]:
Yeah.

Ben Wilson [00:31:47]:
I was like, yeah. That's what AI is gonna do. Got it.

Michael Berk [00:31:52]:
It's interesting to think about how that could also accelerate the the growth of humanity too. Just the do you guys see the 3 body problem or read the books?

Alex Levin [00:32:01]:
No. I haven't done it yet. Is it good?

Michael Berk [00:32:04]:
I didn't read the books. I I, like, started to zone out midway through the first chapter, but, apparently, the best things ever, if you have the attention span. But I thought the show was solid. And 1 of the core tenants of the show is basically humans are gonna evolve at sort of an exponential rate. And, I think that that's speaking to this. If you can start automating the boring and let people focus on the fundamentally human, fundamentally creative aspects of the job, it's really exciting to see this exponential curve. Like, in the next 50 years, we could be completely in a different space relative to the next 2 or the prior 200 years. So Yeah.

Michael Berk [00:32:42]:
We'll see.

Alex Levin [00:32:43]:
If we're really being right, at the birth rate today, max population den will be in, I think, it's a 100 years to 300 years depending on who you talk to. So after that, the population is gonna crash, not slowly, but, like, astronomically. And so if we haven't built robots and AI and do and all this stuff that will enable us to have a lower population number, like, that population is gonna be in trouble. Or we have to somehow magically increase, you know, birth rates from whatever the 1.8 it is now back up above 2.1.

Michael Berk [00:33:14]:
Yeah. Yeah. And, also, like, what is the carrying capacity of the Earth that's relevant too, but not quite machine learning topics. Yeah. So, Alex, I wanted to shift gears a little bit back to, your work experience. It seems like you've done a, like, very broad array of jobs ranging from sales to marketing to growth to product. How has that set you up to be a good leader or a bad leader?

Alex Levin [00:33:42]:
Yeah. We definitely I first of all, I believe in a liberal arts undergrad education. Whatever it is you study, I think as long as you really are interested in it and deeply engaged with that topic, like, it's a great way to spend 4 plus years. So, I encourage people to do that. I do think, like, it was good that I understood, like, every company would be a technology company and made that shift into being a product manager. By no me, I'm not an engineer. Like, I and I don't pretend to be 1, but, like, I'm close enough to be dangerous. Like, I I, you know, I understand in the areas that we work in some of what's going on.

Alex Levin [00:34:15]:
And, you know, if people today still tell me they wanna be a CEO, I say, well, like, how good are you in technology? Like, if you don't understand the underlying technology, you think they're gonna make you the CEO of the company? Like, that doesn't happen anymore. That those days are gone. So, you know, I do think, like, everybody should go through some level of training. And when I interview people today, 1 of the questions is, like, have you ever built a website? Have you ever used SQL? Not because I want them to be good at it, but if they've never done that, how are they gonna possibly understand the things that are going on in the company? I think the other sort of experience that was very valuable to me was having seen sort of a big company thought management was and then going to a smaller company where it was a very different perspective, where it wasn't, hey. The more people that work for you, the more important you are. It was more actually, like, the more redundant you can make yourself as a manager, the better you are, which is a very big shift. And it it it's important when we bring people in or managers to teach about this and say, look. If you go and make yourself so important in a specific team that, you know, everything is going through, is that good? And they go, oh, yeah.

Alex Levin [00:35:18]:
Yeah. Because I'm very important in the team. And I go, no. That's not good. In fact, not only won't I promote you, I can't promote you. And they go, what do you mean? I go, I physically can't promote you because you are so integral to what that team is doing. And they go, oh, shit. You're right.

Alex Levin [00:35:30]:
Like, on the other hand, had you, you know, spent your time digging into it, and I believe in doing the work and actually being an IC as part of being a manager, But then ask yourself, how do I make it so that I'm redundant? Whether that's through technology, processes, hiring, whatever, and then come back to me and said, Alex, you thought this just this management job cost you $200 a year. I've turned into something that I do in 10% of my time, which basically means it's only cost you $20 a year. So either hire a lower level person to do my job, or let me do that with 10% of my time and give me a different job. Now I'm, like, praising the ground you work on because I've shown not only can you manage that team, but you have the capability of solving more complicated problems, and I can give you a bigger, meatier thing. And believe me, there always are bigger, meatier things. So the task, I think, of learning what it is to be a manager or in my experience with learning what it is to be a manager, like, that was a big mindset shift. Not saying I do it well all the time, but, like, it was an important mindset shift. And then I think, having been at start ups, seeing, you know, what I liked about what founders did, what I didn't like was good.

Alex Levin [00:36:34]:
Like, I think it's really hard when people go and be a founder the first time having not worked at start ups because they wanna have a reference point. And I'm not saying you have to work at the best startup. Just go work at a startup because you're gonna learn, like, how do they manage people, how do they do decisions, how many founders are there, x, y, and z, and see what you agree with and what you don't. I think without that, like, it takes an extra year or 2 just to get going. But because my cofounder and I had seen all that, we hit a $1, 000, 000 of revenue within 6 months of starting the company, 3, 000, 000 within 12 months of starting the company. Like, we didn't have to spend a year or 2 working on all that stuff because we had already had a good idea of how it worked. So, you know, am I perfect as a CEO? By no means. And there's a lot of areas for me to improve in, but, definitely, like, having those experiences helped me a lot.

Michael Berk [00:37:22]:
What are the things you avoided making mistakes on by knowing how start ups work?

Alex Levin [00:37:27]:
The biggest 1 is the the nuts and bolts of operating a start up. Just, you know, things like that, to us, feel simple, but I think, you find out are really complicated to founders who've never done it. Like, how do you set up an office? How do you hire people? How do you fire people? How do you manage people? How do you create goals for the company? How do you do basic financials? You know, how do you run a board meeting? Like, our the founders we worked for before tell this hysterical story when they were first time founders, they raised some money from General Catalyst, and they, you know, get in a bus and go to Boston for their 1st board meeting, and they sit down. And, first of all, they go say hi to this guy who's actually their lawyer, not the, like, outside lawyer. They didn't even know. That's embarrassing. And then they go, okay, guys. What are we talking about? And the investors are, like, well, so did you bring a deck? They didn't even know they had to create content for this meeting.

Alex Levin [00:38:18]:
And, like, you know, they're they're going, well, we're gonna go print out a bunch of stuff. So they leave the room, print out a bunch of stuff, and they, like, try to, like, somehow do this meeting. And after the meeting, they're walking back to the bus to go back to New York City from Boston. They get a call from their lead investor, says, hey. You know, I know we just gave you $3, 000, 000. You know you can just return Imagine that call. Like, when your first board meeting goes so badly, the investor says you can give them the money back. It's okay.

Alex Levin [00:38:42]:
And they go, no. No. No. No. Don't worry. Like, we understand now the expectation. But, you know, it's not their fault. If, you know, they had never seen what that meeting was supposed to be like, and so that caused them extra, you know, consternation that wouldn't have been needed.

Michael Berk [00:38:58]:
Interesting. So it's on the logistics side?

Alex Levin [00:39:00]:
A lot of it. Yeah. You can't imagine how much time early like, first time founders waste on this stuff. We're just it's not, again, not their fault. They just haven't done it before. Like, I've you know, between my girlfriend and I, we probably hired 500, a 1000 people. Not that we're perfect at it, but, like, we've done it enough times that we can create a process and know what to do and know how to make sure it sort of at least works.

Michael Berk [00:39:24]:
So then where do you spend your creative energy?

Alex Levin [00:39:27]:
Or if any of hard problems in a business. Right? You know, you wanna be focused on those. Like, the the framework I sometimes say to people is, like, if you're leaving the office at the end of the day satisfied and happy, you're not doing your job. And they go, again, that's, like, strange way of looking at it. But that means you're solving easy problems. That means that you're solving something you know how to do. And, look, there are some jobs in certain companies where that's the task. It's just do the job you know how to do.

Alex Levin [00:39:51]:
Fine. But in a company like ours, that's not the task. If every day we're coming in and solving problems we know how to solve, we're not moving forward and people are thus catching up with us. On the other hand, if you're leaving the end of the day frustrated, you're going, this problem, I can't figure out this answer or how do I do this thing? Frustrated about the right things. Like, I don't want you to be frustrated by your coworker. But if you're frustrated about the problem, that's good because you're trying to solve something that no 1 else knows how to do. You're moving it forward. You're figuring it out.

Alex Levin [00:40:17]:
And even if you're failing at first, that's getting you closer to the answer. So I'd much rather you be doing that hard work every day knowing that you're solving something no 1 else has ever done before and moving the company forward.

Michael Berk [00:40:31]:
That makes sense. I'm I'm frustrated 99% of my days.

Alex Levin [00:40:35]:
Yeah. But, again, don't don't let the the the thing to remember is don't go home and go, it's bad to be frustrated. Go home and go, what what how lucky am I as a human being that I get to live in this job where I get to do things no 1 else has ever done before. Like, that's so much better than, oh, yeah. I came home, and I know exactly what my day tomorrow is gonna in my opinion. Like, it depends what you want. Like, the other people want a different kind of life. But for me, like, that's exciting.

Alex Levin [00:40:59]:
Every rock I turn over every day is a new thing that I don't know the answer to.

Ben Wilson [00:41:05]:
Yeah. I think that's excellent advice for anybody at any level of management, be it from a group leader on a technical team all the way to any big company, CEO, CT, you know, any e staff member. And I I've I've noticed parallels between what you just explained and all of the great leaders that I've ever interacted with is they're all thinking about that, but they're putting they're not just thinking about that and then keeping it to themselves. They're imparting that to the people that need to hear it. And I remember early on in my career, getting into tech that I was almost had that feeling of being afraid because in the like, the world that I worked in before, you had to know what you were doing and come up with a formulated plan of, like, okay. I'm gonna spend my time and plan this out and then do a physical experiment. And if I can't just test this thing without a solid hypothesis because it costs 1, 000, 000 of dollars, I'm not I'm not gonna do that. Yeah.

Ben Wilson [00:42:15]:
And then moving into tech, it's like, no. Just try to build this thing. We don't know if it's possible, but you'll figure it out. And then getting that panic of, like, I have no idea what I'm doing. And I it was almost anxiety inducing of, like, moving into that. A really good VP sat down with me and explained almost exactly what you just said. It's like you should feel that, and you should enjoy that feeling because it means you're going into the the unknown and you're doing something complicated. And the reason we told you to do it is because we think you can do it, So go do it, and don't worry about if you mess up.

Ben Wilson [00:42:48]:
I was like, you know what? That that's good advice. He's like, well, thanks. Now go do it. Don't worry if you mess it up.

Alex Levin [00:42:57]:
Yeah. It is it is you know, we're we're lucky. We're so lucky to work in a job where there's new things to be done, and we have the opportunity to experiment and fail. Like, you know, not everybody has that opportunity.

Michael Berk [00:43:10]:
Is it sounds like it's a lot of nurture and not nature, though. Everybody has to learn to like this?

Alex Levin [00:43:16]:
No. I mean, you know, my cofounder and I or I'll speak for myself. Like, I was I didn't even know how miserable I was by the time the last company we took from a $1, 000, 000 in revenue to a 1, 000, 000, 000 and a half. You know you know, it was a public company at that point. I wasn't happy managing managers and managers where we have all this this it wasn't what I wanted to do. It makes me, me, personally, very happy to be in this newer environment where we're learning new things. Not everybody is the same. Like, there are plenty of people that are very happy in that other environment.

Alex Levin [00:43:44]:
So I'm not trying to put my values on other people. I'm suggesting that if what you want is that, faster moving environment, don't look at that frustration or that fear as a sign that you're not good enough for it or you shouldn't be doing it or that something is going wrong. In fact, to your point, that fear is a sign you're doing it right, and you're in the right place. So it's a weird thing to learn that, like, that that's actually AAA positive feeling, not a negative feeling.

Ben Wilson [00:44:14]:
And, eventually, you get addicted to it, and that's all you want. You know? And then you get hired at, you know, fast moving start ups. They're like, yo, everybody around you is the same way, and you're like, oh, this is kinda great.

Alex Levin [00:44:27]:
Yeah. So I think, you know, when we think about the organization, like, 1 of the pieces that we're proudest of is we, Michael Federico, made a decision about what values we thought would make this company successful at this stage. Now that doesn't mean it's for every company or every stage, but thought about those and wanted to make sure people made decisions in that way. So when I think about the values of the company, it's not ping pong and parties. It's how do we make decisions? Because, you know, if we can make sure that people coming in agree with the way we make decisions and like that, this is all gonna go swimmingly. You know, when people come in, even if they're great, and they don't agree with the way we make decisions, we should be very honest about that and say, maybe this is not the right cultural fit. You know, just as an example of Google, you know, I have a lot of friends there who loved it, but I find that it's very consensus driven, very slow in decision making, not very database in in some teams. Like, there are certainly database teams, but many teams, it's about things other than data would frustrate me.

Alex Levin [00:45:23]:
And I don't think I could work in an organization that was consensus driven and not database.

Michael Berk [00:45:28]:
So how do you make decisions?

Alex Levin [00:45:30]:
Within Regal, you know, we focus a lot on, you know, first, what's right for the customer. 2nd, like, how do you do things quickly? Meaning, like, you know, be 80% right rather than, you know, be a 100% right. How do you use data to make your life easier? And the as we're sort of on a more engineering sort of focused podcast, let's say, you know, you have 2 engineers who are trying to decide what database to use for something new, and really what you're gonna do is run some SQL query in this new example. It's easy. In many companies, they'd argue over which database is better. My suggestion, don't argue. Just say, hey. What we're trying to do is run the SQL query faster.

Alex Levin [00:46:05]:
Run-in 1, run-in the other, see which is faster, and you made your decision. So by turning it from an opinion based debate to a database debate, you've solved it because you've moved faster and not had this, like, tete tete. Right? You know, I'm not saying that opinions are bad. People think usually when I say you have to use data that opinions are bad. No. What I'm saying is use opinions as a starting point, not an ending point. So if there is no data and you say, hey. At the last company, x, y, and z work, great.

Alex Levin [00:46:32]:
You've just like, that's a shortcut to get started. But the second you started, then start creating AAA way to use data to see if you were right or to alter what you're doing till you find the right thing for your company. So, you know, we're not saying always data, but, you know, often, use data to make your life easier. And then, you know, at our stage, for instance, we talk about growth being sort of the solution to all problems. You know, more growth for our customers, more revenue for us, more opportunity for employees. So that's kind of the tiger. That doesn't happen forever. But at this stage, like, you do wanna make the decision at least to more growth rather than something else.

Alex Levin [00:47:07]:
And, you know, there's other values, but those are just some of the ones we think about making sure people understand as they're going and making decisions about what to do.

Michael Berk [00:47:16]:
Chris, look there. Well put. 1 more question. So Ben and I have been chatting a lot about your prior topic of the importance of making yourself redundant. And typically, that can be achieved through mentorship and sort of teaching the people around you how to do your job. How do you think about mentoring? Do you have sort of a a to do list of these are the 7 things that this person needs to be able to achieve? What's what's your approach?

Alex Levin [00:47:48]:
Yeah. Honestly, I'd say it's something that I'm not as good at, especially in this environment. I think in more stable environments, it's something that I'm better at where you have more time to focus on that. You know, at our stage as a company, you know, we pick what we do is for every new employee, 1 of the things we do is at 90 days, we say, here's the value. You're living the best, and here's 1 thing, basically, I want you to work on. So, like, you know, good and bad. But, you know, not bad. Just the thing that we think would have a lot of impact in in in what you're doing.

Alex Levin [00:48:18]:
And so within that, then we try to, support them, and it's not always mentorship, right, by me. It might be, hey. Here's a class. It might be, here's a book. It might be, here's somebody I want you to meet with once a month that's from outside the company. It might be me just holding them, you know, keeping them honest, saying, hey. We said we were gonna you agree. We said we're gonna work on your ability to negotiate internally for assets or whatever.

Alex Levin [00:48:39]:
And, like, here are 3 examples in the last week. How did they go? So I don't know that that's me being a good mentor as much as it is me just sort of being the, mirror, you know, coach to some extent, which is a slightly different, thing. But, yeah, as I think the company gets bigger, my role and and sort of the teams go more stable, my role shifts and many managers will shift into being more of a mentor when you then have more of a middle management layer. We don't really have that middle management layer as such. Like, there are middle managers, but it's not a big enough company that there there's enough time spent on mentoring, I'd say, honestly. Got it.

Michael Berk [00:49:17]:
So it's a bit it's a function of organization size. And as you get bigger and more stable

Alex Levin [00:49:22]:
On my opinion. You've talked to other people, and they'll say, you know, there are other CEOs I know where they don't do any individual contributor work. And if they want something done, they mentor the person. I we believe a little differently. Like, we believe all managers should be spending 20, 30% of their time doing IC work, and part of getting, you know, to the right answer is being in the details. That doesn't mean I need to be the decision maker on it. That's a different role. But being involved as an IC understanding and digging into it is a very different mode than mentorship.

Michael Berk [00:49:50]:
Ben, how does Databricks operate on the balance between IC like, deep IC work and mentorship?

Ben Wilson [00:49:58]:
Once you get to a a certain level in the IC Oregon Engineering, you are strongly encouraged to be a mentor. So that's separate than management. So managers like, that that's a weekly task for them. Every direct report, they're having a customized conversation with that person. It's going through, like, what are you working on? What are you struggling with? How can I help you? What do you wanna do next quarter? What do you wanna do next year? Are you happy? You know, typical management questions with employees, make sure you know, their number 1 priority is make sure that people are happily working, and they're gonna stay there and continue to do great work. And they don't regret, you know, signing a contract to work here. That's the big thing. And if they are unhappy, how to make them happy if they're good? But for the mentor mentee thing, it's it's dual sided as well.

Ben Wilson [00:50:59]:
So most mentors are themselves mentees in our structure, And it's done in such a way that your folk the mentor is is adapting what they're imparting, be it technical or soft skills to their mentee for whatever that person needs. So you need to be, you know, senior enough and hope to have seen a bunch of stuff. And most of those people at at least the data engineering are people that have done the I've worked for big tech company. I've worked for 3 or 4 startups. I've worked for another big tech company, and now I'm here. So they know the dynamics of the their mentee's team. Even though they don't might not know all those people, they know where all those people came from in general. So they can give them advice on, like, this is how you should interact with these people, and this is how your team does what it does and why people do this sort of thing.

Ben Wilson [00:51:57]:
And it it it makes for it sort of creates this environment that's unique here that I haven't seen in at many other places where there's this camaraderie. It's like everybody knows that everybody's going through this sort of help each other out thing. So everybody just is helping people out that they don't even know in almost a a friendly way. Like, there's no there's nothing coming from on top, from c suite down saying, you will work as a team or we're all 1 family. I've never heard that sentence at this company, and I don't think I ever will. Everybody just gets it. We're we're a tribe, and you help out your tribe members in whatever way they need. So it's super successful here.

Ben Wilson [00:52:47]:
I've seen that forced function at other companies just create misery and make people's hate what like, oh, I got a meeting with my mentor today and see how bad this is. And they have a checklist that's printed out, and they have to go through bulleted items from HR. And everybody's like, this is such a waste of time. So it works here because people

Michael Berk [00:53:12]:
care. Interesting. Okay. Cool. So I will quickly summarize. Lots of cool insights here. Some things that stood out to be were startups are successful when the startup is right and others people don't see the potential, but that industry is still growing. Don't look at number of employees.

Michael Berk [00:53:32]:
It's a bad metric. For outsourcing your work, you can use the or, like, 7525 split between 2 providers and then make that a dynamic ratio if 1 is doing better or worse. And then for promotions, make yourself redundant. If you're essential to your existing team, you can't be promoted out of your role. And then finally, for Regal's decision framework, they look to be more in the 80% right than the 100% right category. They also start with opinions, but ensure that they use data to validate the opinion, and then growth is the north star. So, Alex, if people wanna learn more about you or Regal or anything else, where should they go?

Alex Levin [00:54:08]:
Yeah. You know, please visit regal.io. You can always email me at hello@regal.io. I'd say, particularly, you know, if you're a consumer organization, the big contact center, you know, trying to figure out how to better engage your customers throughout, you know, know, either new customers or existing customers throughout the life cycle, feel free to reach out. We always love chatting about it.

Michael Berk [00:54:27]:
Cool. Alright. Well, until next time. It's been Michael Burke and my cohost, Ben Wilson. Have a good day, everyone.

Ben Wilson [00:54:34]:
We'll catch you next time.