AI services, so hot right now

0:04Today, we're joined by Jaclyn Nelson, the co-founder and CEO of TribeAI, the AI-native services firm that forward deploys top engineers into the Fortune 500 to build AI products. And she's also my partner at Coalition. We're going to nerd out on all things enterprise AI, like why the bottleneck to AI transformation isn't selling, but adoption. And how early are we? Great Chat is brought to you by Mercury, radically different banking loved by over 300 ,000 entrepreneurs. We use Mercury to run this podcast and love recommending it to other founders and fund managers. It's free to get started. Try it for yourself.

0:41Visit mercury.com to learn more and apply online in minutes. Hey, guys. Hi. Jackie, welcome to the group chat. I am thrilled to be here. Let's chat it up. My group chats have like cross-pollinated because, of course, you're in my coalition chat and now you're in my great chat. So welcome. Welcome to the pod. We're thrilled to have you here. And I feel like a lot of the stuff that's happened in the last week is like very in your world. Like, I mean, what isn't? AI news these days. It's a weird moment. I mean, SpaceX S1 maybe being like the thing that everyone has been waiting for and salivating over.

1:21did you do a deep dive or has your has your Claude done a deep dive I did ask Claude and Claude was like oh uh SpaceX is going public I don't think that's true that would be a big moment let me check and I was like wait did we just go back a year and then it like you know it sort of did its search in real time and it was like well many of these sources are speculative like it took a lot of cycles to get to having any like ability to summarize sort of the key points the main reaction. So I think like upgrading, the more I pay Claude, the less value I seem to get right now. So I'm not sure what's going on.

1:58Something I've been noticing is like, if I have like an experience like that with Claude, I'll just jump to ChatGPT. Or then if somebody like, if I don't get like a good response or have like a sort of a bad interaction, I'll just jump to a different one and I just kind of rotate all day. And I'm just wondering like what that does for the long It's going to be like Uber and Lyft with comparison pricing. Totally. It becomes a little undifferentiated. Although, I mean, Claude has all of my great chat memory. And so I like to sort of prompt it in there. Anyways, well, maybe SpaceX is giving Anthropic the shitty compute and it's showing up in the product.

2:41I don't know if that's a thing. But okay, maybe like without Claude's help and with more manual research, I mean, it's an anticipated valuation of$1.75 trillion, which is insane. Starlink is like the cash engine and accounts for 61 % of total revenue. And then Helen, I feel like you were most focused on the TAM. Well, that slide was just so funny, right? It was like, this is the largest TAM of all time. And you're like, okay. I saw a couple other folks have this reaction, but it's like, the thing that I was reminded of was, I think the last time Elon put together some sort of like prospectus or forecast, it was after he acquired Twitter.

3:25And it was like, here's what I'm projecting revenue to look like in the future in something like four years from now or something. It wasn't that much, that too far in the future. and it was like two things. One, I think there was an 8x revenue growth, but then more than half of that would have been non-ads. You know, let's fast forward because it has been four years, I think since he published something like that. And it's like total revenue is down something like 75 % or something, like something pretty significant. And it's still majority ads. So I think people are just like, well, you know, for him to follow through on what he says is kind of not, just using that as a moment.

4:02And so I thought that was kind of interesting. Just last time we saw a TAM slide, you know. Early stage investors love to sort of gently make fun of TAM slide because every TAM slide is like, like the numbers are ridiculous and don't necessarily, like it's like you have to, as a founder, you have to have a TAM slide. But the TAM slide is often ridiculous. And like this is validating, I think, for all of those early venture-backed founders that even when you get to the scale, your TAM section is going to be ridiculous. It doesn't matter. Whether you're pre-product, pre-revenue, or you're going public with an implied$1.75 trillion valuation, the TAM section is always – maybe the point is for it to be absurd.

4:49I think it's something like $28.5 trillion. It's the size of the US economy, which is amazing. You know, on the other hand, like data centers in space. Yeah, it's pretty crazy. It's wild and ambitious and all the things that it should be. But like, it's not insane. I think that it's just we're having a hard time sort of like grasping the possibility of that. But I think that like it's consistent with what I think SpaceX is trying to promise. And maybe it's possible. It's also where we are in the market. Like we have OpenAI and Anthropic looking to go public really soon. Everything's going to be in the trillions.

5:32Billions are no longer interesting. Elon's never one to want to be one-upped. So I just kind of think there's a lot of jockeying. On that point, Jackie, I do think the best move Elon has made is to be first. Because how much capital is out there, and we've talked a lot about how this race is as much about like how much capital can you accumulate as anything else. And SpaceX being first to market is going to be first to tap into those capital pools. And so Elon is outmaneuvering the other players in that respect. And we'll see. And then the hilarious thing is like Anthropic is part of what makes this story so compelling, right?

6:13They signed up to spend$1.25 billion per month and rent and compute from Colossus. And so that's like$45 billion over the contract period, which is through like spring 2029. So they're actively making SpaceX's IPO sexier, which I just love it. Everything's so incestuous. But then OpenAI filed right away, or like appears to be filing. And so I think that's the like now who's going to be next is the big question. That feels like the more cutthroat race to me, like Anthropic versus Open AI. Like SpaceX has a lot of other stuff in it, whereas like Anthropic and OpenAI are much more sort of peers, competitors.

6:59One's a spite company of the other. So that's the real race. What is the, I've heard some responses to this, but asking you guys, like what was the logic of sort of sort of ducking all of the companies inside SpaceX? Like SpaceX itself, okay, yes, incredible. And then it's like some of the others, I mean, XAI, right, still such a long-term bet, but like a hotly contested race with OpenAI Anthropic. And then, yeah, X, which still feels early days in what its ultimate form will be. And so, like, what was his party line and what was the party line on like kind of like why all those things were grouped together?

7:43My hunch back to the like, this is all about capital and fundraising is people will always give Elon money because he makes good on their investments. And so I think a lot of the turduckening, which I don't know if you coined that, that's fantastic, is about like, oh, like people who gave him money for XAI or money for X. He's jamming these things in so that they can have a return. And you can always tell a story. Now we have a story around data centers in space. You can always find the narrative. But I do think a lot of it is like Elon is generally a good bet. And even when the assets don't look good, like Twitter or X, to your point earlier, those investors will probably do well in the SpaceX IPO.

8:31And so I think he really plays the long game when it comes to fundraising. That's my best. This is the best shot at him looking like a genius. I think there's a narrative, true, not true, truth doesn't matter anymore, that this was the plan all along, that these assets actually together are what creates this massive vision and value, and that independently, they're far less interesting. But if you actually take the social graph, you know, with the AI capabilities, with the compute, like that's where you get, you know, the insane power and ambition. Again, truth being irrelevant and like, you know, did narrative come after or before?

9:14But I think there's like a beautiful story here of like a really visionary person when like no one else believed or it made sense before. Well, Jacqueline, one thing about me when I work with founders, I always say like, you have to make it seem inevitable that, you know, this was the plan the whole time. And so, you know, you validate, validating that point. But no, yeah, I think, I think that that's, that's a good point. And, you know, you can't be a techno optimist without frankly rooting for, for everything that is wrapped in this traducan. So here we go. Let's go. I think you could, I think there are, I think there are nuanced ways to be a techno optimist.

9:52Elon has always had a vision for like an everything app, right? So even back in his like PayPal days or pre-PayPal, pre-that merger. He's wanted to have one sort of app to rule them all. And now he's doing it more at the infralayer. And so, Jackie, to your point, it does actually feel like part of a grandmaster plan. And the pieces have just come together in ways maybe none of us could have imagined a decade ago, certainly two decades ago. I'm surprised he hasn't jammed Tesla button yet. So maybe that will be like a post. Sure. He tried. He still has the famous quote. We were promised flying cars and we got 140 characters.

10:33And my like secret theory has been like Elon is just building a company or an empire based on that tweet. Like we're sending Teslas to space and we have the hundred, like you can have it all. It's not an either or. So. See, genius. Genius. Against this backdrop. Google IO felt a little, like that? I don't know. I am not deep enough in the Google universe to assess personally, but it felt, I don't know, it felt very measured. It felt a little bit more incremental. Part of this was I listened to the hard fork interview with Sundar and Sundar is just like, he's the opposite of Elon, right? He's so thoughtful and calm and not hyperbolic in any way and acknowledges like where they're behind or like, and so I, I, maybe it's just that I hadn't recalibrated to like a CEO who's telling it like it is versus one who's painting an insane picture.

11:36Yeah. Two, two nuance. But, um, but yeah, in the midst of all of this AI craziness, Google IO kind of felt like business as usual, but not like a giant leap forward in the way that Gemini 3 was. How much of this do you think is kind of like the iPhone effect where like we've reached diminishing returns on like what these models can do and iPhone effect meaning like every time like Apple does a keynote and talks about stuff in the new iPhone, I'm like, I'm good. I was good with that camera four generations ago. Like I don't need to see like the hairs inside my kid's nostrils. Like I'm so good. And so now I think we're like, holy crap, AI is incredible, but it's like everything's going to feel really incremental to what you said.

12:23I don't buy that. No? No. I think there are going to be step function improvements from here. I was expecting a big week and I was like, I was ready. And this is from someone who I'm pretty against conference driven development. I think we get the forcing function of releases, but I think we actually just get way worse products as an industry. I really think it works against all of us, but it exists, so there we are. But I did think that this was kind of like a moment where Google could have really created a lot of noise and really assumed a position. And part of my theory, and I will admit, I recently launched a partnership with Google.

13:04I am ex-Google, so I am closely monitoring things and certainly not unbiased here. But if you really believe that new center of gravity is going to happen with AI integrating directly into Google Docs and email and sort of, you know, PowerPoint and all kinds of Google slides, whatever, it makes sense to me that Google would step up and say, like, you're not eating my lunch. Like, we're going to do this too, and we're going to do it better. And I just don't think they leaned into that at all. I think it wasn't actually the lack of, like, step function improvements for me. I think it was like the lack of clarity and the lack of like, here's what you do with us and here's how to use our tools.

13:45It almost feels like we went back to, you know, the GPT versions of models where there were all of these, you know, fragmented models and different names and no one could keep track of what to use for what. Like, I think they just kind of like lost. They had an opportunity and they didn't fully capture it. So like a story problem more than anything. And your naming point resonates, but also reminds me of like my one sort of like fun takeaway from the IO announcements was that they're sort of like sunsetting the Gemini brand, right? And like, I sort of felt like they were laying the groundwork for Google to sort of like take this 2.0 meaning, which is like the future of search being inside of a chat versus, you know, the browser as it is today.

14:32And I have this fun, you know, sort of long Google theory on how they could win the AI race, which is like helping advertisers sort of make that transition so that they, you know, continue to dominate basically search. But in order to do that, as a public company, it's going to be very difficult because there's going to be a transition period. And so like I always have this like weird hot take that like they should take themselves private. And the whole alphabet setup was like designed for this exact scenario in mind. So like Wall Street doesn't freak out when they transition and like there's a little bit of a dip.

15:06So I saw like the sort of maybe the, you know, quiet sort of sunsetting or like maybe the emphasis on Gemini is like, oh, maybe this could happen. But that was kind of like my takeaway. So for that to be one of the takeaways, Ashley, to your point, not like blockbuster announcements. I got two things from the headlines. One was Google continues to sort of like chip away at its cash cow, which is like traditional search. And I think Google I.O. sort of provided an opportunity for a lot of publishers and content creators to come out of the woodwork and talk about all of the diminishing returns they're seeing from search.

15:42And so it just sort of fed that narrative. And then also their sort of biggest story on the model side seemed to be like, we're faster and cheaper. We're that alternative. And actually to the iPhone point in a different way, it's like, it's sort of like the Android story. Like not as sexy, but actually like much better for mass adoption, much more accessible to people. And if they had leaned into that, if they had been like, listen, sure, they're all going to, they're going to be these like frontier labs over here, you know, pushing like, you know, a quad launching mythos and moving everything forward.

16:15But like we're going to be AI for the masses unabashedly. And we're going to make that affordable, you know, not just for individuals, but also for businesses, small businesses, large businesses, whatever. And over time, this all becomes like a little bit commoditized. And we're actually the ones that can play a volume game like that could have been an interesting story for them. But maybe that's a hard story to tell if you want to get the top talent or retain the top talent in this like very sexy were addicted to step function improvements time. I think you can also argue that they actually leaned into, to your point, a consumer story, not an enterprise story.

16:50And so what I was expecting them to do is to lean in on enterprise because that's so clearly where OpenAI and Anthropic are like encroaching on their biz. But I think actually it makes so much sense that their biggest step function improvement was in a video generation model, which is very much geared at consumers and geared at kind of creatives. And I think, you know, then you're betting that people keep building on the Google suite and ecosystem that then you can sort of tackle the enterprise use cases in a sort of secondary way. And so I think, Ash, that kind of goes to your point on like, hey, we're going to go for like the mass strategy.

17:30And I actually think that's really where OpenAI started. But since Anthropic moved the war to Enterprise, to B2B, they then shifted focus. And then maybe Google says like, hey, I see an opening and I'm going to take it and like go gain a ton of share, be the Kleenex, again, be the Google Helen, right? Google search and then kind of come back to Enterprise because we have so much surface area. I love that. I just want that narrative. I want something like Chris that makes it all sound strategic versus, yeah, versus sort of like, Like, here's a grab bag of things that we've collected. And that almost goes back to the conference-driven development, which is, hey, you're throwing a lot, like, what actually happens inside these companies?

18:14You're kind of throwing a lot of spaghetti at the wall and then seeing what makes the cut for the announcements. And I think what makes the cut are like, hey, what actually, like, put up the best results? What's going to tell the best story? And I think in some cases, like, and I'm not talking about Google, I'm talking about any large tech companies with conference-driven development, you don't really know until kind of the end. And so I actually think you almost like you miss the chance for some of that narrative if you were actually doing that bottoms-up development. Right. Like you're like stitching it together towards the home stretch.

18:46Jackie, I kind of love that as a point because I also, I think I'm thinking back to my like enterprise software days and we used our annual conference BoxWorks as like a forcing function to get everyone excited to be ready to like ship, whether it was something in GA or the beta of something, you have your customer's attention. Once a year, you have a stage and you should take advantage of it. And I think the news cycles and the development cycles in this moment are so much faster that maybe the old conference model makes far less sense for this kind of approach to product launches. Because you can get people's, if you're a Google or an Anthropic or an OpenAI, you can have anyone's attention on any Tuesday or any Thursday and you can make a moment and like give people things that like they can play with.

19:38And like, it's actually like it used to be you needed to package a lot of stuff together to create a story. And now actually it's a little bit overwhelming and dilutes the story. And so, yeah, maybe the whole sort of conference model, unless maybe it still makes sense for an Apple where you have some like big hardware moments or something like that. But for a Google, And maybe for an OpenAI, like their next Dev Day, like I remember, Helen, were you with the last one or maybe we were just following along? It was like a barrage of things. Developer Day to me, and I've said this a couple times, it almost felt like a fire hose intentionally because that was sort of like peak VC brain of like, well, I don't know if I want to invest in any of these companies because we don't know what OpenAI has been up to, right?

20:20And so they basically took the stage and said like, hi, everyone, don't bother because we're going to do all of these. And it was very smart. I thought it was the right thing. But I think that's kind of. Yeah, come join us. Come join us instead. You know, best talent. Don't start a company. Don't bother. Just like come work here. Right. And it was great. You have a room full of developers. Like, yeah, that probably really worked. I love it. Different seasons of AI and they are passing so quickly. So I think like a recurring theme on this podcast is like, where are we actually in AI adoption? You see these insane stories around valuations and fundraising and ARR that are just like, you know, record-breaking, mind-breaking.

21:04But the adoption side of things is like a little bit more I'm a question mark. And institutional lag is something we talk about a lot. Like, you know, go on Twitter and people have their crazy open-claw setups and maybe are experiencing LLM psychosis. But like, what's it like in the rest of the world? And so I think we can talk about that broadly. Like where are we in AI adoption, especially in the enterprise world where you play? But then also you run Tribe, which is an AI native services business. And services are kind of hot right now. Like in the past couple of weeks, like OpenAI and Anthropic have both launched deployment companies.

21:42And it's becoming clear that like services might be the real bottleneck for this AI transformation, not the technology itself, not can they get people and enterprises to buy it. And so we thought it would be fun to bring you here to chat about this and give us the lay of the land. Maybe just to start with the headlines that are kind of fresh, like walk us through what are these deployment companies and why did OpenAI and Anthropic feel that they needed to launch them? Yeah, it's such a fascinating moment. And it's really these two companies, OpenAI and Anthropic, admitting that their technology is amazing, but not enough to solve the bottlenecks that enterprises face.

22:24and that those bottlenecks are so significant that not only do they have to hire tons of people at each of their companies, including forward deployed engineers, but they actually need separate standalone companies to go tackle the diffusion through parts of the economy. And in this case, they've teamed up. So OpenAI built DeployCo. They did it with a handful of private equity firms. Anthropic did the same. The structures are a bit different. But really what matters is these are standalone entities that are going to be run as separate companies. They're going to have deep connectivity and connection back to the model provider.

23:06The incentive for the model provider is diffusion of their models across a ton of companies. And the built-in distribution is really through the private equity firms. So in both cases, the partners of the private equity firms have two challenges. One, they've probably been investing in a bunch of software businesses that may not actually look like the economic model that they had projected. So already they are having a challenge in their book of business. Private equity is traditionally built to have zero zeros. So venture meant to have a lot of zeros and some outliers, private equity, your ball is just a much tighter band.

23:49And it's really built upon these strong fundamental businesses that have some upside and where they have a playbook they can go run to go get that upside at scale. One, that changes their book of business. So services becomes a really interesting opportunity for them. Two, they have an acute pain and need, which is that they have portfolio companies they have invested in that they will not be able to sell unless they are AI enabled. And so they have to actually implement AI into these companies in ways that actually create real economic value. And so how are they going to do that at scale across their portfolio companies?

24:30It's the same way they've done things across like cloud migrations. When I had started Tribe and I went to some of the private equity firms that are engaged in these deals, their big playbook on how to actually get the returns projected on the operating or value creation side, it was go put these companies on the cloud, go put in some basic business intelligence, and we can move the needle just enough. That's going to be the playbook here. And what OpenAI and Anthropics see is like, hey, built-in distribution, I can go get my tech into these companies. If we don't do it, they're going to go do it with someone else.

25:09And also, you get more service providers who are trained in actually deploying your technology and can feed more of the market. And that TAM, that problem set is so huge. It's something they can't handle on their own. And it's something that, frankly, the services ecosystem isn't prepared to handle because you have the, you know, big consultancies that aren't AI native and you have, you know, some smaller ones like Tribe who are very AI native but aren't yet at scale. And so they are actively trying to build up this ecosystem while also building their own solutions to these problems as well. But isn't this like a playbook we've seen before?

25:54Because I feel like cloud migration had something really similar too, right? Like all of the cloud providers had these huge services arms where we didn't call them forward deployed engineers then, but this is not something completely novel. And to your point about breaking them out as completely separate entities, I read that cynically as like, oh, well, the economics behind like, you know, per employee revenue are much different in a services business than something that's, you know, exponential the way the model companies are. I think you nailed that. They're like, we can hire more people. We can do it like funded by someone else.

26:32And keep the math clean, right? And we keep the math clean for when we go public. A hundred percent. Yes. How did we get to this point where services became such a bottleneck, right? Because if you, like, Jackie, I'll just talk you up for a moment. You've been building Tribe since, what, 2019, 2018? Like, you were building the, like, AI lab for enterprises before we were even using AI. We were still, like, as the term, we were saying machine learning and well before generative AI. And so I think you, like, saw the writing on the wall that every enterprise company needed to become an AI company.

27:06but they wouldn't have the talent in-house to do that. And so there needed to be some way to like guide them through this transition, build AI enabled products, et cetera. It's not like this is a brand new realization or idea. Is this technology hubris? Is this like, you know, Silicon Valley is so tech will solve all problems. We don't need people to solve problems. These mottos are going to get so good that like it's offensive that we would even need like a services component? Is that it? It's part hubris for sure. And I think the other part of the equation is I think they thought they could do this through the large consultancies.