Joe Lonsdale describes Palantir's early years selling to U.S. intelligence agencies, where who introduced you mattered more than the product. The company flew to Washington bi-weekly for demos, iterated based on feedback, and built secure on-premise software before cloud services existed, requiring top-secret clearances and extensive security audits.
Chapters
Early days of Palantir, going into these meetings with the CIA and big government organizations. Were you in these meetings, or did you send— you mentioned—
Yeah, no, it was Stefan and me, and then Karp. And then we both— the 3 of us kind of ran stuff for the first few years on the business side as well. I was head of product, but then I shifted over to head of business. And I replaced myself with someone who's way better than me. But yeah, the first few years, I was doing a lot of these. It's really, really hard dealing with big institutions. There's all sorts of tricks of the trade. This is another reason why you need probably people who've done it before who understand these things. But a big institution, it turns out, one thing that was not intuitive to the 21-year-old version of me is it really matters how you're introduced and who they think you are. So it's one thing I thought, gosh, I have this thing that they need, and I can walk in the door and show them, and shouldn't the world just work where if they need it and it's helpful to them, they're going to listen? And it turns out that's not at all how the world works. It turns out that if I'm introduced by the former head of the CIA in the right way, then it's— who's looked at it, they got to be way more serious than if I'm introduced by the older brother of a friend who used to work— who works there or whatever. So it's like there's all these subtle status things about how you come across. There's all these subtle things about how you present the company, you present yourself. Government people, and I think this is true of most big institutional people, there's a dialectic there where they, one, they really want to be feeling like they're on the cutting edge, but then they also really want to feel like it's safe and like it's proven and like someone else has done this already. So when we ended up closing our first pilots, like, Karp figured out a way to like make the FBI think the CIA was already moving ahead and make the CIA think the FBI is already moving ahead, and they both felt like more safe, and they were jealous the other guy was going first, and they both did it, you know.
Wow.
So it's like, there's all these like— it's really stupid because there's like, there's like the actual substance of the business, which does matter, but then there's also all these subtle things about how humans just institutions work that are very— they're just— people are complicated beings. They're not always logical.
What was the customer discovery process there like? You guys knew that there was a problem with data in these organizations, or as you started talking to them, you started pivoting whatever software you were building? What was the first software you built? How was that?
Yeah, and I wouldn't recommend building companies like this normally. I think with Addepar, it was a lot more logical, where we had early customers using it and iterating with us. With Palantir, you couldn't really get early customers using it, so you had to go just talk to them, and you had to see what their objections would be, see what they wanted, see what their requirements were. What we did, I think there's one of the early years when we started doing this, there was a period of about 12 months where I had 26 round-trip flights on JetBlue to DC. That was before JetBlue had business class, by the way. Back and forth, back and forth, economy. We would just every 2 weeks, every week and a half, we'd go out. And we'd show them the latest we'd done, and we'd get their feedback, and we'd talk about it, and we'd see what they wanted, what they cared about. And then we'd go back, and we'd just iterate and code and build as fast as we could and create something that was based on what they said, and try to make them feel like they had ownership over it, make them feel like they were part of the process, and then impress them with how quickly we were building. And then eventually, after iterating a bunch with all these different people, we got to the point where they didn't really have objections, and they were willing to try it out.
With everything going on right now with controversy in terms of data handling, I'm guessing handling government data was probably something very sensitive back then and now. How'd you guys get—
Yeah, well, I mean, even now, a lot of what Palantir does is not in the cloud. Back then, there was no cloud, right? It was 2004. This was before the idea of the cloud. So we built our own version of what would now be called an internal cloud for them. I created that as an SNI, as a lot of others did at the time. And yeah, you just had to be on-premise. We had to get people who had top-secret compartmentalized security clearance. It's a mess. I mean, dealing with the government is a whole nother level of just obnoxiousness when it comes to these things. So yeah, it was a lot of work. I do think that a lot of things in the world are broken because people are too careful with all the rules around data. If you're a startup and you're trying to work with the Fortune 500, There's oftentimes, it's just like 6 months of just like extra work. There's like these security audits you have to get ready to go through. You have to hire people who know how to deal with the security audit. It's just like, it's a whole set of things that are very difficult to go through if you're going to be working with the really big companies or the government on these things. So it doesn't mean you can't build things in that area. It does mean you better have people who know what they're doing working with you if you're going to go after these types of customers.
How many employees did Palantir grow up to be while you were there?
While I was there, I don't know. I kind of transitioned out slowly after 6 years. I'm still an advisor, so I don't really have a hard point in my mind, because I started building Addepar 5 and a half years in, in '09. But I think it was probably like 300, 400 when I stopped being full-time, and I think it's about 2,200 now. I don't know. Maybe that's secret.
