Latest / The AI Ecosystem Podcast / Network Effects: the Agentic Intelligence Moat

Transcript

Network Effects: the Agentic Intelligence Moat · The AI Ecosystem Podcast

• Bob Moore: Hey, welcome back to the latest episode of the AI Ecosystem Podcast, that one podcast where we talk about AI and not ecosystems very, very often. But man, are we going to buck that trend today because Lindsay and I are deep in a philosophical discussion here around the implications of everything we've talked about in our last few episodes, rippling its way through to... Lindsey DeFalco: Mm-hmm. Bob Moore: like ecosystem like growth, concepts like the future of what it means to do work in partner ecosystems and partnerships in general, and of like these new paradigms around how AI is change the jobs to be done ⁓ in ⁓ Partner Island, also how maybe it'll build these like rocket ships off of Partner Island into other exciting places. So ⁓ let's dive right in. What do you think, Lenz? Lindsey DeFalco: Yeah, maybe we can name this episode the episode where we actually talk about ecosystems. Bob Moore: The one, yeah, the episode that should be episode one, but I just had to make a stupid AI video before we could really go down that road. Lindsey DeFalco: Yes. ⁓ Yeah, I thought it was really interesting coming off of last week's episode, which like in isolation is really depressing. And I think even ⁓ the article, I actually had my whole team read that article because I was. Bob Moore: In fairness, in context, it's also very depressing. Lindsey DeFalco: Yes, that is true. Yeah, it's just kind of depressing. Yeah, I had my whole team read that article and one of them came back to me and was like, it was really hard going back to work after that. And I wanted to talk a little bit about that because I think it's interesting because it's so easy to get, well, first of all, the way the article position I think is really well done where you're like in the reality and you feel like you're in it, which Bob, have you read it yet or no? Bob Moore: I ⁓ got the gist. ⁓ Lindsey DeFalco: Yeah. But coming out of it is funny because like you go back to work and I'm sitting there and I'm talking about features that we're building. I'm on calls with our customers and we're, know, we're talking about working ⁓ these year, three year deals. We're talking about, you know, integrations that maybe involve AI and maybe don't with customers. And it was a, it's such a whiplash from like this possibility that we talk about, back to reality. And I think there's a few things there and one is like, How do we make sure that we're with AI? ⁓ But then the second is also like, what is the reality? What companies will What companies are sitting here still doing these things and ⁓ they fully be replaced? Which other ones won't and how? And a lot of the unanswered questions we talked about. that's kind of some of the stuff I want to touch on is just like a little bit more grounded in our current reality of where we are. Bob Moore: Yeah, it feels like there's so, there's this ⁓ of kind of this belief that there's this new generation of companies that will kind of help define the way that A in B2B, how work gets done, but maybe more broadly in humanity, like how life gets lived. ⁓ And, know, those are forthcoming, maybe aside the core model provider companies and what they're doing, right? It's kind of to be determined. very likely the most impactful companies that meet that description haven't been founded yet, haven't been started, they're coming. And there's kind of a belief that that's a thing. ⁓ then there's just, yeah, like we're over here signing like, you know, million ⁓ three contracts with businesses for our company as we know it and love it today. And like those companies are gonna get tremendous value out of Crossbeam over that three years. But over that three years, it might also... ⁓ Lindsey DeFalco: Yes, Bob Moore: they might change radically and we might change radically. And the question is like, in order to enter in those kind of long-term relationships on either side of that contract, there's gotta be this belief system that there's still gonna be a there there on the other side of it. there's gonna be something to build toward or something that makes us matter to them and then matter to us over that time period. like, is that? What gives them that confidence to make these long-term commitments to us? And what gives us that confidence that we still wanna be in the business of? ⁓ of ecosystems and of exchange of data between companies still ⁓ mattering. And maybe that helps narrow the focus of the discussion a little bit because there is a little bit of this, like, on your question around what companies are gonna survive, I do think there are a lot of semi-viral ⁓ thought leadership things floating around the world right now that define some of those things that it might make for a company that's gonna make it out to the other side. Lindsey DeFalco: there's lot of about motes and what the motes are. that always makes me laugh because a mote is a word that Bob has used for years when talking about a ⁓ And I, I famously a decket, not that famously, but at one point really leaned into the mote theme and there were, ⁓ it was very, you know, like Roman or like medieval with Bob Moore: Thank you. Lindsey DeFalco: Knights and armor and a castle and a moat. And then I really lengthily what a moat was to a bunch of people on a call anyway, moats. So yeah, there's there's a bunch of different. ⁓ Yeah, there's bunch of different ⁓ moats that are in like think pieces and it's kind of gone viral. And a big one is ecosystems. And I think that's been that's been something that, has been talked about terms of Bob Moore: I feel so validated, My day is here. Lindsey DeFalco: virality before, but it's generally just been talked about as a good way to have a viral product and not like the only way to have a product that is defensible. And so I think there's like a very obvious bucket of companies that are easily reproducible and those are ones that are, you UI workflows. And I think that'll continue to be a battle of like taste and experience and how well you're actually, you know, serving your users and your customers and kind of like some of that more standard. like SaaS product stuff. then you get into networks and like networks become even more because it's not just about your ability to expand virally, but now it's creating this moat that is really difficult to replicate. like, Bob, I'd love to hear your, ⁓ always, you've kind had this take that you've said about like the era in which crossbeam specifically was created being an advantage because of the macro climate, the funding environment and all of that. like, I think ⁓ AI changes that calculus for lot of companies that don't have of these modes that are being described, but it kind of doesn't in a lot of ways for ones that do. Bob Moore: Yeah, so you mentioned two things there. You mentioned ecosystems and you networks. And I think I'm at all these ⁓ kind think pieces and commentaries around like what companies seem to be persisting or having a very high chance of thriving even in the AI era, even if they were founded in a legacy time period, ⁓ network effects does just kind of like bubble up. to the top of every single one of these lists. So much so that we went from maybe a month ago, people had a list of like, here's the eight times of kinds of companies or attributes of companies that are gonna survive in the AI era. Network effects kept being number one on those lists. And now there's these think pieces like, here's the eight different kinds of network effects and why they are or aren't gonna be valuable for your company. like people have leaned into the network effects conversation enough that now they're kind of appropriately so splitting hairs on the difference between. Lindsey DeFalco: Mm-hmm. Bob Moore: For example, data network effects, which is really just data economies of scale and saying like, Hey, if you have a million customers, you can look at how those million customers use your product and learn how to build a better product. Cause you have more, you know, like just scale makes you smarter and you can use the data from scale to build better things. Like that technically is a form quasi of network effect. I don't believe that to be a network effect in the truest sense. To me, network effects are things that inherently make the core value proposition of the same product, even if it doesn't change, better or more powerful as more people join it, right? Like Instagram is a network effect company. Like if they never changed their application, it's just better with a million people on it than it is with 10,000 people on it. Crossbeam is the same way. If you sign up for Crossbeam and none of your partners are using it, you're gonna have a pretty bad day. You're gonna need to go and convince a bunch of people that you work with and collaborate with to... to join this thing and get set up and you're gonna be sitting there waiting while that's happening. But now that Crossbeam's got, you know, going on 40,000 plus companies that are on the platform, you show up for Crossbeam, your partners, at least some of them are probably already there and you're gonna be able to understand and realize the value proposition pretty much immediately as soon as you get the first one connected. And that is it, that is a network effect and that is a data mode. And I think the argument here around why network effects... businesses might be so persistent in the AI era is that network effects create this very interesting kind of walled garden of data layer are second party in nature. And this is where this distinction of first party data, second party data, third party data becomes really, really important. first party data is your data. It's the stuff you know. In the B2B context, that would be things like what's in your CRM system, right? It's like who you've talked to, maybe like the transcripts of your call recordings, what opportunities you have open right now, a customer of yours and who's not, what customers have you tried to win and you lost a deal with? Like these are in walled garden and they're protected to you and they're kind of your secrets and no one else has them. And that's great. And that's first party. It's very, very, very useful and informative. And it sits at the top of the funnel too, right? Like. who's ever filled out a form on your marketing website, who's attended your company events, all that kind of stuff. ⁓ so as a class of data, first party data is actually great and really valuable. My first company, ⁓ RJ metrics was basically like a first party data analytics engine. There's your RJ, you're laughing ⁓ your RJ metrics reference for the day. ⁓ but yeah, it's pretty square on the bingo board. It's a guarantee. Lindsey DeFalco: It's a free square on the bingo card. Bob Moore: you know, entire industries have been built around taking people's raw first party data and like translating it into actionable intelligence. ⁓ like, you know, which accounts are probably the most ripe to upsell and, know, who's been engaging with you so much that they're a qualified lead or something like that. Right. Great. Lindsey DeFalco: And that's that's like PLG. That's literally what like all of what PLG was founded on is using all that. Bob Moore: That's a great point. Yeah, really is. PLG is like, just get them in the door. And once they're in the door, you kind of own this lens on how they're interacting with your product and how they're journeying through your conversion funnels. And all of that is first party data. It's like, what buttons did they click? What features did they use? Your competitors will never have that. They'll never know how folks are interacting with your product. your data. It's first party. So it's great because it's like, is your secrets and it's also extremely specific to you. So first party data is a gold mine, but there is like a limitation on first party data, which is if you don't know it, you don't know it and you can't artificially grow it or manufacture it. Like first party data is extremely expensive to accumulate because you either need to some kind of a mechanic or loop or other thing that drives people into becoming part of your first party data by signing up for your product or attending your event or filling out that form or talking to your sales rep. These are things that cost the counterparty like time, social capital, just like things that are hard to extract from them. So first party data doesn't come free and it's something that it's almost like once you have it, you've already kind of like gone over some of the biggest hills in getting a company off the ground. ⁓ Lindsey DeFalco: And I'll say too there I think who's ⁓ worked in company where you needed some sort of, you need to ask some sort of question about this type of data has had the moment where you're like, ⁓ shit, we're not tracking that. And you can't get it back. ⁓ that's always like, it's one of things too, that you, ⁓ once you have and if you start by having it, you're significantly more ahead than if you ⁓ are trying to it later on. Cause you can't. retroactively get it like you can with some third party data, for example, that I think we're to get into. Bob Moore: Incredibly true. Yeah. mean, there are, it's funny if you look back at even some of our earliest customers at a cross beam, whole host of businesses that were just in the business of like, just grab everything and throw it into like cold storage and you can go analyze it later. Heap was one of those, one of those companies, one of first customers ever. like their whole business model was like, we're going to grab the entire Dom every single time any one of your customers like goes and visits your website. It's like, Lindsey DeFalco: Yeah. Bob Moore: We're just going to throw every single JavaScript event that happens, every single like thing that gets, you know, pixel that gets clicked on all of the surrounding context on what was, what CSS class did that button belong to and everything else? Not cause you asked for it, but because in six months you might want to say, Hey, I wonder if the red button is performing better than the green. And if you're grabbing all that stuff, you can go retroactively do it. That is like the, aggressive form of first party data collection, but you're right. A lot of that stuff, especially at the top of the funnel with like interactions and products and websites and things, it's just like super ephemeral. It's like, if you don't grab it and figure out a way to store it in a way that you can later go explore it, you've kind of, you kind of missed the boat and it's gone. it's a really, really weird thing. so that's one class of data, right? That's first party data. Let's skip over second party. Cause that's kind of our punchline here. I would bury the lead. Let's go to third party data. Third party data is data that is basically observable. or acquirable from the outside. A great example of that would be you can go to any number of services. Many of are our partners, so I won't mention any by name here, but we all work with them and know them and, paid money for their services where you can say, hey, ⁓ who's the CPO at Crossbeam and what's their email address? and it'll come back and it'll say, that's Lindsay DeFalco and her email address is blah, blah, blah at crosswind.com. ⁓ We're email choice by the way. ⁓ So, ⁓ you know, it's basically data enrichment, right? Is one example of third party data. ⁓ Other ones might be like, I don't know, reverse IP lookup services, right? Like, hey, I know all these IPs that came and visited my website. I want to know who owns them and like, you know, de-anonymize them. Well, it's just a giant. Lindsey DeFalco: Thank you. Bob Moore: table somewhere of like reverse lookup stuff. the important through line about third party data is ⁓ who wants that data and is willing to either work hard ⁓ spend money can get their hands on that data. So ⁓ I want to know Lindsay's email address, because I want to sell to Lindsay, I can pay ⁓ favorite provider to get Lindsay's email address and they'll charge me a certain amount. And if my competitor is also trying to sell a competing product to Lindsay for the same reasons at the same time, they too can also pay that competitor and that competitor will give them the same exact information. So the thing about third party data is that unlike your first party data, it's not private. It's not yours. It's not proprietary. It's not something germane or specific to you. It's just kind of information that some other party or yourself, if you built some script to scrape it or however else you get your hands on it. has accumulated or detected and anybody can get it. And in that way, it's like undifferentiated. However, what's great about it is that it's actually pretty actionable in a lot of cases. Like it'll complete the picture on how to actually email Lindsay. Very, very actionable thing. It'll complete the picture on who the executive is that holds the title at the company that I care about. Do I want to make sure that company has at least a thousand employees or a hundred million dollars in revenue or whatever else? Third party data is kind of the enrichment layer that allows you to do that. So As a component of good decision making, it's kind of this essential information layer. It's just so overwhelmingly commoditized. And with the advent of LLMs, especially ones that can access the internet, if you're looking for information about one very specific, narrow company or person, it has also become cheaper and easier than ever to just get that data as kind of like a point answer. ⁓ And because of that, there's a bit of like a race to the bottom line. Lindsey DeFalco: Mm-hmm. Bob Moore: quality and pricing and the market's more crowded than ever and you know third party is kind of this really interesting minefield right now where the data is as useful as ever maybe maybe more so but the differentiation is smaller than ever between the companies that do it. Lindsey DeFalco: Yeah. And I think that's, mean, third party data has been interesting for that reason for a while, which is like, you know, when you had, just a few really key providers of it in the market, it's like everyone was targeting the same accounts. Um, and you know, we talked a lot about that when crossbeam started, which is like, is, is outbound dead. There was that whole movement called near bound. I know if you guys remember that shout out to near bound. there is a. Bob Moore: Never heard of it. Lindsey DeFalco: dilution of the value of the data when everyone is using the same data to target everyone. And so, you on that side, it's, it ⁓ its But I also think about it like breadth, right? Like if you think about first and we'll get into it second party data, it there's a certain there, but it's also you're not going to have every single company in the entire world visit your website so that you can have information about them and how they visited your website. So even now, I think like, like you're saying in terms of having a breadth of context to feed into these ⁓ and for decision making. That's where, regardless of whether or not it's kind of diluted because everyone has the ⁓ information, it still provides a ton of value because you get ⁓ breadth that you're not going to get from any proprietary data set that you're going out and curating yourself. Bob Moore: Yeah, yeah. And I think, you know, when we talk about the SaaSpocalypse a little bit, one company that I have a ton of respect for, especially Henry, the CEO here, I think is just an absolute dynamo, but who's been beat up a ton in the public markets lately is ZoomInfo, ⁓ which is probably of the more prominent brands here. But if you look at this business, you know, just in the last six months, and I guess maybe year to date is also relevant, like year to date, right? So we're sitting here in March. Last three months, the stock is down over 35%. And what's even more shocking here is the market cap, right? Like the understood valuation effectively of this business is almost $2 billion on the dot. If you go back and look at their income statements, they did $1.25 billion profitably, by the way, with $125 million of net income last year. So they are trading at less than a 2X revenue multiple on the public markets right now. which is pretty wild. And I think maybe just speaks to this idea that there's a... there's a limitation on the expectation of long-term value that's gonna come out of companies that purely play in the third party universe, despite the fact that that data is undeniably like valuable as an asset class. That's kind of fascinating. And we've seen Zoom Info, they acquired Chorus a couple of years back, which is an early Gong kind of call recording competitor and like. you they have a bunch of offerings that are not just third party in nature. And I think that's, you know, that's, that's where I'd be focusing if I was in that business right now, because I think diversification out of just that third party category is, is, is pretty essential. which brings us to second party, unless you got any more comments on third. Lindsey DeFalco: Yeah. ⁓ yeah, I was just gonna say, think it's like more of a macro comment on, you know, when we are talking about all of the theorizing about the SaaS, SaaSpocalypse is like, we are in this ⁓ moment right now where builders are really having a heyday ⁓ and anyone worked in early startups knows that generally the people that build are not the people that scale, ⁓ which why there's, you know, a complete, usually swap out of the leadership team at some point. Again, this is in like traditional, SaaS models when you hit certain levels of scale, it's different people that are creating these zero dollar businesses that are running these 100, 200 plus million dollar ARR businesses. I don't know if that's still true, which I think is interesting. ⁓ But I've always like the most fascinating thing to me about what's going on ⁓ with the ⁓ startup with ⁓ AI right now is the fact that like Creating things is easy and scaling them is hard. And you can create anything you want. The second, I created this dashboard yesterday using Claude Cote. It wrote this app for me so I could run a dashboard in Snowflake. I was like, ⁓ this is great. ⁓ it to a bunch of people on the team. Immediately, the engineers were like, ⁓ actually makes us look significantly worse ⁓ it is because of the way you're showing this metric. Bob, responded and you're like, I don't know. I kind of want to see it in this view. And in this view, do we have it? in a time series we have it in this, like stuff is great if you don't have customers. When you have customers, then all of sudden you've got this flood of ⁓ feedback and people that you can disappoint. so that's another thing that I'm like, all of the kind of counter discussions to all the SaaSpocalypse stuff is like, is this a bubble that's going to pop? And we're very much in the builder phase right now. A bunch of cool stuff is being created. ⁓ And it's like, what does that look when there are customers and you've got to maintain it. so, know, I don't, I don't know what that means, but ⁓ you're thinking about companies like Zoom Info and you know, they've like built, run, scaled this company. Are other that are popping up, gonna make it in the long haul? I don't know. Bob Moore: Yeah, it's a little bit of ⁓ you know, it's like Slopfest 2026 right now. ⁓ It's just like, ⁓ think, I don't think we've gotten through an episode of this podcast without talking about AI Slop through some different lens, but you bring up, yeah, it's like a great example. Like your, ⁓ thing about your dashboard is that it worked and it's valuable and it was used to like express a point, but here we are, I don't know, 24 hours later, and I'd probably just put it in the Slop category because there's a, Lindsey DeFalco: Yeah, exactly. ⁓ Whoa! Bob Moore: Like you're going to have to make a choice. ⁓ You have to make a choice about whether this becomes like an institutionalized tool is kind of repeatable in nature and customizable in nature. And if that's true, you kind of have to productize around it. And maybe the answer in the AI era is that you develop a skill inside of Claude that knows how to generate dashboards like that. ⁓ allows users to kind of, you know, have some discourse with our data. And then it spits out a brand new dashboard every single time with exactly what they want the way they want. And it's accurate. that skill matures and evolves over time. But idea of a dashboard as we know it is kind of like moot because that dashboard inherently, even like the metaphor it comes from, it's like. It's these rigid physical like speedometers that are sitting in front of you in, you know, arranged in a certain order that you come back to and look at every time you're driving the vehicle or flying an airplane. like dashboards are dead. It's there's a, can have analytics output and analytics report, but it's so easy and so fast to customize it to the needs of a team or a person on an individual day Like the dashboard is really about like measuring the same thing over long periods of time with some consistency. like, I think dashboards have kind of been for so long as both these point solutions to answer questions and these like long running ways to track business progress. I mean, more specific than that now. It's like a, it's fun and it's exciting, but man, is it a paradigm shifter. Lindsey DeFalco: Yeah. Yeah. Which is funny because I'm like, dashboards are dead. Maybe. I don't know. I still like looking at graphs. sure, maybe I can ⁓ you Claude or whatever to make me a graph in line and that's all I need. And maybe I get like 90 % of what I needed from just querying the from, you know, like a chat interface. But I think, yeah, this is part of the like, what are there things ⁓ as ⁓ humans that we because been doing them ⁓ or... ⁓ what are all the different ways that we actually like to ⁓ experience the data. So yeah, more unknown questions, but shall we get back to ecosystems? Here we go again. Bob Moore: Yeah, yeah, yeah. So look, let's round this out, right? We talked about first party data, which is your stuff that you know about your people that you've engaged with. Third party, which is like what the whole world knows about everybody, including people you haven't engaged with and have. And then we get into this very interesting kind of traditionally narrow set of data, which is second party data that people don't talk about much. But second party data is basically what the people I know know. about the people that I care about. So literally like the second party in that relationship, right? It's like, for more practical purposes, I think it's safe and fair ⁓ say that in most cases when people talk about second party data, they're talking about partnerships and partner ecosystems. So if run Crossbeam and Lindsay runs Lindsaybeam and are partners with each other and we sell into the same market, two complimentary products, or maybe Lindsay runs a services organization that helps install and stand up and service my product. We are selling to the same people very often at the same time, very often co-selling in concert with each other. But I know things Lindsay doesn't and Lindsay knows things I don't. She might have contacts at that company. She might have closed a deal more recently and learned about some new procurement process. She might know more about how ⁓ users interact with her product. And while I know how they interact with my product, I've got my first party data. She's got hers. And there's some very interesting Venn diagramming going on there between that knowledge set. And if we can figure out a way for she and I to collaborate on an account, then we both get the superset of our knowledge about that account. And it makes us both smarter. And what's interesting is that second party data gives you the benefits of first party data, which is that it's this walled garden knowledge that no one in the world can get their hands on except for you and your partners in this case. without the limitations of first-party data, which is that you only know what you know about the people that you know, right? So you kind of break down those walls, but you still get kind of the quality and the proprietary nature. And then in the same way, it kind of has the benefits of third-party data, which is it's net new information that you didn't know otherwise. Some of it about accounts who already know, some of it Lindsay might be willing to tell me about her customers who are not my customers or who are in my sales pipeline. So it's like net additive potentially. But again, without the limitations of third-party data, which is that anybody can... have it just by throwing effort or dollars at it. Like no one in the world can buy a partner ecosystem, at least not directly. So because that, you get into this really, really interesting world where get out your bingo cards. It's almost like a two by two matrix, right? Where on one dimension is how much does it expand your knowledge, which is what makes something actually actionable and adds to the scope of what actions you can take and what playbooks you can run. And then on the other dimension is like how proprietary is it? How much is it yours for the sake of you? And it's defensible ⁓ and something that your competitors can't get their hands on. And you kind of get the two corners of that two by two matrix for first and third party data. It's that upper right quadrant that second party sits in and it's so valuable. And this is what brings us all full circle back to where we started around what companies are going to survive in this era. When you think about network effects, most companies don't, especially in B2B. Most companies do not actually have a true network effects model, but there is one thing that every company in B2B has has network effects embedded in it and entrenched in it by nature, which is their partner ecosystem. It's like by investing in your partner ecosystem as a B2B company, you are inherently investing in the network effects of your business for a whole host of reasons. Go to market wise, product wise, and ⁓ of aggregate market defensibility and positioning wise. are networks and they give you network effects as you grow them and invest in them. And in the AI era, if you invest successfully in ecosystems, what you're doing is also growing that second party data asset. And that second party data asset unsurprisingly is this differentiated knowledge that effectively can serve as agentic intelligence. It's this knowledge set that you can feed to your agents, whether they're directly taking actions or they're just turning your humans into super humans in terms of how they do their work. That agentic intelligence coming out of the second party data layer is your data that is both actionable and scalable and proprietary to you. that is how these dots get connected where some people talk about networks, some people talk about network effects, some people talk about ecosystems as being like this survivability secret weapon in the AI era for B2B companies. And this is why, because ecosystems are inherently network effects that now create an output that is this data asset that is proprietary, agentic intelligence, which allows your agentic capabilities to actually be differentiated and matter and standalone, right? And when all that comes together, I think the answer for everybody is like, if you're not tapping into that second party data yet, you're not even at the table. Lindsey DeFalco: Nailed it. think we got our snippet for the, for the socials. Yeah. Yeah. No, I think it's, ⁓ something that ⁓ ⁓ meta about all of this to me is like the, ⁓ ability to actually see all of that data together, the first and second and third party data, that is what of these LLMs ⁓ are doing. that is what they're doing. And so. ⁓ Bob Moore: Yeah, I'll pick your bumper stickers. Yeah, you can print them up. Print them up now. Yeah. Yeah. ⁓ you're still right. Yeah. Lindsey DeFalco: It's funny because like, I, is no reason why I will go into Crossbeam and ask Crossbeam AI chat a question or go into, you know, our, what's it our forecasting tool and ask their AI chat a question. I'm just going to go into Claude and I'm going to ask Claude a question and it's going to go look at our first party data, our second party data, and it's going to go look at our third party data. And so it's like, you know, And I think I've mentioned this in a couple of our other podcasts potentially, but like when you started Crossbeam, you talked a lot about why it was the right time to start Crossbeam because of where the API economy was at the time and like the, the actual ability to build the product because the technology existed to connect the dots. And I think we're in a, ⁓ same spot, different timeline now ⁓ where this whole of the fact that like first, second and third party data are this. agentic intelligence layer and they don't need to be treated separately or consumed separately is all possible because of the ⁓ economy now, right? Like it's not APIs, it's MCPs, but now you can ask one question and get the answer from all three different sources. And like that's gold. Bob Moore: Yeah, you're so right. My whole rant that I just gave on first, second, third party data, I could have given that rant five or six years ago. It's not untrue in terms of the definitions of these things and why they're valuable. So ⁓ the why now kind of comes in two flavors. And one of them is the, companies need network effects and this is a place they can lean into them to create their own data modes. And that's great in the AI era. But then I think your point is actually the better one, which is that there's an ability to process this information that just wasn't there before. the consumers and people could leverage this second party data really used to just be partnerships teams who kind of live on partnerships island and have all these challenges politically and operationally internally where they don't get invested in as heavily and ⁓ their ⁓ attribution isn't as and they're kind of ⁓ not always treated as... major needle movers inside of organizations in the way that maybe they deserve to be, ⁓ or maybe in some cases they don't. ⁓ And that creates all this tension where the ability to actually leverage and parlay this second party data into outcomes, measurable outcomes as compromise. And partly because there's a lot of data there and you got to pick your bets and humans can only hold so much context at one time. Well guess who can hold a lot more context at one time? These LLMs and... ⁓ point is so spot on because what ends up happening for the first time ever is ⁓ because Crossbeam built a new feature, but companies and own internal operations teams, companies and their own AI agents, ⁓ and frankly, increasing universe of third party services companies that we're seeing. There's this company, GTM Fabric, that I think is doing some really smart stuff on the services side about kind of what happens at the intersection of second party and third party data. And how can you use the coalescence of that information to kind of build go-to-market programs and kind of operationalize around this intelligence that's coming in. And whether you're doing that with humans or you're doing that with AI, it's unquestionable that the AI layer is actually making it possible to turn, you know, this giant universe of Venn diagrams of data that we used to just have to figure out how to process or prioritize basically into signals. Lindsey DeFalco: Mm-hmm. Bob Moore: that can initiate actions and further research and recommendations and tasks and next best steps that either agents can take or agents can recommend that humans take. So the AI era creates this unlock whereby an extremely complex multiplication of diverse data sources used to be next to impossible for human intellectual limitation reasons, for internal political reasons, for prioritization reasons is now effectively free at the margins. So now you can take your first, second, third party data and basically have something constantly crunching it, looking for the very things that represent the next unit of value for your company. And is really, really big for its second party data's moment as a result of that, because in the past, when you had finite resources, it was one of the hardest things to incorporate, one of the hardest things to stand up. And now it's there and it's processable. And it turns out that it's actually more valuable than the other two. Lindsey DeFalco: Yep. Bob Moore: but it's the most valuable when it's combined with the other two and the three are brought together under one hood. And thanks to MCP and a whole host of other things, now that can be done without even thinking about it. Lindsey DeFalco: Yeah, I actually, went yesterday we ⁓ getting our whole internal, we're setting a, can I say that we're setting up Claude? I just did. We're setting up our own ⁓ company, Claude, ⁓ it's actually to all of our different internal products and ⁓ services. And I this thing yesterday because I was curious about how it would respond. I said, tell me, ⁓ Bob Moore: Yeah, sure. Get out of the bag. Lindsey DeFalco: based on everything you know, what's our next dollar account that we'll sell? And also went and asked that same question ⁓ in couple of different AI chats for specific tools. And it was funny because as expected, they me very different answers. And so I asked our forecasting tool and it was like, ⁓ well, there's this and this in the forecast, but there's nothing that definitively tells me an answer here. ⁓ and I asked him cross beam and it had a similarly, you know, second party data focused answer. ⁓ and then I asked him Claude and it was like, well, looking at your forecasting tool, looking at what's going on with this account was your second party data. And then pulling in this other third party data and information I can find elsewhere. And then it gave me this really smart answer. So it's like, ⁓ ability to actually look across all of those and have access to all that data is the thing that ⁓ ⁓ huge unlock here. And I think like, go ahead. Bob Moore: Yeah. Good. I'm going to tangent so hard you should finish your thought. Lindsey DeFalco: There you go. ⁓ Okay. Well, maybe I will too then, because I was going to say, I'm glad you said signals, because I think a company that's been doing this for a long time is Clay. And they've really tried to like abstract this concept of a signal. many people that are running account-based marketing think signals as this old, like either third or first party data primarily, because that's what's been sold as these data sets. so ⁓ Bob Moore: I'm Lindsey DeFalco: I think Clay has done a really good job of taking this concept of like, anything can be a signal if it is just that, right? Like if it's an interesting piece of information that can move a needle for you, regardless of what that needle is. And now what we have is the ability to not just have things that are signals where, you know, at this moment something happens and you can do that. But it's like, you have all of this context where all of the previous signals across all of these different sources are sitting. And that's like wealth or gold mine of information. Bob Moore: Yeah, I think your tangent and mine are actually the same tangent, which is great, because the question I was gonna ask, and your explanation of Clay is actually a good example of this, which is we're sitting here as kind of the preeminent second party data provider in the market, but it's worth asking the question, like, is all this stuff actually more powerful if it's under one hood, right? Like, what are the limitations of having your LLM in the middle needing to call Lindsey DeFalco: Great. Bob Moore: one MCP server to go get your CRM data and another one to go get your call recording data and another one to go get your second party data and another one to go enrich all that with your third party data. You know, there are limitations there because if you want to ask a question that's predicated on some filtering dimension or combination of rules ⁓ exist across different sources, ⁓ becomes actually pretty difficult for you to... do that because you need to get like a super list out of one to then enrich with the content of the other to further filter down. Whereas in a classic like if I was writing a sequel query and I had all this stuff in a warehouse I could write one where I could join a bunch of tables write one where clause and get my answer. MCP is much better at delivering a finite record or a finite response or you know kind of something that's distilled down rather than saying here's the 200,000 things and by the way I've performed some action on all of them. And when it does do that, it's very costly from a token standpoint in terms of the central brain that's calling all those MCP servers to aggregate that stuff. So, but if everything's under one hood, if you had your first party data, second and third kind of living in a central agentic intelligence source, then that work can actually be done by that source in place. then the MCP can kind of reap the benefits of getting that out the other side. And I think where, when you talk about Lindsey DeFalco: Mm-hmm. Bob Moore: Technology brings these waves of bundling and unbundling. Lindsey DeFalco: Yep. Bob Moore: AI is giving us a lot of unbundling, a lot of re-imagining what even matters and what the nature of software and value is. But I think in some cases, it's going to precipitate some very interesting bundling of traditional industries that have started out in an unbundled state. These tools get more horizontal, the leverage you can get by having it all under one hood becomes more obvious and prominent. Lindsey DeFalco: Mm-hmm. Bob Moore: and what used to make sense as like six different software categories within a meta category is going to end up all getting distilled under, under one. And with friendly, we've seen this already. Like if you look at the, like Gong and you look at sales loft and you look at outreach and you look at zoom info. Five years ago, these were all these very specific point solutions where it's like zoom info is the third party data broker. Gong's a call recorder. Outreach is this automation thing. I should use Clary instead of SalesLoss. Clary is like a forecasting tool. Look at what all these businesses are now. you know, all in each other's backyards. And, ⁓ you they've all got call recorders ⁓ and all do third party data enrichment and they all are doing forecasting in some flavor. And that's a coincidence. And it's almost like this very necessary precursor to the AI era where not only do I think that's going to continue, I think it's going to like really, really, really accelerate. And there likely will be a bunch of market consolidation happening. And I don't know what that means for folks like Crossbeam, what it screams specifically is that we should probably be smart about the way that we play ball and partner and exist in the first and third party data universes as well. Almost just like existential responsibility as a business. Lindsey DeFalco: Yeah. Yeah. And think it's like, you know, we talked in the last episode a little bit about the, we were bragging about our computer science degrees, I think. and the benefit it brings us in this, in this time period. And it is, uh, it's so true. Cause you know, I've been working on this project with Claude cowork and I brought in a bunch of data sets from various sources. Some of them were CSV files and some of them were direct access to, you know, snowflake. Other ones are from our Salesforce, et cetera. Um, what I noticed when Claude was working is that. It was actually taking all of the data and like creating its own database of data and then querying that data. And it's doing stuff like that behind the scenes. And most people don't care about that, right? But it, does bring up this question. I think there's companies that are solving this in a really interesting way. Like, you know, Snowflake itself is spending a bunch of time, creating these AI solutions on top of just having a database. there's this company called Omni, which I know you're, uh, familiar with Bob, which. Yeah. Bob Moore: Yeah. Love Omni. Go buy Omni, everybody. Lindsey DeFalco: which is fantastic product where they're also spending a lot of time creating this AI layer on top of databases and data warehouses. And it's like, there's this extreme that we're talking about where, everyone's like, whoa, MCP. And you're connecting all these MCPs. You hear the creator of, ⁓ you know, ClaudeBot, RIP already ⁓ acquired OpenAI. And ⁓ he talking about like, ⁓ I don't care about MCP connections. I'm just going to go access the data. And it's interesting because yeah, we're all creating these MCP connections, but then, the narrative we have, if you blow that out to like 10 different products, that one question is going to take like 10 minutes to answer. Cause you're beholden to the speed of the servers that these companies are creating. Then you're beholden to the speed at which all that data can actually be joined together and then queried against. ⁓ And then it's like, don't we just dump all the data into one big warehouse and then go ham. ⁓ And so it is is interesting because then, again, the bundling, unbundling, ⁓ happens is ⁓ asking too much of people to know what to do with the data. And we enter this cycle that we just continue to be in ⁓ as think, which is like, what can we do with it? And then what should we do with it? ⁓ And I think it's interesting because as much as we talk about all of the data living under the same hood, when we're talking about first, second, third party data, a lot of consumers Bob Moore: Mm-hmm. Lindsey DeFalco: the point consumers of the data are different personas and different people. And so I think we still end up in this world where like, sure, having them all together is and they need to be queryable very easily together. ⁓ then you also end up in this world where like, you're going to have people that, you know, we have partner people that want to create ⁓ mapping views with a specific partner and talk about accounts. And then we're going to have people who want to look at the status of a certain deal who are managing a team of sellers. And so this is like the ⁓ The continual cycle that I think is hilarious to watch in technology is this like, let's see what we can do and go crazy with it. And then we abstract things away to a point where we're asking too much of the consumer and then we're like back to creating applications to tell them what to do. Bob Moore: Yeah. Yeah. And in this journey, it's harkens back to our last episode of like, is anybody going to have any jobs anymore? ⁓ in, 2028, we're this very interesting period now where people do have jobs and you know, who's buying our software people. we're yet to encounter, ⁓ a single agentic buyer that has kind of like unilaterally decided that it needs our product and bought it and gone through procurement and like, you know, hopefully some people are finding us through. Lindsey DeFalco: Yeah. Bob Moore: LLM's recommending cross beam and whatever else, all that's doing is landing on the desks of humans who actually are taking action. 2028, as as things are moving, does feel aggressive to think about a timeframe that not being true anymore in the majority of cases. ⁓ Not that it won't true at all, and not that we won't see instances of it and it won't start to come through, but like at the end of the day, the slop problem is primarily a slop problem because humans, are not consistent about how they perceive and interpret and are skeptical of information that's presented to them. ⁓ and they do better with guardrails ⁓ and do better with consistency and repeatability and with rituals. And, the LMS don't have that limitation. like, If we were building purely for LLMs, think there's a lot of things that we would do to take the guardrails off or make the guardrails different in nature based on where their limitations are relative to humans. But we're just not in that world. like everything we do sooner or later, like if you start to feel like, ⁓ is this slop? It's probably because it's being consumed by humans and they have some needs or expectations around it. And if we're not willing to maintain it or make it robust enough for big human audiences to consume then we have created. We've created some AI slop and it's contributing to the mess. ⁓ I think we do have years ahead of us until that problem is kind ⁓ moot. It's pretty heavily correlated to the 2028 vision of the future, right? It's like inversely correlated. Lindsey DeFalco: Yeah. Yeah. And I wonder like ⁓ things like skills, actually think are interesting in Claude because of this, because it's like, you already see that need for making things persona specific in a container. And sure they've, you know, named them skills and it's abstracted, but like, I will be curious to see how many of these, how of these ⁓ LLMs like fall into this trap of, of building for personas at some point, or they're to be like skill libraries where it's product manager, be a product manager. Bob Moore: Yeah. Lindsey DeFalco: and you're already seeing content like that out there and it's, is this funny cycle because we're already building, you ⁓ skills that are really persona specific and guiding people for what to do. and then, and that way, like ⁓ we really up coming, ⁓ you know, don't know if it's full circle. It's like a, ⁓ whatever that is that makes a circle on the inside. But, yeah, I think are like a, a toe dip into that, that are kind of interesting depending on how people ⁓ consume them. Bob Moore: Yeah, you're right. It's really so much of what you see. And I think it's kind of click baity too, because I think a lot of the people consuming this stuff are figuring out how to AIFy their jobs. But the thing behind the thing is that their job is not the thing they should be AIFying. the giant concentric circle around their job ⁓ ⁓ the ⁓ of influence that their job has on the broader organization. Because Yeah, a bunch of these really narrowly defined skills. If we were building the way work gets done from scratch in an LLM native way, we would never have the jobs that we have today with the divisions of labor that we have today in the way that they were. It's just not how these organizations would come together. like level of ambition and extreme intellectual curiosity need to exist. it's especially the bigger company you get into where you try to do that kind of thing, you're going to be like, rubbing elbows and brushing shoulders and like whatever the bad version of that metaphor is with people next to you who are thinking about their own jobs and how to defend them and how to expand scope. I that's the conflict that we're all kind of into, but I think it's a trap. you are like a, yeah, you're a product manager. ⁓ that traditionally exists on a team with designers and engineers and middle management and stakeholders that exist on other teams. And the way that you are thinking about AI changing how you do work is to build a product manager skill that writes, you know, product requirement documents and kind of stops there. Then ⁓ may be falling into a trap, feeling like you're doing the average in the future when the reality is someone starting from scratch. likely going to build something significantly more powerful that does the product design and engineering work all under one hood. And you're T-shaped or you're, I heard somebody say E-shaped recently, which is kind of interesting. It's like, instead of your horizontal on a bunch of stuff and deep in one thing, I guess it's like your deep end horizontal on multiple things ⁓ because you're aided by AI. Lindsey DeFalco: ⁓ Yeah. Bob Moore: I don't know, somebody will send me a meme about it at some point, I'm sure. I think that's the world we're headed to. Lindsey DeFalco: Yeah, I think all of that is where my, my skepticism about the 2028 number on that reality is at least is like, again, like the to kind of circle back to the beginning of this episode, going back to work after talking about all of that stuff is such a stark difference. And I think that's even, you know, our company is going to take a month to pause everything they're doing and say, if we started from scratch as an AI native company, how would this work? And I think the The good ones probably will and should, ⁓ Bob, we should probably do that. ⁓ technology, technology transformation is already this thing that's been like a buzzword at large enterprises for a long time. And it's, ⁓ it's like a process, right? Like they hire consultants to do it. It's this long thing. And so the timeline for AI native companies, either like cannibalizing and ⁓ eating an entire sector of, companies, I think that timeline's a lot longer than, we think it's going to be. Bob Moore: Yeah. Lindsey DeFalco: ⁓ And maybe shorter than companies realizing it themselves, but it's all, I think we got a ways to go. Bob Moore: Yeah, yeah. Well, this is a great kind of wrapping up spot, I guess, for this episode. think this is the whole ballgame, right? And hopefully what we've been able to do here is take AI and partner ecosystems and network effects and data and the future of work and kind of pull it all under one hood. And really the spirit of this podcast is this conversation, right? And extensions of it. And of course, whatever other. things distract us in a given week or two. So hope you enjoyed this episode and we'll be back with more. Maybe next time we'll do some more down in the weeds kind of tactical stuff about bringing some of these AI ways of doing work to life in this AI ecosystems era. Lindsey DeFalco: Yeah, thanks everyone.