Jun Kwak, co-founder and CTO of Overjoy, joins us to talk about why he believes every B2B SaaS company will have to become a service business, how Overjoy went from selling software to delivering sales outcomes for consumer brands, why tangled data is the strongest moat against the frontier labs, how AI agents could weaken wholesale marketplaces, why brands need websites AI agents can read, why small brands chasing big box stores often abandon their most profitable channel, whether agents will turn premium products into commodities, and what three national debate championships taught him about running a company.
We’ve shared some highlights below, together with links & a full transcript. If you like what you hear/read, please leave a comment or drop us a review on your provider of choice.
Subscribe to The OSVerse to receive your FREE copy of The Infinite Loops Canon: 100 Timeless Books (That You Probably Haven’t Read) 👇👇
Links
Show Notes
Jun Kwak & Jim O’Shaughnessy
Jun Kwak: Co-founder and CTO of Overjoy. Connect on LinkedIn
Overjoy: The AI-powered wholesale growth platform Jun co-founded with Alex Han. Overjoy
Overjoy on X: Follow Overjoy on X
Jim O’Shaughnessy: Host of Infinite Loops, founder and CEO of O’Shaughnessy Ventures, and author of the New York Times bestseller What Works on Wall Street. Follow Jim on X
Full transcripts: Transcripts and highlights from every Infinite Loops conversation. OSV newsletter
Books & Ideas Mentioned
Zen and the Art of Motorcycle Maintenance by Robert M. Pirsig: The source of Jim’s story about a friend outraged that his BMW motorcycle was fixed with a scrap of aluminum can. Zen and the Art of Motorcycle Maintenance
Influence by Robert Cialdini: The classic text on persuasion Jim cites when asking whether AI will master it. Influence
Sebastian Barry: The Irish novelist whose McNulty book A Temporary Gentleman Jim discussed with his AI agent. Sebastian Barry
Consumer Reports: Jim’s model for a trusted product-rating source in a world of AI shopping agents. Consumer Reports
Collapse of Silicon Valley Bank: The 2023 bank run that locked up Overjoy’s seed funding. Collapse of Silicon Valley Bank
Faire: The wholesale marketplace Jun cites as the market leader. Faire
Electronic data interchange (EDI): The ordering system big box retailers require, and one reason small brands fear that channel. Electronic data interchange
Highlights
On Agent Marketplaces
Jim O’Shaughnessy
We mentioned Amazon, right? They’ve turned off access to the agents. I think that ultimately is not going to be a winning proposition. And my guess is, and I’d be interested in your point of view, that Amazon itself will have its own group of agents. But I don’t know how much the consumer is going to trust that those agents are going to actually find them either the best product or the best price or whatever. How do you see this whole thing unfolding? This is kind of a broader question than just your specific company […] is there an opportunity for a new marketplace that caters directly to agents, for example?Jun Kwak
Yeah, that is definitely a possibility. But I think these marketplaces require both sides to be signed on to that platform. The analogy would be: would you want to only crawl through Wikipedia articles or Reddit articles, or would you want to get information from any source on the internet? I think that’s the analogy here, and the reason why even...Jim O’Shaughnessy
I like that one. That’s really good.Jun Kwak
Maybe I should be using that more. But yeah, basically, the reason why... I mean, when we first started, we thought, there are these big wholesale marketplaces, and would it make sense for these brands to set up their own wholesale portal? The truth is, there are always buyers that are not on those marketplaces...Jim O’Shaughnessy
Right.Jun Kwak
...and you still want to do business with them, and that’s why you always need a direct channel. And agents make it even easier for the buyers to find, keep buying from and build relationships with these small and mid-sized brands on their direct channel as well. So I do believe in that direction, and in a sense this actually disintermediates the marketplaces that were able to build that walled garden because of the limitation of human attention, which is now gone with agents.
Edited Transcript
Jim O’Shaughnessy
Well, hello everyone, it’s Jim O’Shaughnessy with yet another Infinite Loops. Today I have Jun Kwak, the co-founder of Overjoy, one of the first portfolio companies that O’Shaughnessy Ventures... As opposed to our earlier investments, this one was one of the first that O’Shaughnessy Ventures as an entity made. Jun, welcome.
Jun Kwak
Thank you for having me.
Jim O’Shaughnessy
So I thought that what might really be helpful for us and for our listeners and viewers today is for you to give kind of an account of the journey, right? I think a lot of people thinking about founding a company, especially young, might have a lot of questions. And I think that, having seen how you guys have progressed over the last several years, you’ve got a lot of answers. But first off, tell us a bit about the company and why you decided, “You know what? I think rather than trying to get a job at McKinsey or whatever, I think I’d rather start my own company.”
Jun Kwak
So yeah, happy to share a bit about Overjoy first, and how we decided to start Overjoy instead of getting a job elsewhere. At Overjoy, we are basically delivering sales and revenue teams to consumer brands. We actually started as a startup building a software stack for small and mid-sized brands to scale their wholesale channel. When we first interviewed a bunch of brand founders, even though they started their business as a direct-to-consumer model, which is, imagine you shoot Meta ads, Instagram ads, and direct them to your own website, a lot more of these brands were trying to sell to other businesses, like, say, local boutique shops or other local stores. And that’s a pretty expensive channel, because you need to invest in a full sales team and you need a dedicated software stack to run that.
So that’s how we started building tools to help brands scale that channel. And that went pretty well. We scaled and were able to get to a decent revenue milestone, but we still felt like we weren’t really solving the biggest pain point for these brands, which was that they still needed someone to run the software stack. So that’s when we started talking to more folks in this space, and we even saw some of our users taking our software and turning that into a consulting business and selling it at a 20x, 40x price. So that kind of helped us land on more of where we are today. But happy to share more about that in detail.
Jim O’Shaughnessy
So tell me a little bit about the process in determining what made you target this particular market. Was it a big search across all of the available opportunities? Was there just a glaring empty space there? What was it that made you do this, as opposed to any number of other...
Jun Kwak
Yeah. So I think that from very early on, I had a... So I guess I can start from my journey as a 13-, 14-year-old.
Jim O’Shaughnessy
Please.
Jun Kwak
So I grew up in Korea, and at that time somehow I got into debate for the first time in English class. And that was fascinating, because if you’re in Korea, even though it’s a peninsula, it’s basically an island, with North Korea. So basically the only way to learn about how the world works is through news, and debate felt like you can actually talk about those topics with logic. So I was fascinated by that and I started going to debate tournaments, and I ended up winning three national championships in Korea at age 15. And then after that I came to the realisation that you get to talk about all these topics, but nothing gets solved. So that’s when... One of my friends was into game development, so I thought it was cool to build stuff on your computer. So I started self-teaching myself how to code, and built my first game, built my first app. And that was kind of the second half of my high school. I actually quit debate, all the public speaking, mock trial, everything, and I went on learning how to code, and felt like if I get good at this, I can actually go build things to solve problems in the world. And yeah, I ended up going to Yale. I was one of the first 10 Presidential Scholars from Korea, so that helped me move here to the States and start my journey here. And I first felt like I should be a good engineer, so that was all I was focused on throughout Yale.
I ended up working at Google AI, Snapchat, Robinhood. And I think that’s when I started feeling the itch of, “Huh, I feel like I can now build something that people will actually use. What if I actually go try to build my own startup?” So I had a stronger itch to build my own startup, but I did not have a clear picture of what that space would be. And I think in that sense, going through a new grad accelerator programme helped a lot. So I ended up going to the SF Bay Area. We had two big hacker houses, and that’s how I actually met my co-founder, Alex. We went through so many different spaces. It’s embarrassing to talk about right now, because there were so many bad ideas that we explored. Even in terms of different areas, we went from fashion design tools to crypto tools to all these different sorts of spaces. But then I came to realise that you should really like the people that you build for. And after going through different domains, when I was talking to these brand founders, they’re all super passionate about what they built, what they make, and they’re just really cool people to hang out with.
So even though we had a very vague, murky idea of what problem we wanted to solve, it was pretty clear that we wanted to build for the brand founders. I remember talking to one of our first customers, his name is Tyler. He founded a brand called Cinnamon. They want to be the Hershey’s for cinnamon and horchata products. And I still text him, and he’s still very passionate about his brand. And I think from very early on, even though we had very little idea about the space, it was very clear that this is the space that we do want to be in, and the people that we want to be around.
Jim O’Shaughnessy
So, quite the underachiever. By the way, what did your debate experience teach you that you brought over to becoming a founder? Because I found one of the rarest things that I look for in founders are people who don’t want to just yap about things, right? Because debate, I love debate, I think it’s great and I think you can learn a lot of really good skills in it, but you’ve got to want to actually do something about it. But what did all of your debate experience teach you, and what did you bring that sharpened you as a founder?
Jun Kwak
That’s a great question. I just thought of two things. One is, since I’m CTO of our company, I lead the product engineering team, and there’s a lot of overlap between computer science and debate. Code is essentially logic. So when I was making that transition... I actually never thought I would be in a STEM field as a kid. I really didn’t like math, but coding, computer science, it felt like a lot more overlap with debate, in that you need to be logical, you need to understand how machines work. So that helped me through that transition.
Another thing is more about how I work with my co-founder, Alex. So, fun fact: he’s also a debater too. He did debate as a high schooler. And one superpower that we have as a team, I think, is we are not that emotionally attached to the arguments that we have. And this is almost different from ego. It’s more like, how can you handle disagreements? How can you handle these attacks on things that you strongly believe in? And because we both come from a debate background, I think it’s easy for us to detach ourselves from what we are saying, and also from the attacks that the other person is throwing at me. So that oftentimes helps our arguments tremendously, because basically we’ll talk about what to build or how to build a specific search feature, and we both feel fine flipping our stance an hour later because, logically, the other side makes a lot more sense. So I think it helps a lot for us to uncover the truth behind what to build, how to build, how to sell, all these problems that you face as a founder.
Jim O’Shaughnessy
Yeah, I think the ability to pivot, the ability to have an open mind to a variety of paths forward, it increases your cognitive space enormously. And if you have, like, “We’re gonna build X,” right? And then X becomes the only thing, and “Nope, that doesn’t serve X. Nope.” You can miss a tremendous amount. And you actually experienced this, right? Because you started realising there was a whole different segment that you probably should be servicing. Talk a little bit about that, about the services part of the business.
Jun Kwak
Yeah. So basically, more and more we are calling ourselves and branding ourselves as a service company, because what we are realising is small and mid-sized brands lack the resources to spin up a full sales team. And even though you give them the best tools, it’s hard for brands to use those tools, or even orchestrate the agents to get to the most, say, AI-pilled or agent-pilled sales team. So this was a pilot experiment that we actually started running earlier this year. The whole time, we’ve been giving them search tools, email campaign tools and CRM to create a full sales pipeline and turn them into opening orders. “What if we just deliver the outcome?” was the thought that we had. So we had a first set of brands where they still had access to the software, but there’s no need to do that. Just talk to our team and we will run it. Just tell us what you want: what outreach targets you want, what messaging you want us to benchmark. And with that plan that we agreed with the brand, we just went on and launched the entire outreach campaign and outreach sales process. And it turns out the brands were happier that way. Brands got more visibility into how things were going. And it was also not that far from how brands were working with, say, in-house hires or other agencies. But we were able to deliver that in a way more efficient way, but also with better outcomes, because we owned the software stack, we owned the agent stack.
So yeah, now we have three main services. We have an outreach service, where basically we deliver interested buyers to these brands. We have a conversion service, where we own, say, 100, 200 active deals in their pipeline, and we run all the manual work of email follow-ups, sending them sample products, figuring out the pricing and discounts for these wholesale buyers. And the last one is reordering and expansion, where what’s interesting for these physical products, unlike, say, software sales or insurance, is that they’re not on subscription. You need to constantly go through the sales process, even for your existing customers. So every time there’s a new product launch, you need to let them know, get pre-orders and upsell your existing customers on these new products. So we have the third service that’s responsible for that. And with that, we are seeing two-person brands that had never had a sales hire be able to expand their existing accounts through our service.
Jim O’Shaughnessy
So I want to come back to that, because I’m very intrigued by that part of your build-out. But let’s do a case study. Somebody is a small brand, let’s say they’re here and they’ve got a little shop, and they don’t know about you guys yet, but then they meet you and they hire you. What do they look like when they discover you? What kind of operation are they running? What are the kind of sales numbers, et cetera? And then what do they look like after you’ve been fully embedded into their company?
Jun Kwak
Yeah. So the truth is, we work with small brands that are less than 1 million GMV, to 1 to 50, and then 50-plus, and the needs for these are all slightly different. Say it’s a very small brand, say one or two founders just run the brand. Oftentimes they don’t have a working wholesale engine. So they come to us and they want us to build out the entire...
Jim O’Shaughnessy
And let me stop you for a second. Can you give us an example of what they are selling?
Jun Kwak
Oh yeah. So one of our brands, they sell leather goods, say, leather coasters. And they are a two-person brand based in Minnesota, and they have an existing channel, but because they don’t have a sales team, they just have their products on online wholesale marketplaces. The market leader there is Faire, and because of how marketplaces are structured, they tend to pay a very high commission for the sales that they get through those marketplaces. And there’s very little control over how you can increase your sales through those marketplaces, because you don’t have an in-house salesperson. So when they came to us, they basically had 10,000 customers. They only had opening orders, and they never got reorders from those accounts.
And imagine you try to do this as a founder: you would probably dig through all your sales history. The biggest problem actually is it’s really hard to combine your email data and your order data in one place. You might think, “Oh, we can use agents to do that,” but to run that at scale, on 10,000 accounts from five different sources, becomes an actual data problem. So oftentimes it’s really hard to build that source of truth for someone to go in and figure out a very simple question of: which of my accounts should I go out and tell about our new product launches? So when they actually came to us, we took over that customer management side. And every day we have 50 to 100 customer win-back campaigns, where we know exactly what they purchased before and what is the best product that we can recommend for them to purchase next. And with that alone, they were able to go from zero to $15,000 in Overjoy-attributed revenue in the past 30 days. And that number actually keeps going up, because we are building up the account management motion for the wholesale channel.
Jim O’Shaughnessy
And give me an example of what the workflow looks like that determines: okay, so they bought this leather coaster or whatever. How many different products are you researching that are available for this company to offer a returning customer? And then what is the algorithm or the agent that determines, “No, this is the one we pitch them,” and why?
Jun Kwak
Yeah. So basically, when brands onboard to our service, even though they think that they are only getting the sales service, for us to power that, we do onboard them to our entire software platform. So when they come in, we sync with their Shopify, Faire, QuickBooks and all their email communication channels as well. And we build out their system of record, where they can see their whole list of accounts. And with that, what we are able to do is start grouping them, and basically building out their customer profiles for their wholesale channel. So as we group them, we can then see what the top-selling products are for each of these groups.
So say I’m a brand selling protein shakes: I can group my customers based on fitness gyms, yoga studios and grocery stores, and then rank which of my products are the best-selling products per category. And then, on a per-account basis, these agents can go and look up all the products that each account bought and has not bought, and recommend the best sellers.
Jim O’Shaughnessy
Okay, so you look at what they bought from your client, then you look at what they bought elsewhere, and then you find the most adjacent thing that your client is offering, to offer to them. What does the funnel look like, in terms of you... You are now part of this smaller brand thing. How do you do A/B? Take me through. Is there A/B testing, or is it just, “Yeah, no, they’re buying a ton of this, and this we offer”? Do you have to have a price advantage? Do you have to have a quality advantage? Is there a secret sauce?
Jun Kwak
Yeah. So for any sales outreach, there’s target, there’s message, and there are a lot of tactics on what the call to action should be. So those are the three things that we do experiments on, on behalf of these brands. I would say that the experiments actually happen more for new-account campaigns than for customer campaigns. But what that looks like is, the same protein shake brand will come in, and we’ll actually launch three different campaigns for different ICPs. What’s actually interesting is, even though a lot of brand founders think that they would land better in certain accounts, because that’s what they saw in the retail channel and wholesale channel, oftentimes, depending on how they bundle their products and how they price them, what lands for wholesale is completely different from their direct-to-consumer channel. So we do...
Jim O’Shaughnessy
Let me stop you there. And my guess is they’re not aware of that.
Jun Kwak
Yeah, exactly. Which is actually a hard thing for us to convince them of, and that’s why we do run experiments. So when they first onboard, they fill out this outreach form. That’s our seed for their brand brain, or internally we call it the brand wiki. So each brand founder will come in and identify, “Oh, I want to reach out to CrossFit gyms, and I want to reach out to grocery stores, and I might want to try yoga studios, but I’m not sure how it will end.” Then, with that, we have our local business database that maps all the local businesses across the US and Canada, and we are able to find the most similar retailers to the ones that they’ve identified.
And with that, we start having two or three different campaigns with slightly different email copy, and we change the brand voice, we change what the offerings are, because sometimes you want to just go and send sample products, sometimes they want to give them more of a catalogue and let them know more about the brand. So we will run these experiments, be able to return it back to the brand, and double down on the ones that work better.
Jim O’Shaughnessy
So it’s an iterative process. And do you often find that your clients... do they just go along, or do they fight with you, or what’s the normal experience like?
Jun Kwak
Yeah. So I think this is actually the beauty of how we’ve transitioned from more of a software and subscription-based business into services. Because if you go two years back and look at our pricing page, we did the very standard SaaS seat-based pricing. Brands come in, they pay for, say, two seats, five seats, and they pay for credits, which is essentially how many new leads, new potential email contacts, I can go and run outreach to. Now, fast forward to where things are now: it’s more outcome-based pricing.
Brands pay for 20, like 40 interested buyers, and there we actually abstract out the parts of identifying the targets, creating... It doesn’t matter if you do three campaigns or five campaigns, and run it on 500 buyers or 3,000 buyers. Essentially we deliver the outcome, as if this is how you would work with, say, an SDR or a sales agency. It’s more of a sales process that runs on autopilot.
Jim O’Shaughnessy
Who are your competitors?
Jun Kwak
So right now there are more of the newer players. But I think on the sales side, we haven’t lost deals to anyone in our space. I think what we are more excited about is, I think there are a lot of adjacent problems that these brands... In terms of competitors, right now in the pure sales space, we don’t see any dominant players that are competing with us. There are one or two newer players that we are still taking a look at, but we haven’t lost any deals for our current service offering, on the outreach side and the conversion and retention side of our services. What I do believe is a lot of the SaaS companies in CPG will also move towards a service motion.
And even for us, our customers are asking us to expand our offering to inventory management and accounting and other back-office problems that they internally face. So I think it’ll be pretty exciting in the next 6 to 12 months to see how this pans out. And I think there are players that are on you to build accounting software or other types of software stack that brands use that are also trying to expand their service to more of a broader offering.
Jim O’Shaughnessy
So let’s take a moment and envision the person out there who is watching or listening to us, and they’re contemplating founding their own company.
Jun Kwak
Yeah.
Jim O’Shaughnessy
Take me through, like, the six things that happened that were completely not on your radar, and how you dealt with them.
Jun Kwak
Great question. I don’t have six in my mind at first, but I’ll start with things that just pop into my mind.
Jim O’Shaughnessy
Sure.
Jun Kwak
I do remember when SVB went down and basically all our funding was locked up in this account, and I actually remember calling you, Jim. We were like, “We don’t have any cash flow. This is it.” And that was actually right after we closed our seed round, and we were like, “Okay, finally we can relax. We can now not worry about money for a bit and go build cool stuff.” And yep, that happens. And I remember calling you, and fortunately over the weekend everything was sorted out. But yeah, I remember you assuring us that you actually went through multiple bank runs, and how this will be handled, and that was a big relief, because as a first-time founder I did not expect something like that could even happen.
Other things: because we build a software stack, we have a lot of integration partners, and there was a case... I would not mention the exact name of the company, but with one of the integration partners, we have one shared developer account where all the engineers share logins, and no one really checks the inbox. And they were sending us all their updates and how we need to make sure we are compliant with the integration partner’s app settings, and no one was checking it. And one day all of our customers got an email, a very scary email, that, “Oh, the Overjoy app will be removed from this app store.” And we had to reassure our customers that it will not affect the process, and we were able to sort it out. But that was also one of those moments where we did not expect that things like that could happen. So yeah, a lot of this firefighting.
I would say one more story there would be... I do remember very early on, our search... I think this is also when all these language model providers, OpenAI, Anthropic, they do require you to go through a moderation endpoint, so you don’t accidentally try to look up harmful content. But what happened was, one day we got a warning, and then they actually stopped our API access. This was very early on, when it was just me and Alex and our founding engineer at the time, but they basically cut our access to OpenAI, and we were digging through the data, and it just turned out, in our local business search product, there were some people that were looking up adult stores, and there’s no filtering on whether that hits the OpenAI endpoint or not. So we had to make sure that restriction was lifted, which we did through the people we have in our network. But all these things were definitely things that we did not expect, did not sign up for. But it’s just part of being a founder, I guess.
Jim O’Shaughnessy
And I think also, for example, the Silicon Valley Bank, having been through a lot of that, would freak any new founder out, I think. And what about your clients, though? Did they get wind of all of that, and what was their reaction like? Because I also want to hear a bit about, like, are your clients all of a certain type of personality? Obviously they’re not technical, would be my guess. Correct me if I’m wrong. But is there kind of a translation tax that you have to pay to get them to understand, “No, no, wait, you can do this so much better”?
Jun Kwak
Yeah, I have literally three different thoughts that just popped into my head at the same time. So there are definitely... I think these macro changes also do affect our customers too. It could range from... I remember when there were all the changes in tariffs, and all these small and mid-sized brands were affected by that, and they had all these questions. It’s very unpredictable in terms of what the next six months of inventory would look like. And we actually saw those things do affect them more emotionally, even though logically you should keep your outbound engine, you should keep running your sales process. But then most brands, especially as a two-person brand, get scared of all these macro trends, and they would actually pause or change their strategy for how they approach wholesale.
Those are more economic trends. Even on the technical side, a lot of brands struggle when there’s a new launch. Grokbot comes out, Claude’s new update comes out, and they want to use these agents, they want to use these frontier models in their workflows, but they do struggle a lot with how to incorporate that into their products. We even faced that too when we were rolling out our sales agent products earlier this year, and this was before we had the service motion. So we do all these tutorials, we try to teach how these brands can adopt and use these agents to run order processing, run outreach and run customer management. And it was actually pretty difficult to explain the best practices when the best practices change in real time. What we recommended a month ago is no longer the best practice, because with new models we actually do make changes to how things work.
So that’s actually one of those things that contributed, that helped us make that decision to be a lot more service-forward. Because, especially when things are changing so quickly on agents and AI, I think it’s a lot easier for us to deliver the outcome with the best orchestration of software, automation and agents for these brands, than for these brand founders to understand what’s the cost difference between Opus and Fable, and “How should I... What is MCP? How do I connect them to my workflow?” So that actually was our conclusion, and that seemed to land better with our customers lately.
Jim O’Shaughnessy
And are they of a type of personality where they’re just very eager to outsource? And I mean, your services are growing wildly. How do you scale that, though? How is that going to scale? Let’s say you continue with the kind of demand you’ve been seeing recently. Scale it for me.
Jun Kwak
Yeah. So with one of our new services, the full sales service, in the past four weeks we actually had 4x operational efficiency. Now, how we do that is a couple of things. I used to only lead the product and engineering team, but four weeks ago Alex and I made the decision that I should also be leading the account management team, because internal operations is basically product when we are a service company. So now I’m going to talk about what this unlocks, but then also all the new problems I’m faced with. The unlock actually is that now it gives immense clarity to a product person in any company like us. Because now you know what the service you’re trying to sell is, and what the price point is, first, before you go and build all the software and product that leads up to it. In a sense it removes all the guesswork from the product development cycle.
Because most of tech startup 101, how you build a product, is you talk to users, you have these hypotheses on what those problems are, and basically, if you build X, it will solve Y. And then you take it to the design and development phase, where you try to figure out the best mock-up, the best prototype to solve that problem. You implement it, you run QA, and then you bring it back and see if that actually yielded results for your customers. Now, in the service world, you can actually go straight to what these brands want, with an actual sales expert delivering that service. And in-house, you keep labelling: what are the things that are making this person inefficient, what are the things that we can actually automate with our software automations, and what are the things that we can actually improve with agents?
So that’s basically... it flips how we build products. You go from outcome first, even on the internal product building side, and you actually go backwards and keep improving the efficiency of how we deliver the service. So that’s what it unlocks for us. Because before, Alex and I would have all these debates on what will happen if you launch this new ordering product. We don’t know, we don’t know how it will land, we don’t know how people use it, but we’ll just need to test it out. And you spend all this time trying to build that out, and then it may or may not work. It basically removes the guesswork and adds a lot of clarity on what you’re trying to optimise for.
Now, internal operations-wise, it just means that I have a whole new set of problems I need to deal with internally. And what we came to realise is our account managers are essentially the experts. The reason why services work is because what we believe is, for all of these services, whether it’s outreach or sales management, you have, say, 5 to 10% where it requires human judgement. Whether it’s, does this fit with what this brand wants to represent? Does it fit with their strategy for this year? All these strategies that require human judgement and review. And the rest tends to be more amplifying that and actually executing that plan.
So that’s essentially how we’re trying to design our services and how we deliver them, where each account manager owns the strategic part of all these sales processes. And then the actual work that is involved there, making sure all of the leads that we do market research on are qualified based on the criteria that this brand sets, making sure all the samples that get sent out match the budget that we actually set, all that actual work that needs to be done on our platform becomes more and more automated with agents and our software stack.
So basically what we essentially did is we built a product-service loop, where we match each account manager with an engineer in our team, and every day we review all the inefficiencies, all the time sucks of our internal operations. And that feeds into how we improve the agent quality, and adding more things to our product roadmap to improve the workflow. So that actually led us to getting the 4x efficiency improvements on our full sales service offering. And right now I can’t share the exact margins, but basically we are able to offer services at close to software-company margins right now.
Jim O’Shaughnessy
And as I was listening to you: are there any universals? Because you’re obviously dealing with very disparate brands, right? Like leather goods and protein powders, you wouldn’t think, share much in common. But is there a universal? Is there an Ur-pattern, if you will, that underlies marketing anything?
Jun Kwak
That is a great question. And right now, what we’ve done is we have these loops, even among these agents, because essentially we have the full data of what all these brands have tried and what works best. And of course there are differences based on their category or their size, but you get to then see all the trends. There are things that we used to recommend anecdotally. Say, if you attach a large file or image to your cold outreach email, it’s not going to land in the inbox. But then we never knew that for sure. Whereas once we hit a critical mass of outbound emails and we actually ran analytics on the best patterns that we should recommend, that actually turned out to be true. But then there are also some counterintuitive things. It actually turns out emojis on subject lines help, but emojis in the body are bad. So we get to actually learn from what people tried and all the actual outcomes, and we can constantly improve what these agents can do on the outreach side.
And we’ve done this in a pretty semi-automated, somewhat manual way. But soon, what we are headed towards, even within how we use and improve our agents, is for all the agents that run on these brands, they’re going to be able to do... basically, I think the space is now calling them self-improving agents. Now they can actually see the patterns of these agent runs across these brands, identify the patterns that actually yielded better outcomes, and constantly improve. And I think that’s why, whenever we launch new agents or services, we never try to launch a standalone one that doesn’t integrate with our stack, because we want to make sure that data flows in, we can actually see the full outcome and whether that was very good or bad, and that flows in and can improve our software and agents continuously. So that’s how we determine which space, which new service, we’re going to launch. It can actually have a compounded data effect of improving our agent quality with those added data sources on our platform.
Jim O’Shaughnessy
And the brands that are your clients, do they grok all this? Do they really understand what’s going on? Or are they just like, “Yeah, no, you do it, we love the fact that our sales are going way up”? Or do they really understand the process?
Jun Kwak
So I would say yes and no. We do expose our software stack to brands. This is actually the reason why brands prefer to work with us than some third-party agencies, because they tend to be pretty secretive about who they contact, and they have their own walled garden of contacts or buyers that they work with. For us, it’s all about transparency. We market ourselves as basically an embedded sales department for your brand, so you should have control over all the data and all the software stack that we’re building for you. And oftentimes they get surprised looking at the CRM that is well maintained three months later, because they only thought that they were getting inbound leads. So yeah, probably not real time. But then we don’t hide anything. We try to give them full exposure to the underlying things that we use to run it, because essentially we are constantly using our software and agents to run this.
But what you pointed to does relate to an interesting new problem that we are facing, which is that how we interface with brands is changing as we evolve from a software company to a service company. Because what we are realising is a good chunk of our customers are now just talking to us on email. They just tell us what they want, and things get done. Right now our product manager, Ramil, and I are thinking a lot about: what is the new interface that we need to build for our brands? Because it is different. The assumptions are different. We’re not having those customers as software users. They’re more like clients for the service, and what we need to report back or summarise or explain about what’s happening is changing because of this transition.
Jim O’Shaughnessy
Has a client ever come to you with a really good idea that you guys thought, “Wow, why didn’t we think of that?” and then you actually incorporated it in your stack?
Jun Kwak
So actually, having the main sales email end with, “Would you like samples? Can I send this to this address?” is now a pattern that a lot of our brands use. But that was actually from a few people who really had this idea that even the brands that sell products that are really hard to send samples of need to create the smallest, cheapest product and send physical samples for the buyers to engage with and understand their brand. So that was actually from looking at and working with sales managers in these brands that had those hacks. And we were able to have more brands benefit from those sales hacks.
Jim O’Shaughnessy
And go through your customer acquisition funnel. How are you attracting customers? Is it word of mouth from your existing client base? What other ways are you getting new clients?
Jun Kwak
Yeah, I’d say right now we are still primarily outbound, but we’re increasingly having more inbound, as now we have 3,000 customers working with us. And I think now there’s more word of mouth and referrals coming in. But for a long time our primary sales engine was our outbound engine. And so basically what we give to our customers is essentially the same as what we use in-house to reach other brands. And what I often tell our customers is: cold outreach email is all about qualification. If you reach out to a store that is a good, compatible buyer for your brand, then it’s a business opportunity. If it’s not, it becomes a spam email.
So it all comes down to: can you qualify and do very deep market research to find the best retailers that match your brand? So that’s what we’ve really tried to crack the code on in the last couple of years, for our outreach engine but also for our brands’ outreach. And Alex went through four big iterations of making sure we can do that. There are a lot of technical innovations that allow us to do things like this. Because models got cheaper and faster, now we are able to use products like Clay to score all these businesses at scale. But then there was even one more unlock this year with Jeff, which is a System 1 model, where, because it’s built to score things, not write things, it’s way faster and cheaper for qualification of business opportunities. So with that, we actually completely rebuilt how we score these brands on their intent, on whether they are engaging with their wholesale channel. And this week we’re actually building that for our customers, starting with our service customers, because this actually increases the quality of leads that they can reach out to for their outbound sales motion.
Jim O’Shaughnessy
And what is the size of your total addressable market here?
Jun Kwak
So right now there are 3 million e-commerce Shopify businesses that sell physical goods, and only 700,000 of them are engaging in the wholesale channel. And that’s essentially because wholesale is a pretty expensive channel to run. So our first goal is to be able to democratise that sales process and deliver this agent-pilled sales department for these brands. But I think this will be a second Shopify effect, where, if you think about how brands were selling in the pre-Shopify era, you would still have to figure out how to put your products online. You’d have to have developers and designers to set up your own website, and then try to figure out how to get them to pay, all that. Remove the need of having your internal tech department, and just go start selling as a one-person founder. There’s still a big barrier to entry for the wholesale channel. And what we’re essentially doing is allowing even the smallest players to be able to grow their B2B wholesale channel.
Jim O’Shaughnessy
Okay, so let’s say I’ve got, again, a little shop with my wife, and we sell, oh, snow globes, and you’re going to pitch me. And what does that pitch look like? How do you get them excited about, like, “Ooh, wholesale”? “That seems like a big headache to me,” et cetera. How do you get them to become a customer?
Jun Kwak
Yeah. Right now, I mean, our sales team probably has a 10x better pitch than I do. But what I should tell them is: give us a shot, and in 30 days we’ll bring you 20 interested buyers that match your brand, that want to stock your products on their shelves. And it’s not just a one-time thing. We’ll build out an engine where we can constantly get you predictable sales opportunities every month, and you only pay when we hit those targets. And once you do that, you will start having a sales pipeline of 50, 100 and 200 opportunities, and someone will need to manage all the communications with those opportunities.
Now, you can decide to hire someone to run that in-house, or you can try to work with an external agency, or let us run that sales pipeline, and we will handle everything from email replies, giving them quotes, sending sample products, giving them marketing materials to push them down the sales pipeline. And essentially this turns a very tedious wholesale process into almost direct-to-consumer-like simplicity, where you basically just focus on the orders coming in and the opportunities coming in.
Jim O’Shaughnessy
So, Jun, are you at all worried about the frontier labs just essentially cloning what you’re doing and offering it as a skill?
Jun Kwak
So I think that’s the reason why I think all B2B SaaS companies will have to become a service business. Because I think there are only two paths: either you become a service business and try to automate not just the software stack but also how you use that, and build your proprietary data, or you just become a tool that models call, and eventually even that gets replaced. So I think there were a couple of reasons why Alex and I, even earlier this year... I think everyone was pretty shocked by the newer Claude models and how that could impact each vertical. And when we looked at our market, the more data sources you need to deal with... essentially, it’s a data structure and data complexity problem, where if your data is flat, say it’s a single spreadsheet and it explains everything, then it becomes very easy for one agent to just take over and abstract away the software stack.
Jim O’Shaughnessy
Sure.
Jun Kwak
But then, once you have two, three, four, with all these relationships in between... because for us, we have a list of businesses, and then contacts, and then their orders, and then each of these sources have locations, and each of these products have inventory data, and this comes from five different data sources. This becomes hard for someone to build, or for agents to build. And that makes it harder for frontier labs to just... unless they’re actually going into that very specific wholesale market. I think that’s where it becomes harder for model improvements to accidentally kill a space.
But then, not just that. I think on the service side, what we are most bullish on is the amount of data that we are getting just for our space, that is also tied to natural outcome quality data. And we are able to constantly improve our service with that data and improve our agents with that data. So this is how we can keep building our data moats through services. And this is why we’re a lot more excited and bullish than afraid about what the next launches are from the AI labs.
Jim O’Shaughnessy
And I don’t know the answer to this, but I would assume that data set in and of itself is very valuable. Have you agreed with your clients that it doesn’t get resold, so then it simply becomes an advantage for you and your company, as opposed to others joining the field? How much time... You’ve been in the business for several years now. When does that data compound to a point where it’s going to be really hard for a new entrant to reach your level of understanding, because you have all of this data versus them starting a new enterprise today?
Jun Kwak
Yeah, I think for some of the more mature services, like the cold outreach lead generation side of our offering, I think we’ve already reached that point, where we have the best data that fuels that engine, but also the outcomes keep improving. That’s where it becomes really hard. Alex and I talked about how, let’s say we have to start everything from scratch, you would have to go through the exact same path to build out this product, and then build out this current distribution, and then get to a point where it constantly improves the model. We reached that point for our lead generation side. For our full sales service side, I think we’ll reach that point in the next two months, because we’ve passed that pilot phase. We’re trying to build data on these brands that are running the full sales process with us, and we’re now going through the first set of loops to improve those agents with that data. So yeah, I think in the next two months we’ll reach the point where we have that for our entire sales process that we offer for our brands.
Jim O’Shaughnessy
And let’s talk results. What kind of numbers are you seeing, improvement for your clients from prior to when they hired you, obviously, but then with the new service model as well? What kind of lift are you offering?
Jun Kwak
Yeah, so we ran some data analysis, and our service customers have a 3x bigger sales pipeline generated through us than brands that are trying to use our software and run it themselves.
Jim O’Shaughnessy
That’s just, like, extraordinary.
Jun Kwak
Yeah. And that happened when we first launched it, for the first month, and now, six months in, that still holds true. So we were able to scale that level of overall sales efficiency boost as we scaled our entire sales programme.
Jim O’Shaughnessy
But what does that curve look like, right? Does your 3x, which is astounding, does that then just flatten out? Does it continue to improve? What does that look like?
Jun Kwak
So I think that’s hard to say, what the next low-hanging fruit would be or how we will get there. Right now we’re focused more on how we can optimise for... Yeah. So right now, actually, with the 3x boost, we were able to deliver that for six months. So we are trying to copy and paste that performance boost, not on the lead generation side, but on the full sales process side. So that means: can we 3x their conversion rate, bringing more of those opportunities to closed deals? And for those customers, can we have a boost on the reorder rates? So basically, since we proved out that boost on the lead generation side, we’re trying to bring that to the rest of their sales funnel, down the stack to their conversion and retention problems.
Jim O’Shaughnessy
And how sensitive are you to macroeconomic trends, and, like, what happens during the next recession?
Jun Kwak
So I think for us the biggest thing is: can we be the most affordable and also the best-performing way to run your sales department? So I actually think that... we still have some brands where, “Oh, because they’ve worked with a sales agency for a while, I want to keep them for a bit,” even though they are offering at a way higher price and actually performing way less efficiently than us. So I actually think that, say, if there’s more external pressure on these brands to run it in a lot more lean way, I think we would be the best in the market, because we are able to offer the same service or a better service at a fraction of their price.
Jim O’Shaughnessy
And that’s what intrigued me and led us to be one of your first investors, right? The disintermediation. My thesis has always been... I call it the Great Reshuffle, right? And that all of the old playbooks are collapsing, and there’s so much fat, if you will, from what the old players were charging, the results that they were delivering. You seem to be vastly better on both. You’re cheaper and you’re delivering better results, as I understand it. How are they still in business?
Jun Kwak
I think what we realise is that this space does move... No, it’s not the fastest-moving domain. But I think what’s interesting is the whole trend of agent e-commerce. Because I think that, in a sense, as more and more sellers and buyers are used to using agents, it actually does change how things get purchased. One of the biggest conventional wisdoms of how commerce works is that there are always these marketplaces that tax all these sellers, because they are able to get you customers, they give you the best shopping experience. And that’s the same for wholesale too. There are these wholesale marketplaces that charge 20, 25% commissions on top of their wholesale price, which is already a discount of their products. And it’s because it is a way easier shopping experience for their wholesale buyers. And that makes sense. Even as a shopper, it’s a lot easier for me to just order things on Amazon than to crawl through a hundred different brands’ websites and try to track what the lowest price is.
But in the age of agent e-commerce, agents can actually go through and compare prices on 200 different websites. And I think that’s why there is actually a world where more and more brands will want to sell a lot more direct, whether it’s retail customers or wholesale customers. And that’s why we actually launched our storefront products last week, and we are actually offering that for free. These software vendors used to charge hundreds and thousands of dollars just for their wholesale portal, because brands that were able to do wholesale had the budget to purchase expensive software. But what we are seeing is more and more small and mid-sized brands will want to set up their direct wholesale channel. And buyers are also going to be able to buy from lots of separate, scattered direct portals, because agents are able to bundle that for the human buyers.
Jim O’Shaughnessy
And I thought that the age of agentic buying and selling would potentially be very deflationary, because the previous market was very inefficient, and agents don’t get tired, they don’t need a coffee break, et cetera. But what do you think about big brands? We mentioned Amazon, right? They’ve turned off access to the agents. I think that ultimately is not going to be a winning proposition. And my guess is, and I’d be interested in your point of view, that Amazon itself will have its own group of agents. But I don’t know how much the consumer is going to trust that those agents are going to actually find them either the best product or the best price or whatever. How do you see this whole thing unfolding? This is kind of a broader question than just your specific company.
Jun Kwak
Is the question basically how these larger marketplaces would respond to this, and what that...
Jim O’Shaughnessy
But also, is there an opportunity for a new marketplace that caters directly to agents, for example?
Jun Kwak
Yeah, that is definitely a possibility. But I think these marketplaces require both sides to be signed on to that platform. The analogy would be: would you want to only crawl through Wikipedia articles or Reddit articles, or would you want to get information from any source on the internet? I think that’s the analogy here, and the reason why even...
Jim O’Shaughnessy
I like that one. That’s really good.
Jun Kwak
Maybe I should be using that more. But yeah, basically, the reason why... I mean, when we first started, we thought, there are these big wholesale marketplaces, and would it make sense for these brands to set up their own wholesale portal? The truth is, there are always buyers that are not on those marketplaces...
Jim O’Shaughnessy
Right.
Jun Kwak
...and you still want to do business with them, and that’s why you always need a direct channel. And agents make it even easier for the buyers to find, keep buying from and build relationships with these small and mid-sized brands on their direct channel as well. So I do believe in that direction, and in a sense this actually disintermediates the marketplaces that were able to build that walled garden because of the limitation of human attention, which is now gone with agents.
Jim O’Shaughnessy
And I mean, it becomes almost a philosophical question. But it seems to me that your insight is absolutely correct. Who wants to just look for information on Wikipedia? No. But I don’t, I want to be able to look as broadly as possible. But then, if we take this argument and continue advancing it, what’s going to happen to advertising, right? Is it going to become very bespoke, in terms of highly targeted to a specific thing? Is it going to go broad, right? And what happens to physical locations? What do you think? Because I’m not a mall person, but I was talking to a friend and they said, “Hey, I went to a mall the other day and it was, like, this really...” Why? But what they said was, it was a ghost town, there was nobody there. More of that? Less of that? A resurgence of, “No, damn it, I want to go store to store and do it the super inefficient way”?
Jun Kwak
I think, actually, I want to take a step back and label that as more of a sales and marketing problem that all businesses face, because which sales channel, which marketing channel has the best alpha does change over time. In a sense, brands were affected by direct-to-consumer, which was based on social media marketing, being saturated, and that actually forced a lot of these brands to change their business model. I think what’s important for these small and mid-sized brands is to be less dependent on specific channels, and be able to adapt to the different changes in purchase behaviours and sales channels. For example, we keep seeing fluctuating predictions: “Oh, would people buy more from malls and big box stores, or would they go back to local?” If you look at the headlines, it literally changes back and forth every two years.
But I think what’s important is for brands to be... and this is where, if you are running this process in-house, there’s one person who’s managing direct-to-consumer social ads and wholesale email campaigns. What we actually saw was, in November, you would start seeing these brands moving their wholesale managers to become customer support for their Shopify website, because they get a spike in their traffic. So basically, because of a lack of resources, they’re forced to sacrifice and give up certain channels depending on how things go. And I think where we can come in and help is, basically, we build out their whole department that is agnostic to these different channels, for them to be able to experiment more and find the best-matching sales channel or marketing channel.
Jim O’Shaughnessy
Do you think that there’s an opportunity for an aggregator? Like, when I was growing up, there was a magazine called Consumer Reports, right? And what they would do is test all of the washing machines or cars or whatever, and then they would say, “This one is the absolute best. This one is the best value,” right? “So it’s not as good as this top one, but for what you’re paying, this is the one you should do.” Given the way the world is going, right? Here’s a real example. Thinking about getting a ping pong table, because my grandson is very good at it. I used to be very good at it. I’d love to play with him. And so I had one of my agents go and look for ping pong tables. You know how many it came back with?
Jun Kwak
Like, if you didn’t specify...
Jim O’Shaughnessy
But then... So I specifically asked it, “Okay, look at rating systems, look at customer reviews, et cetera.” And it does seem to me that there is no single trusted source, or multiple, if you will. Is that going to be a thing? Because with agents having the ability to find every single ping pong table of this particular size, you’re going to have a sorting mechanism based on price, based on quality, based on a variety of these things.
Jun Kwak
Yeah, I think there’ll definitely be a need for verifying the source of what information is most important for any buyer. But the truth is, a lot of the brands are not even... they’re even set behind on that problem. And what happens is, if you don’t have any product information on your website that is agent-friendly, then what wins is louder blogs or Reddit articles that mention your brand. We even saw how pricing for our platform was completely wrong on Google AI Mode, because we just didn’t clarify that on our website, and there are just blogs that mentioned it, and those tend to win over your actual source of data. So this actually makes it a lot more important for brands to control what they sell and how they sell. In our storefront, there’s a big section on their pricing policies and all the brand information that brands get to directly control. Because in these marketplace entries, there’s a limitation on how much story you can tell, and all the information that you can actually put in for your products, your full catalogue.
Jim O’Shaughnessy
And how many of your brands use dynamic pricing?
Jun Kwak
So that’s actually what makes wholesale interesting. For D2C, not a lot of brands use dynamic pricing. It’s hard to maintain, it’s hard to get right. But for wholesale, the whole industry is basically built on dynamic pricing, depending on the strategic importance of the retailer. Even though it’s a very small store, if it has a very high social media presence, or it has the potential to break into the entire region, they would give a larger wholesale discount to those buyers. Same for net terms. With net terms, essentially, depending on how healthy, how big that business is, they are willing to take on more risk and give them longer net terms. So what we came to realise is that wholesale ordering and pricing needs a lot of business context.
So when we first thought, like two years ago, we thought, “Why would we even build an ordering platform? Isn’t that already solved?” But for wholesale, the pricing was the hard part, and it makes so much sense for that data to be right next to the sales data in their CRM. So that’s how we thought we had a reason to build storefronts. And essentially, how can we... Right now it is still being done in a pretty manual way. We built the software infrastructure that allows brands to specify custom discounts and custom pricing for wholesale buyers. But we’re moving to a world where we will have agents and be able to be a lot more prescriptive about, “Okay, given this retailer’s strategic value or their health, you should be giving XYZ in terms of pricing and discounts and net terms for their wholesale orders.”
Jim O’Shaughnessy
And another thing that I wonder about is, again, agents don’t have any of the problems that humans have. We need to sleep, we need to eat, we get cranky, we have bad moods, we get exuberant, we get overly excited. We do all of these things, and all of these things are basically because of our human OS, which is mostly emotionally driven, and emotions probably drive a lot of purchase decisions as well. But in a world where it’s agents, how many things that currently would be considered a premium product become a commodity?
Jun Kwak
I think that actually it will... I always use the word “amplifier” a lot when explaining agents, and same for coding, same for buying. And I think it does amplify the preferences that people have, and that’s also true for buyers making purchase decisions. So even for wholesale buyers, there are the more data-driven ones that look through the sell-through rates, and they would base their purchase behaviours on that. But there are also buyers for whom it’s more important for products to fit their overall retailer vibe, the types of products that they want, the overall atmosphere of that store. And I think that essentially agents could amplify that behaviour.
So I actually don’t think that premium products will be, like, slashed into more of a commodity. Because if that was true, I think you would live in a world where there would actually be less brands, and we would actually prefer to buy more Amazon Basics, because everything becomes... or Kirby products, because that’s essentially what people want. But I think there are always products that, in the equilibrium state, become a commodity, and there are products that require brands, or people prefer to be more branded. And I think it will just amplify both sides of the buying behaviours.
Jim O’Shaughnessy
And do you think that we will ever get to a point where the agents and the stack that you’ve built literally are able to incorporate human emotional reasons for buying, right? I have long believed that most decisions are made emotionally and then they are papered over with rational reasons for why you did it, right? Like, when I was a kid they had these infomercials on TV at night, and I used to love watching them just because of the psychology of it, more than anything else. And you could see that they were literally selling purely on emotions, right? I think it would be really interesting if the agents could in some way quantify that. Or am I just too much of a quant?
Jun Kwak
So I actually did a double major in computer science and psychology. But it’s still a tough... I don’t know the right answer to this question. But I actually do think... I still think that’s true, because if that was not the case, sales and marketing would not matter at all. And the fact that, depending on how you change your subject line, how you explain your product in our brands’ email campaigns, that actually yields different results. That is proof that, depending on how you explain and how you trigger the different emotional reactions for these buyers, it does make a difference. So I think what we’re helping is trying to make that more of a data-driven way. But essentially, I think that is still going to be true on how buyers respond to these different messages.
Jim O’Shaughnessy
I think that one of the things that I’ve given a lot of thought to is we are highly suggestible creatures, we humans. And I think that, especially the newest frontier models... I’ve been having every model since the beginning, when they called it machine learning as opposed to AI, or what does Trump want to call it? Superintelligence. I don’t think that one’s going to work. But I think that they’re going to get really good at persuasion, right? The classic text on it is Cialdini’s Influence. I don’t know whether you’ve read that or not, but we are in many respects highly suggestible. And I can see a future where you just have the AI A/B test ad infinitum. And yeah, of course some of it’s going to be like, “That was just a fluke, that’s an outlier, it doesn’t actually work.” But they’ll be able to figure that out too. And I can also envision a future where people just get so used to their agents buying necessities or basics or things like that, that it bleeds into what we currently think of as premium-type products or services. Which way do you... I’m relying on your debate.
Jun Kwak
Yeah, I’m not sure if I fully got the question. So what are the two camps here?
Jim O’Shaughnessy
The two camps are... the latest frontier models are demonstrably superior to even those of six months ago.
Jun Kwak
Right, right.
Jim O’Shaughnessy
And one of the things that I always test them on are persuasion techniques. They’re getting really good. And I could envision a future where, let’s say I’m trying to sell some premium protein powder, right? And so I have to do all of the research that says the best, and here’s why we’re the best, these independent labs, blah, blah. But I could also see having the models just A/B test different persuasion techniques till they get to the one where the person’s just smashing the buy button, right? “I need this because it’s going to solve all my problems.” As I mentioned earlier, I think a lot of our choices are emotional, and generally people buy things to solve a pain or to achieve some hoped-for outcome in terms of prestige, or their desirability, or their attractiveness, or whatever.
And so I could see one where the models just continually learn from their interactions with we humans until they get to the point where they’re just getting really good at persuasion. But I can equally see people buying basics that they don’t really care too much about. I guess we used to call these consumer staples, right? And then that bleeding into just, like, everything. “I’ll just let the agent get that for me.” Which of those paths do you think is more likely?
Jun Kwak
I see. Basically, I think there are inherently differentiated goods that require persuasion, and then there are commodities or staples where I think there is a right answer. “What is the cheapest toilet paper?” Then you can get a deterministic answer. Whereas, “What is the best couch for our apartment?” There’s no clear answer to that. In the first type of products, I think there are going to be persuasions involved. On the second side, though, I think those are going to be a lot easier for agents to go straight to the answer, and people are okay with just approving whichever cheapest product they bought for those products.
So basically, my answer is, I think it will depend on the product category. There are always going to be what we’ll call luxury goods or branded goods, where there’s a lot more taste involved. And there, I think, there is still going to be human input that is needed, both for the persuasion techniques and on the other side, for purchase behaviours. But then on the other side, for less differentiated goods, I think it will be more of the latter, where agents can actually be autopiloting and going straight to the best answer.
Jim O’Shaughnessy
And then the obvious question also becomes: how do you hack the agent?
Jun Kwak
Are you talking about the seller side or the buyer side?
Jim O’Shaughnessy
Seller side.
Jun Kwak
Seller side. So yeah, I think most of the brands that we work with... I mean, they’re called a brand for a reason. They have a niche angle, either in terms of their audience or the type of values that they represent. And these are important parts, even when we think about their sales and marketing strategy. So the part about how these agents are doing A/B testing on what’s the best strategy, that’s literally what’s happening under the hood for our lead generation services. Because basically we can keep figuring out... we can increase the temperature, try different tactics and see which one lands better.
Jim O’Shaughnessy
So I’m a huge fan of Bob Pirsig’s book Zen and the Art of Motorcycle Maintenance. And he has a story in there that has stuck with me. He rides motorcycles, obviously, and he’s riding with a friend who is driving a BMW motorcycle. And he’s very proud of the fact that he has a BMW, and something goes wrong with it. And the guy is heartbroken, because there isn’t a BMW authorised service centre for, you know, 500 miles, and what’s he going to do? And Pirsig takes a soda can top, takes the little pull circle out, fashions it and puts it on the guy’s motorcycle, fixing it, right? The guy, though, rather than being thankful, is outraged, and he’s like, “How dare you take a soda top and put it on my pristine BMW?” I wonder...
There’s also a thesis of happiness where they ask the question: what would you do if you woke up tomorrow and you were the last human being on the planet? There hasn’t been any cataclysmic problem. There wasn’t a war or anything. It’s just literally all of the other humans disappeared. Everything is as it was. So what would you do? And when they interview people about this question, in the beginning many say, “Well, you know, I’d go to the Louvre and take the Mona Lisa and put it in the palace, and I would live in the palace, and I would drive the fanciest sports car.” But then, as they continued the questioning, they found that after a while most people started saying, “Well, actually, you know, if I really was the last human, I’d probably live in a place that was very easy for me to maintain. I probably would only wear really comfortable clothes. I really wouldn’t care what I drove.” And basically the result was, many of the things that we buy or desire are to impress other people, right? I mean, it’s the whole story of the luxury goods market. And do you see agents having any effect on that sort of thing? Or am I just being too philosophical?
Jun Kwak
I mean, to be honest, when I was thinking about being the last person there, my mind went straight to... there are some Slack channels in our team where there are just agents and just me, right? Or the group chats on Rockbot where there are four agents and only one person there, and just: how do I talk there? What’s my behaviour there? My mind went straight to that point. So...
Jim O’Shaughnessy
Well, let’s answer that question. That’s a much more interesting question. Are you a different person when you are just interacting with agents?
Jun Kwak
I think that is true. Just being in those group chats where I named all the Rockbots as, you know, designer, engineer, product manager, and just talk... Yeah, definitely, how you talk, because it’s very direct. No, like, you know, words...
Jim O’Shaughnessy
No $5 words to make yourself sound smart.
Jun Kwak
And definitely no... embarrassing, but less gratitude for all the work that we’ve done. So that is true. But then, if I think about it, actually, because I’m still reviewing some of the outputs that agents produce, I am so affected by how the agents phrase the work that they’ve done, which essentially is analogous to, say, branding or persuasion techniques. But I think the biggest change for me is how I respond back. It’s very direct, very straightforward. And yeah, that’s the biggest difference.
Jim O’Shaughnessy
I think that actually could be an entirely different podcast, because, you know, I’m a huge fan of the literary fiction author Sebastian Barry. He’s an Irish guy, and a lot of his books are based on old family stories of his family. And I finished the third book in his McNulty series, called A Temporary Gentleman, and I wanted to talk about it, and all of my friends were either otherwise engaged or not around. And so I’m just like, “You know what? I’m just going to talk to my AI agent about this.” And we have a pretty cutting-edge group of agents. And I actually had, like, an amazing conversation. Because at first I said, “Hey, find the earliest reviews of this book, because I have a question about this and this,” right? And it did a great job of finding the old reviews. I think it was originally published in 2012. But then I said, “Well, what do you think?” And, like, literally, all of the training data... I was talking to a PhD in literary fiction, and I’m like, “Wow, that’s a really good point.”
And so I’m absolutely fascinated by how society in general is going to evolve, and how marketing, how brands, how all of that... I obviously can’t prophesise the future. I have ideas, but it definitely, I think, is going to change the way that people do a lot of things. Do you think that these smaller brands can survive without you guys? Meaning, “Yeah, I want to get in, I want to become a wholesale brand as well, I’m going to this agency or this marketplace.” How do they not just get totally obliterated?
Jun Kwak
Yeah, I mean, I think we got to a point where even I was trying to see if I can launch a brand for fun. And I got my fun side hustle, using Overjoy. I mean, I’m not born to be a salesperson, and I don’t know the whole sales space that well outside of how we... before working on Overjoy. But yeah, we were able to get interested buyers through our platform. I feel like, really, a lot of brand founders, some of them have sales backgrounds, but a lot of them are actually a lot more passionate about their products and getting that right. So the vision that I often have is: can we actually help them be abstracted away from all the less differentiated work on sales and back office, so that they can be a lot more focused on product building, brand building? And I think that’s basically what we are unlocking for these brands. A lot of these brands are able to run their wholesale sales process because of us, and be able to have a direct sales channel because of us. And I think that will be a lot more expensive, a lot harder for them to have us profitable. Like even note experiments to unlock that channel. Because how you start that is so expensive.
Jim O’Shaughnessy
And what do you and Alex hope for? If you could have your dream outcome, what is that?
Jun Kwak
Yeah, I think the biggest thing is: can we actually have some of the smallest brands that we have as our customers be able to scale even further, without having to worry about how... by continuing to remove those bottlenecks as they grow out this channel? So we started from getting their first interested buyers, then we moved on to getting them to the first opening order, and then making them become repeat customers. And now they’re running into new problems: how do I deal with pricing, how do I go into big box stores, all these new problems that actually come along the way. Can we keep removing the bottlenecks, so that they can scale them in a very lean way? That’s basically what we’re trying to achieve.
Jim O’Shaughnessy
And just on big box stores... I am the worst person. I’m not a shopper at all. I don’t go to big box stores. I don’t do any of that. Do they survive?
Jun Kwak
Yeah. I mean, there are some insane problems that we do encounter as we look at brands working with big box stores. I do think that there’s an advantage to coming from outside the domain and applying first principles to all the problems that we face. Because you’re looking at how these brands sell to big box stores: they come in huge quantities, they often have net terms where they don’t immediately get paid. They need to first deal with how to source that way bigger order size for these stores, and if they don’t fulfil it in time, through EDI, they get punished. There’s also a big cost for these brands to set up EDI to be able to talk to the order systems of these big box stores. And a lot of these small founders are scared to even enter that space because of that.
And then what happens is, once you are able to start selling to these big box stores, because they need to focus, they tend to give up the independent channel that they actually spent so much time nurturing and scaling. And only when they become big enough do they start seeing the value of these small retailers across the country, these, like, long-tail independents, and then they try to build the sales process again. That’s been a lot of the stories that we see in our brands. So essentially, what we are doing today is that independent channel. Because we talk to these brand founders, they’re like, “Okay, I’m chasing deals in Starbucks, I’m chasing deals with Target.” And the moment they start that, the whole profitable sales channel for independents, they just give up. And that is absolutely insane. They should not do that. So we’re basically solving that problem for these brands, so they don’t give up that sales channel. And then our goal is to be able to keep making other channels as predictable and easy to maintain as what we’ve already done for their local independents channel.
Jim O’Shaughnessy
All right, Jun. I think that you guys have done a ridiculously brilliant job. I love the way that you were like, “Oh, wait a minute, this whole service thing over here makes a lot more sense.” If you’re familiar with the podcast, you know that the final question I give to you is: I make you emperor of the world. You can’t kill anyone, you can’t put anyone in a re-education camp. But what you can do is, I’m going to hand you a magical microphone, and you could say two things into it that are going to incept the entire population of Earth. They’re going to wake up whenever their next morning is, and the two things that you say, they’re going to think they thought up, or had a dream about, or whatever. And then they’re going to resolve, “You know what? Unlike all the other times, these two ideas are two that I’m really going to work on and try to habituate in myself.” What two things would you be incepting in the world’s population?
Jun Kwak
I think my answer to that would be: one is, be ambitious. And then two, be grateful. That’s how I’ve pictured how I want my life to go. I’m going to spend a good chunk of the first chapter of my life doing the most ambitious things. But eventually I think it’s important to transition into a lot more gratitude mode, and be grateful for everything that happened to you and everything that you were able to achieve. One thing that I really like in our team, even as we are growing really fast, is every Friday we have a Friday happy hour, and we do shout-outs, and we share all the grateful things and throw shout-outs at everyone for 15 minutes. And that is always a highlight of my week. And I think it’s important to balance those two, because it’s easy to over-optimise for one of them. So being ambitious and being grateful, those are going to be my...
Jim O’Shaughnessy
I think both of those are great, especially the gratitude one. I think that people really underestimate how much they can change their own life, but also the lives of others, like your shout-outs at your happy hour. Being grateful just keeps your mind in a whole different place. I am ridiculously grateful for everything in my life, and I’m ridiculously grateful that I know people like you, and can enjoy, invest and watch you grow and do things. But also I’m just incredibly grateful that I’m alive at this juncture in human history, because I think the next 10 years are going to be a bit tumultuous, to say the least. But they’re going to be very interesting because of companies and people like you and Alex. Jun, thank you so much for being on Infinite Loops.
Jun Kwak
Thanks so much.