MIKE DUBOE
Hey everyone. Thanks for coming to the Intelligent Marketer. For our third episode today, we had Dmitry Shevelenko from Perplexity, the Chief Business Officer. You know, we came into this conversation with a set of questions that were more merchant- and marketer-focused around what the future of commerce and marketing looks like on these answer engines, with Perplexity in particular. And I would say we came out with a set of kind of unexpected insights on consumer surplus and really, their vision of commerce, in a consumer-oriented way. And I think it ended up getting into really interesting territory that was kind of different than expected.
RISHABH JAIN
Yeah, I'm excited for you all to hear what he has to say about the various ways in which Perplexity thinks about creating surplus for the consumer and making sure that commerce is extremely high trust in a world where there are answer engines, not only helping you answer questions and helping you make those critical decisions, but also when a browser is agentic, what does that mean for how shopping evolves over time and for both the marketer and for the consumer? How does that play out in terms of the information sharing and the trust building that happens throughout the shopping journey? So I'm sure you all are gonna enjoy this one.
MIKE DUBOE
Hey everyone. Welcome back to The Intelligent Marketer. Today we are talking with Dmitry Shevelenko, the chief business officer at Perplexity. It's been a big couple of weeks. New announcements for them with launching Comet, their new browser, last week, and a bunch of stuff that we're excited to dig into around marketing and commerce. Maybe before jumping in, Dmitry, do you want to do a quick intro of yourself and your role at Perplexity?
DMITRY SHEVELENKO
Hey, Mike and Rishabh, thanks so much for having me on. I'm Dmitry, chief business officer at Perplexity. I oversee our external-facing functions, so this ranges from partnerships, comms, marketing, corp dev, fundraising efforts, as well as our vertical initiatives around commerce, advertising, our enterprise component of Perplexity. So every day is varied and exciting and, as I'm sure we are all feeling when we touch AI, every day feels like a month, so excited to dive in on all of it with you.
MIKE DUBOE
Yeah, sweet. Well, for this conversation, we could assume a pretty advanced audience who's already deeply aware of what Perplexity is and some of the stuff you're cooking up. We wanted to really go deep on efforts around commerce and your perspective on agentic commerce and some stuff around the role of Perplexity in the modern marketing stack as well as the advertising ecosystem. Maybe to start, let's spend some time on the agentic commerce and checkout work that you all have underway. Obviously, this week there's been some big announcements between OpenAI and Shopify. Perplexity has had its shopping feature live for some time now. So maybe the right place to start, high level, talk to us about what your perspective is on commerce and shopping and then we'll get into some specifics underneath that.
DMITRY SHEVELENKO
Yeah, so actually, before going into that, this is probably one of the first-ever podcasts recorded using the Comet browser. I'm in Comet right now, so that's a world first here.
RISHABH JAIN
Dmitry, can I ask a follow-up question? Is this the real Dmitry or did you just train ...
DMITRY SHEVELENKO
Yeah, this is Dmitry AI, yeah,
RISHABH JAIN
Okay, okay, okay. All right, we got it.
DMITRY SHEVELENKO
But I am using my Comet assistant to basically give me the answers, so all the hard questions, there is contextual assistance happening. So when we think about commerce, the starting point is — our purpose is to answer people's questions and we really want to be the product that people use when they have to make an important decision. And so when it's a higher-consideration purchase, the ones where you want to do more research, where you want to be more informed, make sure you're getting the thing that matters, the best of it, that's our focus area. Everything we talk about in terms of checkout and that piece of it, it's all much less important than answer quality. Having the seamless checkout, it's a reason for people to maybe want to try Perplexity because they hear about it and it kind of seems new, but ultimately if you're not going to get an amazing answer, it doesn't matter how great and seamless the flow is from there. And so I think [to get] a great answer, you need to have good sourcing, and that's why Perplexity is always invested in building our own search infrastructure and index and ranking and retrieval. You have really good UI UX so that you're presenting the information depending on the query type, emphasizing the most useful components. And so that's really where we spend the bulk of our time right now. If you get that liquidity of people coming to you when they have to make important decisions, that's what gives you the leverage to then do exciting things downstream.
RISHABH JAIN
What are the types of things that you think about in terms of the rails to getting the information that a consumer needs to make the purchase decision? Like you said, outside of the actual nominal checkout flow, there's a wide variety of things that people do before they make a purchase decision. How do you think about making sure that — and outside of citations as well — how do you make sure you present that in a way in which someone feels like they are getting all the information they need to make a purchase? Just like, let's just say a home gym for example, there's lots of considerations around the types of things you would need to know before you make decisions on what you can purchase and whether it'll fit in your space and so on and so forth. How do you think about the different formats of information that someone actually needs to be presented with and what all you're retrieving outside of what a standard consumer might be thinking about?
DMITRY SHEVELENKO
The system is very dynamic, so it's not like ... We don't really hard code any of this. In general, when you're working in a world of LLMs, you have to have an affordance for the fact that the output is going to be probabilistic, not deterministic. A slightly unnerving aspect of using any AI product is you'll ask the exact same question five times and get five slightly different answers every time, right? When you're designing the interface, you need to allow for that and anticipate that that's going to be part of the flow. But a really important part of how we ... we try not to give very long answers and instead we try to anticipate what the next question might be. So one of the most popular parts of Perplexity are the related follow-up questions, and that's where we're really trying to properly anticipate the user's intention and nudge them to give us the information that we truly need to give them that final answer that could be actionable.
And again, sometimes people want everything and that's when we have the deep research mode where you just basically get a full report. But most people want that sort of iterative [experience], they use the follow-up questions, they kind of prompt it themselves and you go four or five follow-ups deep to get to that point where you have confidence that you've arrived at the right place. Now, why we're so excited about Comet is we now have a much richer set of sources to pull from, which is not just what's in our public index and maybe your past chats. We get to see your browser history, we get to pull in information from logged-in sites, so we can see your Amazon transaction history. So all of that, everything that lives inside of your tabs, can be used to give you a more personalized, relevant answer. And personalization is a double-edged sword in that ... Google's been around for many decades and they still — outside of location — they pretty much never personalize the ranking of the links you get. And that's because the balance of personalization is with the trustworthiness and accuracy of the answer, and people don't want to feel they're being manipulated. And so trustworthiness ends up, I think, being kind of the most scarce asset and resource as AI has become more powerful, and we think a lot about that.
MIKE DUBOE
Yes, I want to spend more time on the browser in a sec. Maybe before that ... There's different flavors of conversion that happen on Perplexity, right? Obviously, a lot of the top-of-funnel discovery is happening on these engines right now, and it seems that you're trying to play a role deeper and deeper and deeper in the funnel all the way through to end transaction. What ingredients need to fall into place before a user is willing to finish the purchase right where they ask the question? How do you guys think about building the right rails to ensure that that happens there?
DMITRY SHEVELENKO
I think it's important to clarify that our goal is not to own the end-to-end experience. The metric we're optimizing for is not full end-to-end commerce experiences inside of Perplexity. The metric we're optimizing for is more important decisions. You begin your research journey on Perplexity. To me, the happy path ... The metric I look for is someone who's been exposed to the end-to-end commerce experience, are they more likely to ask more shopping-related queries in Perplexity in the next two months versus a user that was not exposed to that? So that's the A/B test, if you will, that shows us if we're on the right track because I think there's ... People are different. People are weird when it comes to their shopping.
A lot of people just don't ... It's kind of ... I often tell the story of me versus my wife. I am a very simple shopper of like, okay, I have a problem. I want to find the best solution to it and I want to get it as fast as possible and I don't want to ... I want to move it off my list. My wife, on the other hand, she loves the process of going down the rabbit hole of every possible variant, checking every site, double-checking. For her, it's an enjoyable process. And I think there's all kinds of people in between that spectrum, so I don't think we're trying to pigeonhole people down a specific path. We just want them to start that initial framing of like, okay, what are the product's considerations that would be relevant to this purchase? We want to start that journey.
I don't know if you guys saw our release of Pro Perks. So if you're — Perplexity has a Pro subscription, it's $20 a month, $200 a year — and if you're a Pro subscriber, you now get exclusive offers and discounts that come with that. I think of it a lot akin to a Costco for the internet. And what we're able to do is, if you are a free user of Perplexity and, for example, say you're researching hotels, we now have a native hotel booking option. And you'll see that before you're about to book. You see, "Hey, if you upgrade to Pro, you'll save 10% on the booking." Most hotel bookings are more than $200, so just that one opportunity alone is a reason to upgrade to Perplexity Pro, right?
Where we see the agentic world going is that display and performance ads like AdWords will be less relevant and what people will react to — and their personal shopping agent will react to — is, again, personalized offers, discounts, and exclusives, and we want to start building that groundwork.
Effectively, the economic shift is, instead of us capturing the customer acquisition budget that a brand or merchant has, we want to transfer that full value to the user. So we're taking that OTA commission and instead of us keeping it, we're giving it to the user and we're capturing the value through increased subscription acquisition retention for Pro. I think that's ... I say that because I think that will then inform why we build the product the way we do is we're trying to be fundamentally aligned with our users where we don't want a user to ever think we're changing an answer and manipulating them and saying, "Hey, this is better because we're getting some kickback on this." We're always bringing them best-in-class offers and discounts, but we're not ourselves taking a cut of the action.
MIKE DUBOE
So this actually — let's stay in the business model here for a sec — this is actually very illuminating, and I think very well said. You're trying to transfer a lot of the value that might otherwise be realized in ads to the subscription. I think ... And you kind of framed maybe a contrast between end-user value and engagement and monetization to partners on the platform. Is there a world where you could have your cake and eat it too there? One of the things that Rishabh and I have talked about that seems interesting from these new answer engines is you have such interesting semantic context where perhaps there is a way to actually build ads in a way that enhance the user experience and maybe you can have your cake and eat it too. Do you guys think about these types? I mean, I know you have launched ads in a very small capacity. Is that not the path you're going to go down? How do you think about that? Does it need to be a trade-off?
DMITRY SHEVELENKO
I think the thing that you cannot compromise on is user trust. And I think even doing some very innocent experiments with this, you can permanently tarnish the perception that you're juking the stats. Because an answer that you get from an AI is so authoritative and the second people start questioning the biases of that, that go into that authoritative answer, you're going backwards. And so I think there's just so much value to create with the model I described of offers and discounts, exclusives, focus on scaling subscription revenue. Once you're in that position where people have had enough experience, they know that the answers aren't being manipulated, then you can start doing other types of experiments with the right disclosures. I still think also there's going to be a big opportunity for top-of-funnel consideration. The ad product you mentioned was sponsored follow-up questions where we're not touching the answer or anything around the answer, but in that list of follow-up questions, we let a brand insert the question that they want you to ask.
And by the way, the answer to that, it doesn't take you out of Perplexity, it's just that Perplexity answers that question, which the brand also cannot edit or manipulate. So we kind of preserved full integrity and fidelity. An interesting learning is we launched that — and again, this is, I could not have imagined a more user-centric ad product — but we literally asked, "What is the thing we can do that no one can feel we're trying to manipulate them, the answer doesn't change, it's just a sponsor follow-up question." And we still got completely blown up on Reddit for, "Oh, Perplexity is selling out, and ads, and game over." So I think that was ... While the criticism was ungrounded, unfair, I think it informs that the average consumer, they're not in the weeds on these things. They see ads, they assume they're being manipulated. I think it's actually a deeper problem. When I use DoorDash and I type in "best Thai food," I have zero confidence that the first result is actually going to be what the smartest algorithm at DoorDash thinks is the best Thai food. It is probably some merchant that is paying a higher take rate. I'm being manipulated everywhere I'm buying things, and I think we wanted to do everything we can to make Perplexity feel like a safe space.
MIKE DUBOE
As Perplexity inevitably gains the massive consumer scale and attention share, obviously it becomes ... Marketers will want and need to find better ways to actually capture value from the channel end-to-end. How does your role with them evolve? I think packaging some kind of perks into the subscription makes sense. It's hard to scale that infinitely. Obviously, there's an affiliate-like structure perhaps as you get deeper into commerce, but if you're not going to go in ads, how can marketers actually capture more value on the platform? How do you guys intend to work with brands and marketers?
DMITRY SHEVELENKO
I don't know if we need to work with everyone. I think part of how we stand out and be differentiated is we really ... The way we beat Google and other legacy incumbents is not by trying to rebuild a margin structure that is as high as Google's is, right? That's not how you're going to win. It's like, ultimately we need to create consumer surplus and capture consumers just ... The way we've grown is that people love using Perplexity and we just need to stay true to that. It's an unpopular thing to say, but I often get asked, "Oh, what's the SEO equivalent in AI?" People have been throwing around AEO, trying to get the band back together again and be like, "Oh, that was a good run where we felt we had this set of tactics to get free traffic."
I don't know, maybe I'm just too simple-minded, but I don't see a world in which you can have it both ways. If AEO is a thing, it means Perplexity's answers aren't trustworthy because it means somebody can go in and artificially manipulate the answer. And if our answers can be improved, we should be proactively doing that ourselves by improving our index, our ranking, right? That's on us. We don't need anyone to assist in that task because obviously they don't share the same objective as we do. The constructive answer I give people, though — which I think can sound glib, but I think is actually deep — is build the best product, right? Because ultimately, what we look at is what are people ... What are trusted reviewers saying about this brand or company or this merchant. That is the input, right? It's actual people, what their experience is with the product. That's what drives the ranking and the answers. And so I tell brands, I think earned media and high-quality PR and customer engagement where the customers are so diehard that they want to tell the world about how great your products are, that's going to matter. That's going to be the base of an AEO strategy and not kind of keyword hijinks.
MIKE DUBOE
Yeah, I guess it is helpful for brands to know, though, which sources you ... I think there's an inefficiency still where, as we all know, great products do not necessarily get great distribution organically. And so as marketers in today's era are responsible for helping great products get great distribution, it is probably helpful to know which sources get better rankings from ... are weighted more heavily from Perplexity and the others. I guess that's a sliver of what you might call the AEO practice, but are you disagreeing that that is a valuable area for marketers to be spending time on?
DMITRY SHEVELENKO
It is not an agree/disagree. I don't see a world in which us being transparent about how our system works not leading to gamification that then erodes the thing that we're optimizing. I mean, it is kind of similar to how an AI, the most useful evals are the ones that are private and unpublished. That's why we have our own benchmarks for when new foundation models come out. And the only thing that makes them useful is they've never been publicly disclosed, and so nobody can overfit against them. And so we are ... That's kind of invaluable in our ability to independently assess what's going to be good or not good for users. And so I think that the transparency on the algorithm is ... I don't see a way in which opening it up doesn't lead to gamification.
RISHABH JAIN
That makes a bunch of sense. I am curious when you think about now that Comet has been released and people are not only going to get answers and follow-up questions in the Perplexity app, but they're now going to have it assisted through the browser, how do you think about the ways in which you're able to provide the best answer that has a browser-style interface, especially in the context of shopping, relative to the directed Perplexity chat-style interface? An illustrative way for me to describe this would be, "Hey, when I book a flight ticket, I innately have more trust when I book it on united.com than any other third-party site." And so maybe the browser interface has an advantage in that context because I can actually show you, "Hey, I am pulling up the brand's website and the transaction will happen there if you want to transact," and therefore it is a higher trust mechanism of information display. How do you think about how these two interfaces play with each other? Where will certain types of consumer engagements have more value in your guys' north star of the best possible answer for the consumer?
DMITRY SHEVELENKO
The new behavior that I expect — and we're already seeing the data — that's going to be really popular with Comet is the sidebar assistant. So just the ability to, when you're on a certain website, just be able to ask questions with the context of that site loaded up. And I think particularly in a commerce scenario of ... And again, merchants don't necessarily, they won't love these behaviors, but they're super great for users, like, "Hey, is this cheaper somewhere else?" Or, "What do the other review sites say about this product?" Those are the things that are incredibly ... just to have it right there, you don't have to copy-paste, you don't have to ...
That's the behavior that I think is going to be really popular and kind of a new modality. I actually think using all the personal context, your browser history, all that, I don't think we even want people to have to think too deeply about it. It's like you just get a better answer. It should just feel like a more useful answer. Obviously it's transparent, all the steps that were taken and the sources that were looked at. The agentic queries where you're going and saying, "Hey, go buy this for me," and because we're already logged in, we can take those actions or, "Go add these guys on LinkedIn." That will ... I use that stuff, but I think it is going to take us some time to educate the mainstream browser user to expect that they can take these actions. And so I think that will take a little bit more time.
MIKE DUBOE
Maybe let's make this more tangible. So if I am browsing for best running shoes for flat feet in a Comet scenario versus outside of Comet, how might that experience look different?
DMITRY SHEVELENKO
Yeah, so I'd say the primary one would be we would be able to ... Say you're logged in to X on that, in that browser instance, we'd be able to see what running influencers you follow. And so we'd be able to pull in their recommendations, we'd be able to see, have you gone to a specific brand website recently and taken a lot of actions there? And that can be a signal that you have a preference for that brand, right? We talked about your Amazon order history. So you just get a lot more context loaded into the prompt that is useful in nudging the answer to a way that's more constructive. And then you obviously, you now can direct a transaction. And again, this is why we never were with the seamless checkout, we're agnostic as to whether the checkout happens on our rails or merchant's rails. We just want it to be easy and seamless for the user. And so if we know that you're logged into a certain website and you already have payment credentials, that can be part of the follow-up prompt of like, "Hey, do you want to just go and buy it there? We can do that for you."
MIKE DUBOE
Yeah, it's interesting. It raises a second-order question. If the browser itself starts interpreting intent and curating results, how might that reshape the signals that marketers rely on today? Maybe this is a thing for Rishabh as well here, I'm just thinking of the second-order effects if you continue to play a deeper role in these funnels.
RISHABH JAIN
Yeah, what is it that you would recommend to a merchant in that context when Comet as an agentic browser is taking various actions across various pages and various sites, and Macy's comes to your team and is like, "Hey guys, what's the best way for us to think about how we show up and for us to think about all of the signals from the Comet agent coming to our site?" What are you recommending to them?
DMITRY SHEVELENKO
The fail case is where merchants approach this with a fear paradigm of like, "Oh my gosh, all this incredible infra and all these funnels and flows, it's not going to be as effective or relevant in this new paradigm." It's almost like a wishing this isn't going to happen. I think that's a fail case because I think ultimately the consumer wins, meaning wherever it creates more value for consumers, that's going to be the interface that they're going to go with. A colleague of mine, he talks about the era we're in now is like Browser War III. So we kind of had Browser War I was Internet Explorer versus Netscape Navigator. Number two was Internet Explorer versus Chrome. And now we have the AI browsers. And I think it's ... You want to be the first to really lean in and invite the behavior, right? If this is going to be the thing that people shift to and their first ... a hundred really good e-commerce experiences are with your website, that's going to create a lot of stickiness down the road.
And I think where we see a interesting surface area is in the sidebar assistant, we do want to let domain owners be able to suggest what the queries should be when you're on their website, potentially let them trigger the sidebar assistant with a preloaded answer when it's useful to the user and bring in that independent context as an overlay on their web experience.
And yeah, it's kind of ... I think the resistance most merchants will have is ones that rely heavily on some type of advertising as part of their business model because they don't want agent eyeballs, they want human eyeballs for those ads. And then I think the fear that you're not going to be able to — because you're not controlling every part of the checkout — you're not going to be able to build up people's carts and get them to add extra stuff. So the average order value might go down. The right way to approach it is to, in the second bucket, focus on just winning the space, and that will ... There'll be new ways to — with proactive personalization — to actually get even greater average order value where the AI is nudging you to — you don't even need to search for it — it's nudging you to make that purchase. And for folks that are wedded to display ads, I think that's gonna be ... I think that's just the agentic internet is going to have less of that, and that's ... I don't think there's a elegant solution, at least not one that I see.
RISHABH JAIN
I guess one thing I get most curious about in this world where you think about, hey, in the old world you had signals about, like, you click this ad, you click this email, and that's the old world of signals. How would you recommend, hey, there is an agent who most consumers will use when interacting with your website. Here's context information you could make available to the agent that it can decide whether or not it retrieves on behalf of the consumer, but you now have a new surface, which is an agent who can consume way more information than a human. Is that something that you are asking or recommending? I guess one side of a marketer's job is to take in signals from the outside of like, "Hey, here's how the consumer's acting." And then another side of it is, "How do I present information to be most digestible to my customer?" And the customer in this case is an agent that is synthesizing the information on behalf of the end human. So does that side of the equation change potentially?
DMITRY SHEVELENKO
Well, that's where I come back to the agents are ... AI agents are not emotional creatures. That's where it's like, the personalized offer discount and exclusive, that will be the language that ... that will be the agents' love language because that is something that can be computed, it can be something calculated. And so I think that shifting to that model is going to be, I think, the big paradigm shift. I think the other piece of this is foundation models are really, they're really good when you just dump a lot in the context and they will figure it out. And where I think folks probably burn a lot of unnecessary energy and actually lead to worse results is somehow overthinking how you feed it into the context window or what ... Just put it all in there and it will reason through what's relevant, what's not, to the initial user prompt. At least, we've orchestrated Perplexity to work that way. And I think if you're basically trying to over-engineer how you're trying to nudge the answer, you're actually going to make a worse answer.
MIKE DUBOE
I want to tie this and go back to your earlier point on building a great product. Perplexity, there's a bunch of different signals that you look at to determine whether a product is great and thus make the recommendation in a personalized way to the consumer. There's several commercial claims that brands might make about their product, and I think in a world where brands actually have less control, perhaps there's an increasing value in the role of a certification layer or more verifiable sources that kind of plug into Perplexity. I'm just trying to understand, does there need to be almost a nutrition label of sorts for marketing? Do you think product claims get more standardized? Do you guys have a point of view on that?
DMITRY SHEVELENKO
I think it fits into a more generalized framework of just domain authority. I think if you create ... I don't think it's fundamentally different than a new high-quality publisher that does product reviews. I think it's a form of that, but I don't think we're gonna ... We live in a messy, uncertain world and I don't think there's like ... There's no one-shot solution to solving that. It's aggregating a lot of diverse high-quality sources and synthesizing through it.
MIKE DUBOE
It's always been a marketer's dream to be able to do personalized offers, true one-to-one, and understand the real elasticity on a user level and doing promotions that way. When you have to do blanket discounts that could be brand dilutive, it's more of a complicated topic, but I also know there's some, I believe there's legal issues that prohibit you from doing price discrimination in a certain way. If I go a couple steps forward on this direction you're talking about, what do you guys anticipate running into there? And maybe I'm extrapolating too far on doing one-to-one offers and discounts, but what are some of the conversations you're having around that?
DMITRY SHEVELENKO
I think of it almost like a ... The anthropomorphized version of it is like you're at a bazaar and you're negotiating with a merchant. And except instead of it's you and the merchant, it's your agent that's an AI and the merchant's agent that's an AI. It may get communicated as if it's a personalized offer, but what it really is is a bid-ask system between ... And I think where it gets dynamic, and I think it kind of is on a different plane than some of the restrictions you mentioned, is oftentimes it's not a discount for one product, but it's like, "Hey, if you spend $200 with me over the next three months, this is what I can get you." Or if you buy these three things, one of which I'm trying to offload my inventory of. So I think when it becomes really dynamic ... And we're talking, this is not a two year thing, this is maybe a decade out. And by the way, there's going to be all kinds of categories of consumer purchases where we don't want ... We want to be in control, we don't want the AI deciding for us because it's fun to ... I mean, again, I think back to my wife, she enjoys ... There are certain purchases we really like to make. It is something that brings us happiness, and certainly we'll use AI to be more informed, but you're not going to outsource the execution of the transaction to an AI.
RISHABH JAIN
Dmitry, one thing I can't help but feel when you say this — and I'm not sure how you would think about this is — I feel like the consumer-side capabilities are going to go faster than the merchant-side capabilities. I can imagine someone like you or I using Comet capabilities at a level that is far surpassed where the merchant could get to today. The simplified version of this is like, "Hey, can you show me where this product is cheapest?" and it can look for all the retailers where that product is sold and quickly tell me which retailer is the best-priced option. The merchant's infrastructure I know does not allow for that today. The merchant does not know all of the different prices that its product is showing up at in all of the different retail partnerships that it has at any point in time. And so there's this very interesting dynamic that starts to happen, and Perplexity is solving for the consumer. How do you think about ... Yeah, I guess, do you just see this as, hey, these are great opportunities for other startups to come in and start to help the other sides of the market that get created when companies like Perplexity are creating so much advancement on the consumer side in this equation? Help me think about how you think about that.
DMITRY SHEVELENKO
Listen, merchants have really big IT teams. They're fully capable of ... In some ways ... They're the ones selling, so they have even more reason to get on top of this. And I think necessity is the mother of invention. When they see their older tactics not working because the consumer is superpowered, I think they'll level up. You may have a six-month dislocation, but I think this will become ... I think the best merchants will rise to the occasion, just like the best brick-and-mortars handled the transition to online. I think this is ... I don't think this is ... The best merchants always kind of handle these well.
MIKE DUBOE
Maybe in the last few minutes here I'll shift gears and talk to you as a growth person. You're responsible for growing Perplexity. You guys have been on a meteoric rise, of course, but there's a new paradigm for growth for these types of products and this world. Maybe high level, how does PLG look different in today's world? And maybe share some color on unique insights you've had in growing Perplexity.
DMITRY SHEVELENKO
Our growth has really, on the consumer side, been a function of people loving the product and telling other people about it. And the reality is in consumer internet outside of social products, there have not been that many success stories. It's really hard. I mean, I dunno if the gray hairs come through on the camera, but this is not an easy problem. And you have to really get creative. Obviously we're in a very competitive space with a lot of moving pieces, and effectively all our competitors are either trillion-dollar companies or bankrolled by trillion-dollar companies. And so it's not easy, but user love, kind of, that's...
Without that, I don't think you can even be relevant in the space. So I think that's the foundation. I think on top of that, what we really focus on is execution speed, because the foundation ... This is the analogy I use. I spent time at Uber during some of the hypergrowth years there. And the AI space is kind of like you have all the competitiveness and intensity of the rideshare wars, except it's like GPS, the underlying technology, is changing every two weeks. So you just have new foundational capabilities that come online. And it is a ... Part of our brand is the user promise that whatever is the new thing in AI, the parts of it that can be useful to you will somehow be incorporated into the core Perplexity products in real time. And then that generates the coverage around new features, changes. It keeps you fresh, keeps you in the news.
We do a lot of partnerships with telcos and others, and there's as much benefit from the partnership itself as there is from the conversation around the partnership. And so I think obviously you have Cluely running the most extreme version of this playbook of just being ... generating attention by being controversial. That's not our brand. For us, we want to generate attention by constantly shipping product. That is ... If we were trying to reverse-engineer how we've gotten to where we are, I think the user love and the execution speed have been pivotal. And I think we now have some really interesting plays with enterprise where pretty much every company we would want to sell into, there's already probably five people at that company — minimum — that love Perplexity on the consumer side and probably use the consumer version for work too and just don't tell their IT team. That, I think, will be ... We have, I think, 6,000 enterprises paying for Enterprise Pro. We have an enterprise team of three people. And so that's kind of ... You're only able to do that because you have that motion of inbound coming from your users.
MIKE DUBOE
And maybe my last question, which is in the vein of advice for other founders, you seem to be ... How do you filter for velocity? I mean, you are in a very unique position where many companies, once they hit a certain scale, velocity starts to slow and just kind of laws of gravity set in and all of that. It seems that your team is only kind of picking up speed as you scale. How do you ensure that you have the right people on? How do you filter for that upfront and maybe how do you best enable them as they're getting to scale?
DMITRY SHEVELENKO
I think it's certainly being radically transparent when you're hiring people that this is how we roll, and ... This will sound maybe too serious, but if someone mentions work-life balance in an interview and asks me about that, I'm like, "Okay, this is not going to work out." Because mostly it reveals that someone hasn't done their homework on Perplexity because we're very also transparent about how we operate. If you haven't picked up on that that's not a theme, then you're probably not ... There's bigger issues at play. So I think the radical transparency, set the right expectations, and that way you're a magnet to the people that want that experience and you are kind of deflecting the ones that don't.
I'll give you a tactical example. I don't ... I have many direct reports, but I don't have a weekly one-on-one with any of them. I don't do that. And many of them, I meet one-on-one throughout the week, but I don't have a scheduled weekly one-on-one because my greatest fear is that someone's going to wait six days to bring an issue to me, and we're going to have lost six days on solving that issue or executing against it because that's when the weekly one-on-one is and that's when they would've thought they would've had the opportunity to make progress on it. I put incredible agency and trust on my team with the great responsibility that if something's broken, if you're stuck, I am not going to check in on you, but you better come to me. And kind of keeping a high accountability bar on that is ... That's been important for us. And then I think the last thing is being explicit about trade-offs.
The whole point of priorities is that something is not a priority. And so the thing I always ... My commitment to my team is like, "Come to me with what's on your plate and I will always give you the stack rank." I will ... And that's, by the way, that stack rank might change every 48 hours. And that's part of the radical transparency upfront. As I say, if you want a ... Another thing I find myself saying often is I am highly confident, more than anything else, that six months from now I'm going to have, myself, a top-three priority that today I have no idea what it is, and I have more conviction that I'm not going to know one of the top three than knowing what the other two are. And that's, you kind of need to embrace that. I think that's very localized to AI but definitely even more localized to Perplexity. But that's culturally how we execute.
MIKE DUBOE
Dmitry, this has been awesome. We've covered a lot of ground. Thanks so much for going deep on this with us, and yeah, it's been a great episode, so really appreciate it.
RISHABH JAIN
Yeah, thank you. I think the world is ready for a new realm of how they interact with shopping, with marketing. Yeah, only exciting times ahead.
DMITRY SHEVELENKO
Awesome. Thanks guys. Great conversation. Really appreciate it.