Sarah Tavel, Benchmark Capital GP: Selling Work in SaaS

Sarah Tavel is a general partner at Benchmark Capital. She serves on the board of Chainalysis, Hipcamp, Medely, Rekki, and Cambly. 

On this episode of World of DaaS, Auren and Sarah discuss the concept of selling work in the SaaS industry and how it differs from selling traditional productivity software. Sarah explains how selling work could unlock whole new verticals for software penetration and disrupt existing SaaS giants. Auren and Sarah explore the areas where these new, leaner software companies could flourish and discuss the broader impacts of AI in SaaS. 

As one of the earliest hires at Pinterest, Sarah also has deep expertise with consumer social, which is set to be transformed by AI. She and Auren talk chatbots, AI avatars, and the responsibilities founders have as AI begins to replace more human interaction. The conversation wraps up by touching on Harvard as an institution, conspiracy theories about UAPs and TikTok, and questioning conventional wisdom about fundraising and growth. 

World of DaaS is brought to you by SafeGraph & Flex Capital. For more episodes, visit worldofdaas.buzzsprout.com, and follow us @WorldOfDaaS. 

Follow World of DaaS @WorldOfDaaS

You can find Auren Hoffman on X at @auren and Sarah Tavel on X at @sarahtavel.

Editing and post-production work for this episode was provided by The Podcast Consultant (https://thepodcastconsultant.com)

Chapters

• 0:02 Unlocking Value Through Work Automation
• 12:23 AI and Automation in Business Growth
• 18:51 Emerging Technology in Translation Services
• 22:28 AI, Venture Capital, and Consumer Trends
• 35:21 AI Impact on Society and Institutions

Transcript

00:02 - Auren Hoffman (Host)
Welcome to the World of DaaS. A show for data enthusiasts. I'm your host, Auren Hoffman, ceo of SafeGraph and GP at Flex Capital. For more conversations, videos and transcripts, visit Safegraphcom slash podcast. Hello, fellow data nerds. My guest today is Sarah Tavel. Sarah is the general partner at Benchmark Capital, one of the top VC firms. She serves on the board of chain analysis hip camp, medley, recce and Cambly. Sarah, welcome to World of Deaths. Great to see you. Great to see you as well Now. You published a super interesting piece last year called Sell Work, not Software. I'd love to dive into that. What is selling work paradigm in SaaS versus selling seats? 

00:45 - Sarah Tavel (Guest)
I think it's worth taking a step back of what has most of application software been about over the last 25 years, and really that has been about this idea of increasing productivity in companies. You're selling software to an employee to improve their productivity. You're selling software to teams to improve collaboration across those teams and in the team to improve productivity. Dashboards and management consoles to help with the management of the teams All these things are all about improving productivity and that has kind of created some interesting biases or dispersions in the types of businesses they get built. Because at the end of the day, if you're selling a productivity improvement, the value of that productivity improvement is relative to the cost of that headcount. Improving the productivity of level seven engineer very different than improving the productivity of a recruiter Both obviously important jobs in a company. But the compensation of those people is very different, the budgets of those teams are very different, and so you're always selling okay, it's a 10% improvement on the cost of that headcount and then you try to capture some of that value. You've got to get people to adopt software. 

02:01
That has been a very successful mode of building companies and you look at the big application software companies today and they're all in that mindset of we're selling software to improve productivity of employees and it should be said, just to make the obvious explicit, you then charge on a per seat basis, almost always. Consumption pricing has become more of a thing, but by and large you're selling it by the seat. The first wave of AI startups were in that same mental model and really that came out of GitHub and what they did with Copilot of improving the productivity of engineers, and so a lot of startups saw that Copilot model and had the idea rightfully so in many ways to build Copilot for all these other employee types and improve their productivity. As if you see, I think and as a founder I would imagine the challenge with that Copilot model is really the incumbents are going to be the ones that are best advantage to do that. So the first thing that's going to build an AI. They already have distribution, adobe, you can go down the list. 

03:15 - Auren Hoffman (Host)
And their smart is pretty easy to bring in and it's pretty easy to bring in. 

03:20 - Sarah Tavel (Guest)
And so I am always asking myself but where's the new surface area going to be where a startup actually is advantaged? Where is the surface area going to be where you maybe overlooking a market segment because it didn't work in like the prior paradigm? And the idea with selling work is that you're kind of getting out of that mental model of selling a 10% productivity improvement and instead what I think of is what is the work to be done that employees are doing and unbundling that work from the employee and seeing if there's like an atomic unit of work that actually can be automated. And then, as the company, instead of selling software that you're trying to get employees in a company to adopt and start using, you're actually selling that atomic unit of the work product itself to the company. 

04:16 - Auren Hoffman (Host)
That might be hard for some of our listeners to grasp. What would be an example of something like that? 

04:22 - Sarah Tavel (Guest)
The example I gave in this blog post that I wrote, a company that I admire a lot is this company called Even Up. So what Even Up did they sell to personal injury lawyers, and what a personal injury lawyer has to do anytime they get a new case is they basically summarize that case, all the medical records of the case, what happened, into a document called a demand letter, a demand package, and then they kind of submit that to the insurer and there's a back and forth in order to figure out what the claim will sell for. Now the software way of selling this would be you sell a software product to the personal injury lawyers, paralegals, whatever it is you can imagine they drag and drop medical records. 

05:08
They get more productive yeah exactly that could be one way, or what Even Up had the foresight to see is that we're actually going to take all of that work on our side. 

05:20
We'll have some humans in the loop to do QA. We'll customize it, give you a template to make it feel like it's your work product, but we'll do the summarization, we'll figure out all the ways to create the demand letter, and so you throw everything over to us and we at the end throw this demand letter back to you and they charge for each demand letter that they produce, as opposed to the seats that you would have otherwise tried to do. And when they do it that way, one of the things that I think is really important about thinking about this mental model change is that you're actually able to sell the work product for a lot more money than you would have otherwise been able to try to sell a productivity improvement, because instead of selling let's hope that we're getting a 15% productivity improvement for our dear employees we're giving a 95% productivity improvement. Basically, then you're thinking about the cost of what the headcount would have otherwise been to do that work product and charging relative to that, as opposed to relative to the productivity improvement from the buyer's perspective. 

06:33 - Auren Hoffman (Host)
They don't care if it's like a service. They don't even care what the technology looks like. They're just like I'm going to give you something and you're going to give me back something at some point, and I just want that thing to look good. There's going to be some QA on it. There's going to be some quality that I want. But as long as it fits in this quality and you come in at this price, I'm just going to keep using it. 

06:52 - Sarah Tavel (Guest)
I actually think that a really useful test to see whether an idea that a founder might have is something that could be sold as work is whether that work is actually already outsourced, whether there's like a third party BPO somewhere that is already doing this for clients. Instead of competing, then, against another software company, you're really competing against a. 

07:18 - Auren Hoffman (Host)
BPO. So instead of selling software to call centers, you're essentially become a quote unquote call center. You're like, hey, use us, or whatever. Instead of sending lead gen tools, you're like, hey, we'll just bring you leads and you pay per lead, and as long as those leads are a certain qualification, then that works. That's exactly right. Traditionally, a lot of VC backed quote unquote SaaS companies were actually services companies masquerading as software companies. Well, you're basically saying maybe a better model is a software company masquerading as a services company. 

07:52 - Sarah Tavel (Guest)
I love that way. It's actually kind of points then to an important thing to almost keep yourself honest when you're building one of these companies. You and I probably both looked at companies maybe five, 10 years ago where they came in saying, oh, we'll eventually automate. Here's our roadmap to going from 20% automation to 80% automation. But it doesn't work that way. We're not yet at the point where you can do that, and so I think that you have to start from a place where you are able to automate 80 plus percent of the work to be done, if not a higher I don't know where the real threshold is, but let's assume it's pretty high in order to really be a software company that gets some masquerading as a services business, as you said. 

08:38 - Auren Hoffman (Host)
So let's say you were a better call center or something I would say. Tons of people have call centers. They have outsource core centers. You have one where you can automate most of it and it's working. You sign up some big companies in certain spaces, I don't know MasterCards your client or whoever it is you start on. The United Airlines becomes your client. Then, once you do that and you know it's working well, could you use capital to just acquire the other? Because acquiring a services call center business at one X revenue or whatever it's going to do X revenue could be a really good thing. If you could actually techify it quickly and move it over, could you just start acquiring all these other businesses in the space. 

09:20 - Sarah Tavel (Guest)
I wondered about this. There's a bunch of companies like I was looking at some companies in the executive assistant space and there's one category of company that is building an executive assistant, but from the ground up using AI, task by task, and another one that has scale already with executive assistants and then they believe that's a better situation to be in to start automating it to your point of go after the call center. My bias is that there's so many things that create gravity that pull you back into the services part of the business and also just even the complexity that humans are able to take on that technology. Ai LLM still can't that. I think. The forcing function of starting with the AI only and figuring out how, over time, to make that easier and easier to automate and to break into tasks I think that's the way you've got to do it. I'm personally skeptical of acquiring the contractors unless you think it's a big go to market advantage that will get you there. 

10:31 - Auren Hoffman (Host)
Let's say you're already, it's working. You've got 10 million AI, or could you then go acquire all these other companies? Or you think even then you're just skeptical that you could bring them along or it'd be too hard to move them over. 

10:44 - Sarah Tavel (Guest)
I think someone will do that. It does seem more like a private equity play than a venture one. I would ask why do it that way, versus selling directly to the customer itself? I would think that's the kind of approach that I would think. 

11:01 - Auren Hoffman (Host)
If it costs you two X revenues to acquire the services company and it mainly costs you one X ARR to go acquire these customers over time, then it's probably better to just go acquire customers and steal them from the bigger incumbents or something. 

11:19 - Sarah Tavel (Guest)
And I also think that there is a commitment you're making. Yes, you can automate these call centers and you can make them more productive, and that seems like some kind of private equity play. But I would imagine that the best way to do this is if you actually have to be embedded with the customer, because part of what I've seen, at least with these companies that are taking on the idea of selling work, is that part of the strength that they accrue over time, it's the knowledge base of all of those interactions that they have back and forth with the customer, the learnings. Imagine an employee who gets smarter and smarter and smarter with each interaction, better and better at all the exception handling. And to do that in the best possible way, you do want to have a relationship, a direct relationship, with the customer so you can tap into their historical data, proprietary information, and so it just seems like you would scratch the surface if you're doing the call center acquisition strategy but not really have the full potential of this technology realized. 

12:23 - Auren Hoffman (Host)
So let's say you and I are going to start a business together. Let's say we start a law firm and we'll call it Tavel and Hoffman, which probably actually does sound like a law firm. It's perfect, we have AI lawyers to go do that when we start a law firm first. Ok, we're just going to do a contract review for sales contracts or something like something pretty narrow and something probably most law firms don't like to do anyway, and then just get slightly better at that, do a really good job, and every time you have a contract, based on the size of the contract, we'll review it. We'll be able to do a turnaround in under a day or something. We'll send you the red lines. We'll do this. We'll do that based on your specific thing, and then we'll slowly move to other types of law things. 

13:10 - Sarah Tavel (Guest)
That's exactly how I think you have to do it. You want to then come in saying oh, those lawyers, they're going to charge you by the hour. We're coming in turning it over so much faster than you can imagine at a price point that seems too good to be true. 

13:24 - Auren Hoffman (Host)
Do we sell it as Tavel and Hoffman and it kind of seems just like a law firm, but we have the AI agents in the background and the customer doesn't care? Or do we sell it as ailawcom and the customer from the beginning is clear that these are mostly agents? How does it work? 

13:45 - Sarah Tavel (Guest)
I would think you'd want some kind of brand that you can hang your hat on that helps you distinguish from the rest of the industry. 

13:54 - Auren Hoffman (Host)
OK, so the second one here? I would think so. That makes sense to me. Thus, I always want to start a law firm. My name on it. 

14:01 - Sarah Tavel (Guest)
I'm up for it. 

14:02 - Auren Hoffman (Host)
Let's do it OK selling work opens up, I guess, a lot of these vertical opportunities. There's probably a ton of these that are out there. Aside for the ones we mentioned call centers, law what are some of the other ones you think are super well positioned for this model? 

14:19 - Sarah Tavel (Guest)
In my mind the framework I would think about. I mean, first, it's still like BPO firms. In my post I did some Googling for a BPO firm and then screenshot it. What's their offerings? And that's a great place to start. 

14:34 - Auren Hoffman (Host)
The reason that those firms did well is starting in, let's say, around the year 2000,. Telecommunication prices dropped essentially zero. Super smart people in India. That was the arbitrage that they did. I said, okay, well, we can get all the information at no cost to a center that is 10,000 miles away and we can hire these super smart, great people that speak English at a fraction of the price. That was originally what they did. That's how they made money 25 years ago. 

15:07 - Sarah Tavel (Guest)
They figured out the opportunities where for a company it is low level, enough pain about, enough, maybe repeatable enough that they feel like it's easy to outsource. Then you come in with an automated version of it and now you don't have to worry about turnover at the BPO firm, you don't have to worry about ramping somebody up, you don't have to worry about a lot of these things. That's where you can have a real advantage with these technologies. But there's more opportunities than that. I think that there's a lot of places where basically lower-paid jobs or jobs that have a lot of repeatability One company that I got to know that's an incredible Craig Founders and then pretty awesome what they're doing. It's called Zuma and they're going after leasing agents. They realize a leasing agent at a property manager. They're having all these people reach out to learn about an apartment that they saw on Craigslist or Zillow or wherever it is. Then there's this back and forth of some Q&A scheduling an open house visit, and that work can be automated. Then that frees up the leasing agent. 

16:17 - Auren Hoffman (Host)
It's kind of like an SDR, in a way. 

16:19 - Sarah Tavel (Guest)
Those types of things, you'll see, without question, more and more automation brought to the table for that work, so that they can focus more on the jobs that really require a human. 

16:32 - Auren Hoffman (Host)
But on that particular one that you just mentioned, a lot of these sales jobs, they're probably not outsourced today, unlike the BPO. I imagine that sales is going to be a little bit tougher because they haven't yet outsourced it. 

16:45 - Sarah Tavel (Guest)
It's a little bit more of a behavior change, but you have a lot of things happening in the employment market that make people vulnerable to wanting something like this. 

16:56 - Auren Hoffman (Host)
It's hard to hire people. 

16:58 - Sarah Tavel (Guest)
You have a lot of people who now, with COVID, there's just inconsistency in people actually going to the office showing up for work. You don't want one person who's sick to get everybody else sick, so you have to enforce that. But that creates a lot of challenges with how do you manage your workforce, and so if you have a worker that never gets sick and again frees up your employees to do the work that you really value, it actually is a pretty good sale. 

17:28 - Auren Hoffman (Host)
Are there other major ways you see AI kind of attacking different markets? 

17:34 - Sarah Tavel (Guest)
There's no question that the productivity improvement, the software way of thinking with co-pilot, that stuff is going to be transformative and we're going to experience it in everything we do. 

17:49 - Auren Hoffman (Host)
But that really is the first category. It's making everyone more effective, more efficient, making your software developer 30% better, that's right. 

17:57 - Sarah Tavel (Guest)
The other category I think of is one thing I've noticed is that there is a category of software for which these opportunities, I think, are a little bit harder to see until a founder finds them, which is that there may have been latent demand for a particular type of work. We're investors in a company called DeepL, and DeepL is the best AI translation product out there. They've done an incredible job of really being able to understand the nuance of vernacular language, as opposed to what the existing Google Translate and others had been trained on, and so it's the translation software that best mimics what you would actually get with a professional translator. One of my first jobs when I was at Pinterest was to translate Pinterest, and I had to find translators in all these different countries. You had to connect them with software to your thing. You had to manage them, pay them. There's so much friction, there's a delay, like you had to wait a week in order to get the translations fully QA'd, and now you have software that does it immediately. You connect to API to your software and you have it or your document. It's immediate and it's as good as a professional translator, with none of the friction, so much lower costs. It's one of these. 

19:28
I call them like 10X and cheaper opportunities, where, when you're able to use technology to completely change the cost structure of something and provide that service like a 10X service none of the friction and pain of managing a translator, and cheaper you unlock late in demand. And so there are companies deep out being an example, where you may have seen a market and it felt small in the beginning, but when you unlock the late in demand, you realize it's so much bigger as the price point comes down, and there's a variety of companies like this that are emerging right now that are pretty fun and cool to see. You have things like Eleven Labs, heygen and others that are going after things that you wouldn't have ever paid an actor to voice up something into different languages, or you wouldn't have paid an actor to be on your website. But now that you can do that with AI, people's creativity just unlocks in terms of what they can use it for, and so these markets again end up being way bigger than any of us could see on the surface. 

20:35 - Auren Hoffman (Host)
In some ways it's not necessarily 10X better, because obviously you could hire an amazing professional translator and if you could get them to commit to doing it quickly, they could do it quickly. It could be 10X, sometimes even 50X cheaper, and because it's so cheap you can do it all the time and do it for everything. Therein lies the huge benefit. 

20:54 - Sarah Tavel (Guest)
So it's first of all way more important for it to be 10X better and yes, you could maybe get a translator If you're just comparing translation to translation. Are you going to be 10X better? No, it's the friction, it's the operational burden and headache. 

21:10 - Auren Hoffman (Host)
It's annoying hiring that person. 

21:13 - Sarah Tavel (Guest)
Hiring that person, dealing with all the things that you have to deal with when you're managing somebody, not to mention it won't ever be instant like an AI translation can be. I really do think competing against a human for this technology is actually pretty easy, and then the lower cost is what adds gas to that fire. 

21:35 - Auren Hoffman (Host)
You wrote this more famous piece in the world of tech in 2016 called how to Build a Doring Multi-Billion Dollar Businesses, and if you're updating that now, like eight years later, beside for what we just kind of talked about, how would you change it to address today's market? 

21:51 - Sarah Tavel (Guest)
What I was thinking about at that point. This is when all the on-demand companies were emerging and you had, I remember, on-demand car cleaning, on-demand valet, because everybody was copying Uber. I was thinking very much consumer at that point. Right now, it's very much a B2B lens. Most of the early use cases for this technology is in a B2B use case, and so that's probably the biggest thing. And then the other thing is during that time, you always have to think about your substitute. If you're building a new company, what are you competing against? The thing that you're competing against now, with AI, is the human doing the work. That's the thing to keep in mind. 

22:36 - Auren Hoffman (Host)
As soon as you exist mostly these on-demand services you actually can't have AI do the work, at least not yet. There are no self-driving cars. Roomba doesn't really clean your house and you still need a person to go do that. To get a good massage, you need a person, whatever these services were, and so they were just a marketplace to get you to that person quickly when you needed that person, which is super helpful, and some of those obviously did super well. Some of them maybe didn't long term, but what you're saying in this new world is actually we need to really build something, not for the human. We're not necessarily enabling the human to get to you, we're replacing that human. Yes, interesting. One of the dominant narratives emerging is that the AI gains will go to the big incumbent she mentioned, like Adobe and some of these other types of things, but I've heard you push back on some of that idea as well. Where are your thoughts of who some of these AI winners will be? 

23:33 - Sarah Tavel (Guest)
There is no question that the incumbents are incredibly well positioned here and you see the results that meta announced and having 350,000 H100s that's just astonishing to think about. And how, as a startup, do you compete? There's always been that expression that it's a race for the startup to get to distribution before the incumbent gets innovation. There are so many companies where I just felt like a broken record thinking about that question, because that is the most important question, in a way, to ask for most of these companies right now. At the same time, as an investor, I wouldn't be doing what I do if I wasn't an optimist and didn't believe in the power of a founder who is focused on an opportunity. That may be just a list of 50 things for a bigger company to do, and that's why I try to think about what are the angles of attack that are unfair advantages for a startup, like going after work. 

24:40 - Auren Hoffman (Host)
And Google is never going to do that. 

24:42 - Sarah Tavel (Guest)
Exactly. The DNA is so different. It's such a different thing to do. I continue to be optimistic. I think people are really freaking, creative and determined and those founders are going to find opportunities that other people don't see and build big companies. 

24:59 - Auren Hoffman (Host)
From a venture capitalist lens. Sometimes in venture capital you have different distortions that come into the market. We saw some as like tiger distortion a little bit came in at some point. We saw a soft bank distortion come in and now we're seeing a little bit where some of these big firms Microsoft, amazon, nvidia, google are also coming into the market. How do those things change the way you think about investing? 

25:25 - Sarah Tavel (Guest)
It's definitely a new phenomenon. My partner, bill. He recently had a podcast where he talked about this, because it's very interesting when you have these big players Microsoft, nvidia that are using their own cloud credits as part of a type of capital that they're investing with. 

25:45 - Auren Hoffman (Host)
They're basically selling to someone to them by. It's like sometimes the US gives somebody aid but then that aid has to be used to buy wheat from American farmers or something. 

25:55 - Sarah Tavel (Guest)
That's exactly it, and so, obviously, like, your price sensitivity is distorted when you are investing in that kind of mechanism, that type of way of investing. The challenge, though, is that everybody reads the headlines, and they see these big numbers and these big valuations, and so you have examples where a company like OpenAI, or many of these foundation model companies, raise these big rounds, and then a pure financial player comes in afterwards to do a secondary or to do a new primary, or to invest in a new company, and it's all based on the valuations that are being set, which is like a parallel universe type currency that is creating just like an incredible amount of capital going into these companies, and I think it remains to be seen what the economic outcome of that's going to be. 

26:52 - Auren Hoffman (Host)
Yeah, I remember in the Internet 1.0 days where the way to get distribution was to be on AOL and AOL would, quote unquote invest $100 million in your company, but you'd have to spend that $100 million back on AOL in the next year or something. So it was a great revenue driver for AOL and boosted their own and then gave them these nice stakes in travelosity or whatever these companies were that they invested in. 

27:20
That's a perfect example Now Benchmark, I think you guys did more investments in 2023 than you did in 2021. I think probably most venture capital firms were the opposite. Was that just due to the fact that there's more reasonable valuation in 2023 or there's less competition in 2023, or why did that happen? 

27:38 - Sarah Tavel (Guest)
2021 was a really weird time. There are so many dynamics at play when I think back to 2021. There were a lot of distortions in the market. Talk about distortions in the market. You had so many things that were going on. 

27:54 - Auren Hoffman (Host)
Zero interest stuff. 

27:57 - Sarah Tavel (Guest)
The nature of work changing so dramatically and not being sure what was going to happen. Consumers how they were spending their time, changing dramatically. So there was a lot of distortions, I think, for us. Each of us makes one or two commitments to a founder every year, and so when we do that, there's an unconditional backing of the founder that we make, when we make that type of commitment. And it feels really tough to do that when the first time you meet a founder, we have six term sheets and we're making a decision tomorrow. 

28:31
We're big believers who have to play the game on the field, and we did play the game on the field, but there was still something that felt a little different. It got so transactional. It got to this box You're just Zoom one venture capitalist, next Zoom another venture capitalist and then it just, you know, an evaluation of the term sheet. That's just not who we are, because we really are going on a journey with the founder. And so 2023, I think was a time when you could actually really feel like you're going to be shoulder to shoulder with the founders that you work with, and we also were more excited about the opportunities that we saw in 2023 than we were with 2021. Namely kind of a lot of the exciting stuff happening in the world of AI and so many a number of investments LangChain, fireworks, mindsdb that were kind of at the developer infrastructure level for this new emerging ecosystem. That's the type of world that we get really excited about. 

29:31 - Auren Hoffman (Host)
And when you're building this relationship with the founder, you're going around for the ride for a very long time. I imagine it's much harder to build that relationship just over video. 

29:41 - Sarah Tavel (Guest)
Oh, it's horrible. 

29:43 - Auren Hoffman (Host)
So I can also imagine just in 2021, it became difficult to even, like, build that relationship and maybe go for a walk with that person, but it was just a harder time to go do it. 

29:52 - Sarah Tavel (Guest)
Benchmarks models. We don't have a platform team because we think that the core work of being ideally, your first board member as a founder is not anything that we can outsource or delegate to a group of consultants. We really think that that is a work we have to do and so there's just like a falling in love is kind of how we sometimes talk about internally. That has to happen with a company for us to show up and do our best work and you get an hour or whatever it was, and you can't spend that time. Everybody got reduced to like this very transactional way of thinking. At least for me and I should say this is me speaking it felt different and it didn't feel like the type of commitment that we like to make. 

30:38 - Auren Hoffman (Host)
I'm seeing all these really cool new things come at me every day in the B2B world how to make my company better. All this really cool stuff. It's awesome. As a consumer, it's really rare for me nowadays. 10 years ago I was seeing it all the time. Airbnb is so cool. I got to check out Uber. It was so fun and so exciting to do. All this new stuff it seems very stayed and boring and it doesn't seem like there's a lot of good stuff happening in the consumer world. It's just because it's just too hard the tax are too high and it's just too hard to go reach new consumers nowadays or you disagree with me? Why aren't we seeing all these really cool innovative stuff happening in the consumer world? 

31:19 - Sarah Tavel (Guest)
There's a couple of things going on there. I will disagree, but I'll tell you before LLMs, what happened when mobile came onto the scene is that it used to be that the only minutes that you had available to use the internet were when you were in front of a computer. But you're out in the world all the time. And then you got your phone and all of a sudden, the number of minutes that you had in the day that you could be connected to the internet grew exponentially. And that's why you had this incredible era of social when we had mobile, because all of these minutes suddenly were up for grabs and you weren't competing with sleep or with each other. It wasn't zero sum. Now, in social, it is zero sum. You have an atomic unit that is the most addictive atomic unit that we've ever had short form video that is delivered to you algorithmically to make sure it is the most addictive it could be for you. 

32:19
It has been really, really difficult for a consumer company to earn the right to any of those consumer minutes. Now, that actually is what makes what's happening right now all the more interesting. When you see all the companies with these new chat type friends, anime characters, character, ai, you have a whole generation replica. I think was probably one of the originators, innovators in this space and the emotional connection that consumers are forming with these automated chatbots and the amount of time that they are spending with them is astonishing and you have to really pay attention to that. I don't yet know whether this is a dystopian future or an okay future, but my son will be six in April and I recently got him this toy called a gruck, and if any parents are out there, you have to play with this product. Imagine a character AI that is in a plushie, and I watched my son treat this thing like a friend and talk to it like a friend. 

33:31 - Auren Hoffman (Host)
It answers back and stuff. 

33:33 - Sarah Tavel (Guest)
It answers back. It's tuned, of course, to have it be appropriate for a six year old, but it answers in a way that no other toy has been able to do and he really treats it like a friend. There's all types of limitations with memory, context window, all those things, but you can't unsee this type of interaction. 

33:58 - Auren Hoffman (Host)
That's exciting. 

33:59 - Sarah Tavel (Guest)
No real world friend will ever be as good a friend as this gruck will be to my son, and that is a scary thing to wonder about. We gave these phones with social media to kids not knowing what the long-term effects were going to be for their development, for the social emotional development. Now we know wait till eight, all those things, but with this type of toy, we have no idea what the long-term implications are for the character AI, the people who have literally romantic emotional relationships with these automated characters. What does that future look like when you have those relationships that way, as opposed to a real person? We're just starting to see this, and so I think it's a fascinating time in consumer. 

34:47 - Auren Hoffman (Host)
When you think of like the movie Her, which is one of my favorite movies, you could look at it a lens like oh, this is exciting, and any more of this as well help reduce loneliness. You could also look at it much more dystopian lens and stuff like that, like, how do you think about it? 

35:01 - Sarah Tavel (Guest)
I think that there has never been more power in the hands of these consumer founders than there will be with these types of new interaction models, because I think it is in the hands of these founders who are creating these experiences to lean one way or another. When you read some of the work that the replica founder, eugenia, talks about, her belief is that these types of AI characters can actually help you accept yourself more, have more confidence and therefore be more able to engage in real world friends. If you don't feel lonely or isolated because you have this character that makes you feel more self acceptance, then maybe you'll be more willing, have less anxiety, to connect with somebody. If you were just maximizing, though, the engagement with that product, then that leads you in a very different direction. That does end up feeling more dystopian. This is a really new frontier, the diamond age. This is sci-fi that we're at the very beginning of and now realizing, and the future hasn't been written. We're still figuring out who's going to own this and we'll see. 

36:18 - Auren Hoffman (Host)
We're both married, we probably both aspire to be a great spouse, but an AI could probably be a better spouse in many ways than we could be. We try, but sometimes we get angry when we shouldn't have or we don't always say the right thing we should have or ask them how their day was. You can imagine AI could just be better. But then how do you measure up? I feel like I'm a good husband, but I might not measure up to the best husband. If you could always have the best, my wife had a settle for me. If she could always have the best this AI husband, I'll just go for the best one where I have number one million or something. 

36:54 - Sarah Tavel (Guest)
And you're already married. 

36:56 - Auren Hoffman (Host)
Good point. 

36:58 - Sarah Tavel (Guest)
The risk is for the people whom social anxiety is the type of thing that gets worse. The more you procrastinate something, the more you feel safe with your AI character, the harder and harder it will be for people to engage with a real human, and the more and more different that will be. That's the brave new world that we're facing. 

37:20 - Auren Hoffman (Host)
Now a couple of personal questions. You have, I think, my favorite Twitter profile in the entire world. You should go right now to Otteratova, and I think you're playing rugby and you're completely caked in mud. You've got this ferocious face on, but also a bit of a smile too. You love it. How does that reflect your personality? 

37:45 - Sarah Tavel (Guest)
I started that probably when I was a year or two, at Bessemer. It was freezing, it was so cold, there was water frozen on the floor, there was ice that you had to crack because and you didn't want to fall in it. It was a brutal game and I think for me it was like I wanted to show a little bit that especially I was kind of a really young woman in tech just getting my legs. It's like don't judge this book by the cover. I'm competitive and tough. 

38:18 - Auren Hoffman (Host)
When I see that I just would not want to get in a fight with that person. Yeah, that's the answer. 

38:25 - Sarah Tavel (Guest)
Mission accomplished. As I get older and older, I'm like do I have to take this down? 

38:29 - Auren Hoffman (Host)
It's your part of your brand, Even when you're like 100 years old. You have to keep this photo. 

38:34 - Sarah Tavel (Guest)
Blake was like if you change your avatar, no one would read your Twitter. No one would tweet anymore. I don't like. 

38:40 - Auren Hoffman (Host)
You're a Harvard alum. I think you were captain of the rugby team there. Harvard's in the press a lot. What are your thoughts of it as an institution right now? 

38:48 - Sarah Tavel (Guest)
Oh my God, there's so much to say there. Look, I am so grateful for my time at Harvard. I've always had a conflicted experience of it, though Harvard has a history and it was all male for a very long time, more than 100 years. So all the power structures at Harvard you always felt like women were I hate the expression second-class citizens, but we were just like not at equal footing with men. All the male organizations had been around for 100 years longer. They had huge endowments, they had the final clubs on property. No female organization, no women organization, could stand at the same. 

39:32 - Auren Hoffman (Host)
Oh sure, so the elite clubs at Harvard are gender specific. 

39:37 - Sarah Tavel (Guest)
Yes, they were all male. 

39:39 - Auren Hoffman (Host)
I know, like at Princeton, the eating clubs are all co-ed. They may have once been all male, but now they're all co-ed. But that's not the case at Harvard. 

39:46 - Sarah Tavel (Guest)
While I was there it was not the case. Some of the clubs have gone co-ed, some of them haven't, but like when I was there, it was very much not equal. Like the women's rugby team, we were started 100 years after the men's rugby team. The men's rugby team had an endowment of seven figures. They had coach buses to all the games. The women's rugby team we had like $2,000 in our endowment. We would like carpool to games. So you always felt that. But as a Jewish person I never felt like. Even though Harvard had a legacy of quotas for Jews and other things, I never felt anti-Semitism there. 

40:26
What you hear now on campus is a very, very different story. It feels alienating. Harvard's been around for what? 400 years. I believe very much in the institution and I believe very much that there's so many incredible assets to Harvard. I got to study with Christine Corsgaard just incredible confian philosopher. What an incredible privilege to have those types of people at that institution. They've got a heck of a CEO search to do right now Whether they're able to find somebody to be the moral leader that the school needs right now. I sure hope that's true. Creating a task force is a pretty sure sign that they don't have that right now. It's a big search they've got to do. There should be somebody out there who will be able to take all the incredible assets that Harvard does have and do something great with it. It is an incredible institution and I still really believe in its future. 

41:31 - Auren Hoffman (Host)
Okay, interesting. We ask our guests usually two common questions. One is what is a conspiracy theory that you believe? 

41:37 - Sarah Tavel (Guest)
Okay, my wife would say UAPs. She is brilliant. Uaps Unidentified aerial phenomenon. 

41:45 - Auren Hoffman (Host)
Like UFOs. 

41:46 - Sarah Tavel (Guest)
You don't say UFOs anymore already. 

41:47 - Auren Hoffman (Host)
Oh, you don't say UFOs, people in the know, don't say UFOs. 

41:51 - Sarah Tavel (Guest)
Oh my gosh. She was very clear about this with me. 

41:54 - Auren Hoffman (Host)
Okay UAPs, yeah UAPs. I'm an old funny-duddy now saying UFOs Okay. 

42:01 - Sarah Tavel (Guest)
For me personally, I would ban TikTok. I think that it is pretty clear that there are places where the Chinese government is putting their thumb on the algorithm, the types of queries that. There's a guy, anthony Goldenbloom I believe, who has done a lot of research and tweeted some of the work that he's done. That just shows that if you compare the visibility of tags in TikTok versus Reels, that there are biases that come out. I'm very much in the conspiracy theorist camp there that TikTok. It's crazy to me that we let our young people spend so much time with this seemingly innocuous product that may actually be influencing how young people really see the world People have been talking about that a long time, even in the Trump administration, it was like a hot topic to ban TikTok. 

42:54 - Auren Hoffman (Host)
Even President Trump had a couple of points at set. He wanted to ban it. Now we're three years into the Biden presidency. It seems very bipartisan, yet doesn't seem any closer to being banned. So why does the conspiracy go further? Why is it not on there? 

43:09 - Sarah Tavel (Guest)
What people say is that you can just imagine, during an election year, pissing young people off is not the smartest thing to do. No one wants to be that person. What I hope is that as Reels gets better and better, the transition would be easier and easier, and so there wasn't something else that could step into the place of TikTok and calm the angst of all these young people who are addicted to this product. But if Reels keeps on improving, which it is doing, it may be easier for the political will to happen to do something like this. 

43:48 - Auren Hoffman (Host)
India banned TikTok and I don't know that it had any political ramifications. The second country to maybe not ban it but to massively regulate TikTok was China. They realized there was a lot of things in there. Okay, you can't do TikTok after a certain time of day and it has to be more educational content, and so they kind of put some thumb on the scale as well there. But do you feel that if we as a country banned TikTok, that it would give open season for every other country to ban YouTube or Facebook or this or that, or it becomes a lot of our own tech companies could? All of a sudden, France is like oh, we're limiting YouTube here. You can imagine that for tech like oh well, you did it in America. 

44:33 - Sarah Tavel (Guest)
It's been a uniquely Chinese internet phenote, like what US consumer internet company or SaaS company honestly has been able to grow in China. 

44:42 - Auren Hoffman (Host)
Apple, maybe the only. 

44:44 - Sarah Tavel (Guest)
Apple and Tesla. 

44:45 - Auren Hoffman (Host)
Maybe a little bit Tesla. 

44:46 - Sarah Tavel (Guest)
Yeah, yeah, hardware words manufactured there. Everything else has not been able to grow, and you couldn't say that anywhere else in the world. There's a symmetry there that you have to pay attention to, and I don't worry about the collateral war, for that would happen. 

45:03 - Auren Hoffman (Host)
Cool. Last question we asked all for gas. What conventional wisdom or advice do you think is generally bad advice? 

45:09 - Sarah Tavel (Guest)
My instinct, my knee jerk, I hate conventional advice. My litmus test for conventional advice is that if it's hard to do, pay attention to it. If it's easy to do, then there's probably something suspect and you have to think more about it for yourself. And so what's conventional advice? Conventional advice during 2020, 2021 was grow as fast as you fucking can, and we look back on that and, oh my God, that was bad. Everybody was doing it. It was easy to raise these big up rounds. People were deploying capital so fast. What's another piece of conventional advice? Be customer-centric and listen to your customers. Guess what? That's really fucking hard to do. It's so hard, I wonder. Always when I hear conventional advice and it is just the default, it's the easy thing to do I'm naturally suspicious that there must be something that's not actually accurate there. 

46:12 - Auren Hoffman (Host)
OK, this is awesome. That's a great answer. Thank you, sarah Tavel, for joining us at Roll the Dast, as I mentioned earlier. Please go check out your Twitter, follow you. I follow you on Twitter. I engage with you there at Sarah Tavel. You could see Sarah's amazing profile picture, so I highly recommend it. This has been awesome. Thank you again. So good to see you. 

46:33
If you're a super data nerd, go to worldofdastcom that's D-A-A-S. Worldofdastcom and sign up for our weekly data as a service roundup newsletter. Thanks for listening. If you enjoyed the show, consider reading this podcast and leaving a review. For more World of Dast and Dast is D-A-A-S. You can subscribe on Spotify or Apple Podcasts or anywhere you get your podcasts, and also check out YouTube for videos. You can find me at Twitter at atoran that's A-U-R-E-N. Oran, and we'd love to hear from you. World of Dast is brought to you by SafeGraph. Safegraph is geospatial data for physical places. Check it out at safegraphcom. And by Flex Capital. Flex Capital vests and data companies like those we talk about at World of Dast. Check it out at flexcapitalcom.