Perplexity builds on models from the industry’s biggest AI providers while orchestrating them on behalf of its users. In this episode of AI Decoded, Proskauer partner Oliver Howley and Perplexity general counsel Nathan Barksdale discuss Perplexity’s place in the AI ecosystem, today’s AI valuations, and how Nathan’s team uses Computer for Counsel as a digital associate.
OLIVER HOWLEY: Let’s zoom out and talk about Perplexity’s place in the AI ecosystem. You build on top of models from companies like OpenAI and Anthropic and Google and X — all of the big providers. So you’re a major customer of theirs. But at the same time, some people might say that you’re competing for the same users for the same revenue line. How do you think about that relationship and how do you navigate it?
NATHAN BARKSDALE: I would say we don’t have any competitors, really. I think early on Google was our main competitor. But now that we have broadened from the answer engine to Computer and other things, Google is only a very small subset. But we’re all in this kind of symbiotic relationship together. The better AI gets, the more competitors there are, the better models get — the more models that are released, it makes Perplexity better at the end of the day.
We average — we’ve had no more than 16 models integrated into the platform, upwards of 20-something depending on releases and what models are performing. All of those models, to be able to orchestrate that together on behalf of the user, it makes our product that much better. And that just means that OpenAI, Anthropic, us — all of these can live symbiotically together. We rely on them to get the earliest release of their newest models. And obviously we’re a customer of theirs, and so that helps them as well.
OLIVER HOWLEY: I think what’s interesting is when I’m asked by another partner in the firm for access to the latest model from Anthropic or X or Google. And let’s say we don’t have an enterprise license, I just refer them to Perplexity.
NATHAN BARKSDALE: Right.
OLIVER HOWLEY: I don’t think there’s ever been a time where you have taken more than 24 hours to release the latest model from the latest vendor, and I think that speed, that access, and that ability for you to improve your product on an ongoing basis in line with the wider market is amazing, and amazing for the users.
Now, while we’re at a macro level, I can’t quite resist asking about valuations of AI businesses. You’re worth a lot of money, but every few weeks, another one of the big AI companies or another big application layer developer seems to command an extraordinary valuation. Some people think that the market is pricing a once in a generation, once in a lifetime technology. And others think it’s just valuing stories more than businesses. So where do you come out on that side of the debate? Do you think these businesses are worth the valuations that are put on them?
NATHAN BARKSDALE: I think everyone is undervalued.
OLIVER HOWLEY: Okay.
NATHAN BARKSDALE: It is a once in a lifetime technology that people are going to keep on iterating on. We’re going to look back at the release of GPT two, three years ago and think that was just a small blip on what has occurred in this AI life cycle, and we’re going to be so much further ahead in five and ten years. But what we are and what these valuations represent is “what is the market in the future for AI?” And it’s really everything — every industry, every business will be affected by AI, usually positively from an efficiency standpoint. If we look at what the AI market is right now, I think it’s about $900 billion today. And if we look at what they said crypto was — you know, a blip or a phase — that’s about $2 to 3 trillion. And social media, which gets these insane valuations from social media companies, it’s about $300 billion.
OLIVER HOWLEY: Wow.
NATHAN BARKSDALE: That has sustained for 10, 15 years off of a very small market compared to what AI will eventually hit in the next few years.
OLIVER HOWLEY: Before we get into how you use Computer, how involved were you and your team in actually shaping what it looked like before launch?
NATHAN BARKSDALE: The origin story of Computer is very much similar to the origin story of Computer for Counsel. So with Computer — in December — we were completely using it internally just as a tool of, you know, how can we help be more efficient? How can we access and connect all of our internal systems so people can get information in a different way that’s more digestible? And then when Computer was created and launched, we thought, well, we should do the same thing for Computer for Counsel. And so the product team embedded with our team on “what are new ways, how would you like Computer to be curated, tailored to a lawyer in your day to day work?”
OLIVER HOWLEY: I actually remember when we were discussing — it was either the answer engine very early on in our relationship or Computer where you said, “Wouldn’t it be great if we could have Docusign connected to this?”
NATHAN BARKSDALE: I imagine a world where we’re not removing Docusign — I think Docusign is one of the most important connectors in the legal landscape. But how can I stop the many steps it takes me from getting a document to Docusign’s interface and then signed by the parties? And with AI, I just want to dictate, send this out.
OLIVER HOWLEY: I have horror stories from paralegals in our team spending hours and hours—
NATHAN BARKSDALE: Yeah.
OLIVER HOWLEY: —preparing Docusigns. And now, just like this, Computer fixes it all.
NATHAN BARKSDALE: Exactly. Exactly. I think for in-house teams, having access to legal knowledge and case precedent with companies like Midpage is amazing. I think Lexis and Westlaw have owned that market for the most part and it’s quite costly. And that’s why the primary use of outside counsel is getting access to those resources. So I don’t think it completely replaces Lexis and Westlaw, and I don’t think it replaces outside counsel doing that. But we are more knowledgeable internally because we have access to these systems, and it can help us digest motions better and complaints much faster.
OLIVER HOWLEY: I get great feedback on Computer for Counsel, and actually the search function within Computer for Midpage — which is also a client of Proskauer — and its ability to spot fact patterns in prompts and find comparable cases in relevant precedents. And I don’t think we would be able to do that without the combination of Computer for —
NATHAN BARKSDALE: Right.
OLIVER HOWLEY: — Counsel and these individual connectors that you’ve then curated for the benefit of the entire legal profession.
OLIVER HOWLEY: So we’ve spoken about how Computer and Computer for Counsel was developed. How does that now fit into your day-to-day? I think a lot of people will be surprised to learn that despite the speed at which you ship, the valuation that you have, the market presence that you have, your legal team is still less than 10 people, right?
NATHAN BARKSDALE: Yeah.
OLIVER HOWLEY: And you essentially use Computer as a digital associate, or a very efficient digital associate, to work alongside the team. What does that look like in practice for you?
NATHAN BARKSDALE: Yeah. So we don’t ship anything we don’t use ourselves. And I think that’s important, going back to the ethos of how we hire: we need people that are curious about the future of these products and how they work. Someone that is rooted in the ways of the past 10, 20 years they’ve been practicing and is not willing to engage in new ways and try new things — that really doesn’t work for us, because we need people that are going to find new solutions and basically be a test case for what our business puts out there. For our team, we have a weekly legal meeting where we sync together as a team. Part of that meeting is dedicated to people presenting their new use cases and how they’re using our new products, and we’re shipping so fast that sometimes people are presenting on things that half the team didn’t even know was a product yet, or was out there. It’s allowing all of us to work off each other’s use cases, because I have a dream for Docusign integrated with Perplexity where it makes sending out signatures much easier, but our regulatory team, our commercial team, they all have different use cases on what’s working for them. And when you combine that across the whole team and allow people to piggyback off those ideas, it’s truly magnificent. And the sharing of that information, for example, has really helped us accelerate our growth, even though we’re a size of like 10 people, because we have these powerful agents. Each person is working with 50 sub agents that allows them to do their job more efficiently. And that doesn’t mean we’re slowing down our growth — we’re going to continue to hire. It just means that people have more capacity to do more with potentially less.
OLIVER HOWLEY: It’s amazing. I know our use cases are, well we actually have an internal program at the moment where people feed in useful use cases which are then disseminated amongst the firm so other people can take inspiration from them. But we have ones, everything from tracking BD—
NATHAN BARKSDALE: Yeah.
OLIVER HOWLEY: —emails across accounts and prompting us to follow up, through to, I think the most impactful one from David, one of my partners who you know very well from our scraping practice, he used it to analyze a mural that was going to go up in New York for potential IP infringement risk and IP ownership risk and help him do a risk analysis for the benefit of a client, within 24 hours, and the client came back and said “What are you, an encyclopedia of all trademark and copyrights in the world?” He wouldn’t have been able to do that without Computer.
NATHAN BARKSDALE: Right.