Latest / The AI Ecosystem Podcast / The Road to A2A: Why Agent-to-Agent Starts With MCP

Transcript

The Road to A2A: Why Agent-to-Agent Starts With MCP · The AI Ecosystem Podcast

• Bob Moore: We're back. Hello. Welcome to Lindsey DeFalco: Hello. Bob Moore: another amazing episode of the AI Ecosystem Podcast. It's been probably a whopping two weeks since we put one of these out. So only about eight years of progress have been made in the history of AI and humanity. and we've got a huge list of things to talk about, but mainly centered around our own evolution of how we think about the AI market and AI strategy as it relates to partner ecosystems and software, which is the whole whole theme of this podcast. So with me as always this is Bob Moore, CEO, co-founder of Crossbean. And with me as always is Lindsey DeFalco, our chief product officer. How's it going, Lindsey? Lindsey DeFalco: Great. I'm excited to talk about this. Like you said, it's something internally we've been discussing for quite a while now. And I think you can feel some of the shifts in the market in terms of what people are doing, what they're building, what they're not building, even though there's still a ton left undecided in terms of where thing things will land. But yeah, I'm excited to get into it. Bob Moore: So we have talked a lot in recent episodes about this acronym A to A or agent to agent. and it's really fascinating to us. And particularly because we're in the we're in the business of network graphs and connections between companies. It's only natural that the connections and connectivity between agents kind of strike a chord with us. but I think what we wanted to do today was think about A to A as the destination and spend a little bit of time on the journey. because Today, where we sit is we have an MCP server that's probably one of the most rapidly consumed and demanded product surfaces that we've ever released. and I think we have a roadmap around the evolution of that MCP server and potentially even participating in building some of the agents that might consume it. And only then do you start to get into the realm of where A to A starts to play ball. And I think each of these steps along the way. is instructive and interesting for anybody that's trying to understand how to build the value chain from where the world is today to where there might be opportunities in AI tomorrow. So we wanted to take a little bit more time and surgically break that stuff down. And frankly, the that evolution of I don't know if it's a evolution of what, but evolution of the market that we want to walk through mirrors our roadmap really, really well. and So maybe the best place to start is to start with what we already have. And Lindsey, I wanted to ask if you'd be okay, just sharing a little bit of the data and insights that we've been looking at lately around consumption of our data through MCP as opposed to consumption of our data through our API, which has been around since 2019. Lindsey DeFalco: Yeah. yeah, this is fascinating to me. And I know like we're in this time period where everyone has access to a ton of technology and data that they didn't have access to before. And, you know, people are trying to figure out what on earth to do with it. one quick anecdote, I was at a conference, a partner conference a couple weeks ago and you know, been to a a bunch of those in my time and I have never had more partner people that are talking to me about data architecture. And I think that's such a good representation of where we are right now with this technology is that people that have no technical background or experience are getting exposed to these concepts when they're trying to figure out how to build a solution that are ultimately technical concepts. And so it's interesting because we're in a time period where kind of like everyone is a product person. and people are learning the the different skills and and knowledge required to do that. so I say that because like Bob mentioned, you know, we're doing I I was doing kind of a normal usage check on how people are using the MCP, and then kind of zoom that out to just how how much of our user base is programmatically accessing Crossbeam versus humans actually using our application. And, you know, that's largely informs what we're building, how we're building it, and you know, where we where we index on all that stuff. and it was quite fascinating. So we've had an API for I don't know what, like since day one, Bob, right? Like we've had an API forever. Bob Moore: Yeah, I think we may have even built an API before we built a UI. I Lindsey DeFalco: Yeah. Bob Moore: think back in Buck just texted me, by the way. our our founding CTO, Buck Lindsey DeFalco: yeah. Hi Buck. Bob Moore: Ryan, who really built out the guts of what Crossbeam became, was very very API forward and and I wouldn't be surprised. Yeah. Lindsey DeFalco: Yep. Yeah. So we've had an API for ages, right? So technically anything that people are doing right now with, AI and MCPs, the the same data was accessible via via API. but our usage of our API has never been something that's, you know, transformative or that will like absolutely blow your mind. so essentially what I did is I mapped out the programmatic, the percentage of our Customers who have programmatically accessed Crossbeam. so either via API and then now using our MCP, but in such a way that we can identify that it's not a human, right? Like a human's not hit hitting send, you know, Claude isn't returning a message every point one seconds. so we can tell if it's like an agent that's running something and and hitting the MCP server versus versus a human. and that number went up exponentially with the MCP usage. So you know, looking at like sub one percent of our customer base that consistently pulls from the API and has actually built something off of the API data up to like over 10% of our customer base that is programmatically consistently accessing accessing Crossbeam through the MCP. So I think that's like really indicative of the biggest change in the market that we keep talking about is like That that is even a user base that we're serving now. also, Bob, I sent you this. I got an email today from this tool that is an analytics tool that I love. I won't say the name because you know, we don't want to do a hashtag ad here. but they have a very cute, fuzzy, spiky mascot. and anyway, the email was titled like humans move along, this is for the agents. And then the body of the email was like, Hi agents, we know you're reading these emails for your humans anyway. We noticed that you're using this MCP tool a lot. Guess what? We just launched forty new MCP tools. Here's how you should use them. If you want to do this, do this. so it's just this funny like I think people are, you know, treating agents like their own user Bob Moore: Yeah. Lindsey DeFalco: persona now. Bob Moore: this is so cool. And I I I'm tempted to put more even more hard numbers behind you know, you mentioned this exponential difference, right? And in the consumption rate of MCP. Lindsey DeFalco: Yeah. Bob Moore: it is on the order of first of all, they're they're approaching each if you remove API connections to integrations that we've built and other things. if you just look at things where people are hitting the API directly through things that they've built, not only are we like extremely rapidly eclipsing the number of consumers, the rate at which that is growing. If my math is stuck in my head, right, it was on the order of like 30 to 50 times as fast as the rate at which the API consumption is growing. Was that it it was in that ballpark, right? Yeah. Lindsey DeFalco: Yep, yep. Yep. Bob Moore: so it's it's really wild, right? And I think there's This leads us really naturally to a place where you could draw one of two conclusions. And this is a this is a classic thing in startups with product market fit. You could say, okay, this is great. MCP is clearly the way forward. The way that people are going to consume Crossbeam data is via the MCP. And that's going to be the our AI story is an MCP story. And the evidence would certainly suggest that that's true. But this is where I I invoke my old improv adage. If this is true, what else is true? So if agents are consuming through the MCP at a really aggressive rate and making a lot of requests, then they're clearly finding reasons to query the MCP. And they're finding reasons why, on a repeated basis, the data that's coming out of Crossbeam is kind of inherently valuable. And then you look at the percentage of companies that have ever connected MCP, you look at the long tail of companies that have a million different ways in which they might be able to leverage AI through their own home-built agents, through their own homespun stuff, and actually create systems that are more specifically dedicated to making their use of Crossbeam successful and making Crossbeam fold in with other data sources in a way where it creates the maximum amount of value. And there's an alternative thesis you could come up with, which is that this consumption through the MCP is The early adopter part of a giant bell curve distribution of all of Lindsey DeFalco: Yeah. Bob Moore: the people that might ever potentially consume data through Crossbeam. And there is an enormous meaty middle of more normal, not in the front 10 or 20%, but instead in the the middle 67% that will actually do a better job consuming Crossbeam data if we supply the agents for them. To consume it through. Not to say that you can't use Claude that you can't use all these places with MCP. In fact, we really, really want you to and think it's awesome. But instead to say that we can accelerate the rest of the market on its journey to consuming our data agentically by just handing them an agent. And this would not be an agent where we're building some kind of foundational model. It's simply a matter of we take care of the infrastructure required to stand up an LLM. That sits inside of a harness that is very finely tuned to be able to consume that Crossbeam data and importantly interact with Crossbeam itself. And when we talked about our experiment in the last podcast where we stood up some agents and had them kind of talk to each other, that's what we're talking about, right? It's like kind of a wrapper around an LLM service that specifically is instructed to make amazing use of Crossbeam, optimize your account, leverage the data, identify opportunities, grow your partner ecosystem, curate and fortify data sharing and other policies and things that exist. And if you look at our roadmap, this is the next step is to bring these things to to life. Lindsey DeFalco: Mm-hmm. Bob Moore: Coupled with, I should add, actually, this great. Okay. So we call that the next stage, right? And then here's the if you give a mouse a cookie problem. We're we're invoking every possible metaphor that we ever could. We have if this is true, it doesn't true. And now I'm gonna give you a if you give a mouse a cookie. The minute those agents show up, what did we learn from our air gap test? last time they want more MCP skills and tools. Lindsey DeFalco: Mm-hmm. Bob Moore: Can you talk a little bit about that, Lindsey, and our roadmap Lindsey DeFalco: Yeah. Bob Moore: on MCP and what we're gonna allow to be done through through the MCP server? Lindsey DeFalco: Yeah. And this is all so ages ago, back in I think like twenty twenty-one, we were talking to the company about the concept of a killer app. And I don't know if you remember that, Bob, but it's like the the whole older tech narrative that like you have a platform and your platform will only really make it and be of value if you build the first killer application on top of that platform. and it's funny because I feel like we're in a similar era right now where people are figuring out like, okay, if I have this, you know, AI or agentic platform version of my company, then what is the killer agent or application that people are going to build on top of it? and so this whole thing is really like your classic you know, build versus buy in terms of build with what's available on the platform, buy the application that sits on top of the platform. alongside the the platform. So anyway, it's fascinating to watch history repeat itself with a lot of these patterns that I feel like are just cycles in tech. Bob Moore: Yeah, it's weird too. You're totally right. I had forgotten about that that whole app store thing that we used to talk about. But Lindsey DeFalco: Yeah. Bob Moore: what what's ironic to me, we were talking earlier about did the API exist before our UI existed? And that whole conversation, which is five or six years old about the apps in the app store, is that same argument of like, look, our data's amazing. why don't we just put the data out there and let people consume it? And that'll be the whole business model. And then the argument is made like, look, no one wants to go to the app store until they've seen a killer app, right? And then they can kind of understand Lindsey DeFalco: Mm-hmm. Bob Moore: it. And interestingly now, with through MCP, people actually are kind of going to our app store without killer app. Like they are building their own killer apps because the barrier to building it is lower. But I think it's still it it still does not negate the fact that if you had a killer app, you might be able to multiply. the addressable, the immediately addressable market for the data by as much as 10X, which is my thesis right now, which is that I Lindsey DeFalco: Mm-hmm. Bob Moore: actually think if we were hosting agents that were consuming this stuff, they would be immediately useful and productive for all of our customers. And they too would be consuming data through the MCPs in order to deliver value. Lindsey DeFalco: Yep. Yeah. Well, I think what you just said is like exactly what the whole narrative in the market and like the debates in the market have been, which is like the barrier got lowered so much for people to create their own applications. And then we entered this time period where everyone's like, Well, I can just create my own app my own applications. And n then I feel like now we're finally swinging back to like, okay, maybe I don't want to. Or maybe when I get into that, there's like 10 security levels and governance things that I didn't realize I needed to build in, or my God, our security person figured out that I did this and completely lost their mind. Or, this doesn't cover 10 use cases. Now that I have users, I can't manage, you know, all the feedback that's coming in about it. And like all of the things that are true about, you know, building and maintaining applications internally. and so now I think we're in this like pendulum swing time period where people are like, yeah, okay, I can. I can do this, but do I have all the right knowledge, skills, and time to do it? Or would it be easier if somebody delivered me you know, a packaged experience that is already an expert in this thing and I'm willing to pay some amount of money for it? So again, like classic build versus buy, it's just the the barrier to to build is is significantly lower, which has made the the actual math for that equation different, I think. I don't remember the question that you asked me, Bob Moore: well, whatever it is, it got us where we needed to get to. 'cause Lindsey DeFalco: Mm-hmm. Bob Moore: the yeah, there's a the point you raise here about wow, everybody's waking up and saying, hey, maybe I don't want all of my data from my entire business available to random agents in random locations at random times, like this control era that we're entering into is exactly why A to A is suddenly significantly more complicated than people have dreamed about, at least in a B2B setting, because as you are trying to kind of give a least privileges style amount of access to every agent that might be spun up inside of different contexts to make sure that it doesn't have information it shouldn't have. So it doesn't do things with it that it it might not be advisable to do, then if it's able to talk to other agents, then immediately you have a pathway through which anything that one of them can have, the other one can have as well. And now it flies directly in the face of where a lot of the time and energy and attention of these kind of AI ops groups is going, which is actually like segregating and organizing what can have access to what. And it once you open a communication channel between agents, you kind of break that whole paradigm. so for us again, there's this question of okay, when we started out, we were hesitant to deploy agents because the thought was, well, if an agent can only access Crossbeam data, what can it really do? And We don't want to get in the business of having people connect like 30 other MCP servers to our Crossbeam Lindsey DeFalco: Mm-hmm. Bob Moore: agent just so that it can be useful. But what we're learning really quickly, especially as we stand these up, is there's a lot it can do. There's a huge amount it can do in the same way that anybody, any human who logs into Crossbeam has a huge amount of things that they can do, and insights that they can derive and things that they can kind of get to a point where they have some output or deliverable or action item or signal or Opportunity to increase the scope of what value can be delivered by expanding the network or sharing more data or doing other things like that. So just opening up the keys to that, I think, opens up a lot of potential value. And in some ways, it will actually be easier for a lot of organizations to consume than even connecting Crossbeam's MCP to everybody's Claude instance, which again is is maybe that that wild west zone. We'll look back on that as the glory days at some point in time. which is a really interesting thing. Lindsey DeFalco: Yeah, I think the there's a lot of discourse as well where, we talk about A to A a lot A to A a lot, but I have not heard of a single person who has A to A deployed in production across organizations. It's not a thing that is happening right now at scale in production in any sort of enterprise organizations. And so and there's also a lot of debate about how it will happen. So there is an interesting case to be made for you know, is this actually gonna end up being a hub and spoke model where because of the governance and all of the security required around data that's actually leaving your company walls, are people just going to be more comfortable with like one point where they can go to actually talk to agents from other companies? and is it realistic? Is our whole narrative around I think that we've said here before realistic that like Hey, you're you have a BDR agent and it's gonna go out and realize it needs to talk to its partner and it's gonna go reach out to that partner agent. And if you talk to identity provider companies, you know, they're all gonna say no. You talk to security teams, they're gonna say no. They're gonna say you have a list of authorized agents that you are approved to work with, and this is what you're approved to do, and you know, you cannot do anything outside of that. And so Bob Moore: Mm-hmm. Lindsey DeFalco: That's also something really interesting where networks in terms of ecosystems become really interesting because you if you can provide a place where people can actually interact with each other within the walls of your network, but you are the one secure spot, like Crossbeam, then things can get really, really interesting. Bob Moore: Yeah. so we then come full circle here, which is we spend a lot of time and I wanna give you a lot of credit for just talking to enormous amounts of people that sit in different roles at different companies of different sizes who might be at the intersection of how AI is being stood up and consumed and how agents might collaborate with each other, right? Just really chasing down that agentic collaboration conversation. through founders, through security experts, through AI specific people, through customers of ours, through non-customers of ours. And it was really interesting seeing that trend line of no one's actually doing anything in A to A yet. People believe that it's getting there, but people think that first what's coming is kind of this wave of constraint and control that that will need to exist in order to unlock that A to A future. So from our point of view, if we believe in A to A and we want to play ball there. This is where we get to the next stage of this progression that we're walking through, right? We started with MCP, and then it's these isolated, siloed agents that exist inside of people's Crossbeam instances. And then it's the expansion of the MCP to allow those agents to do more things. And then you very naturally get into this zone where we can open up the ability for these agents only inside of the Crossbeam infrastructure and who by definition only have access to the data from Crossbeam. Talk to each other and collaborate with each other. And it actually becomes a very nice kind of Petri-Dish test case where we can allow agents to collaborate across company lines, knowing full well that we have absolute knowledge of what's okay for them to say and what's okay for them not to say, because the only data they have access to that they might theoretically share is the same exact data that we have rules for, that Crossbeam holds the keys. And is basically the governance policy layer where we know whether or not it's okay to share what sales rep owns what deal with a particular partner. We know whether or not it's okay to talk about a particular deal at all based on what's going on inside their underlying data. So because the agents are then controlled and siloed, the A-to-A activity can, by definition, be controlled and siloed just to the scoping of stuff that we already have companies that have done the work to scope. So This gives us this ability to then stand up an immediately valuable first of its kind thing, which is a giant B2B agent network made above the agents that live inside of these Crossbeam instances with data governance in place. Because the same exact data sharing rules that power Crossbeam can power what these agents are allowed to exchange with each other. And that gets really, really exciting because it like if you talk about. being on the frontier of agentic collaboration. it would kind of be a groundbreaking moment for that, for, for the entire space to bring that to life. And then then you have a foundation on which this stuff sits. And then you get into like that big bright future that we've talked about in past podcasts where you have a much more global policy provider based agentic collaboration layer that's not just about Crossbeam. Lindsey DeFalco: Yeah. And I think when we talk about A to A, we often are talking about one agent talking to one agent. but then with what we're describing, you have two things. You have an agent that represents an organization and you have this policy and governance layer, and then you apply that across the whole network and you have this massive agent network and What are those organizations, what are those agents that represent different organizations gonna do with each other on the one-to-many scale and the many to many scale? Bob Moore: Mm-hmm. Lindsey DeFalco: like things I think will start to get really, really interesting when you have groups of agents that are all working in the same account from different organizations that are coming together. And that's something that I I haven't heard much, much discourse about out in the market yet either that I'm I'm super interested to see. Bob Moore: It does I don't know if this actually if the agents would agree with me or if this is just my human brain making everything human relatable, but I can't help but picture things like Slack Connect as being Lindsey DeFalco: yeah. Bob Moore: kind of interesting role models for this, where you know I've been I've been pounding the table for the last year saying when is somebody gonna come up with Slack for agents? Basically a We've seen agents invade our Slack, but it's it's not a very agent native environment. They're just there and Claude's there and leaves a creepy emoji once in a while. And I I wonder what it would look like if there was a Slack like kind of chat and collaboration environment that was actually built for the agents and the humans are secondary, or put them on some kind of similar footing. and you see little glimpses of this in various platforms that are showing up, but it strikes me that what we're describing doing in Crossbeam is effectively like a giant Slack connect style creation of these collaboration spaces that could exist between individual companies or could exist across multiple companies that are all partnered with one another within a particular ecosystem. And you could drop in as many agents as are appropriate and relevant as long as everybody's partnered with everybody else. And then they can frame what data's there and what data's available. And what's appropriate based on who's in the room. so like, is it even rooms? is this whole concept of rooms like a very humanized thing? And maybe there's some abstracted way in which the agents would would prefer to do this, but I think it does it does raise this interesting world where, yeah, you could you could effectively have a network graph of collaboration spaces that's as big as the Crossbeam network graph, right? So you've got these hundred thousand plus partnerships. Where every single partnership could effectively be yeah, like a like a chat room, a Slack Connect room that goes across company lines. Yeah. Lindsey DeFalco: Yeah. Yeah. And I I do think the concept is because the agents will still be working with humans, which I think is okay and like where we are right now. and ideal. but yeah, I mean, they probably wouldn't use actual chat rooms if humans didn't need to also see and be involved in what's going on. that is something that's interesting. You know, I mentioned that email I got from that product analytics company and it said, it was like, Hey, we know that you just keep using execute SQL. And then it made me think of AWS's MCP where their primary write tool is like run command or something similar. And it's like agents, I think, gravitate towards this like, okay, I'll just write code and execute commands. And then we keep trying to wrap it in human meaning. and they're often just like pushing back and going back to the basics of like, no, I just want to do, you know, the bits and bobs, bits and bits and bytes, bites and Bob Moore: Yeah. Yep. Bits and bobs. That could have been a good name for this podcast. I feel like we had a we really missed out. Lindsey DeFalco: But I think I think I don't know that I want to be called bits, but Yeah, it is but to that point though, when we're talking about the evolution of this whole thing, like it's easy for us to you know, we've had a lot of kind of future looking conversations about, okay, if you know, A2A is the protocol that sticks, first of all, but like the concept of agents working with other agents across company lines happens and you know, they're kind of in this free for all where they're just able to reach out and work with each other. that's all great, but that's so not where we are. And I think that's what has been interesting over the past few weeks is just kind of like grounding everything back in in the reality of, okay, yes, agents are doing stuff, but you know, the majority of people working are still humans and agents are working alongside humans. And what's the actual reality of how humans are interacting with these these agents right now. And That's been how we've centered how we're thinking about this stuff. Bob Moore: Yeah, it is going back, I mentioned the the life cycle of ideas in a startup, and I think we had to march ourselves all the way down this road this if this is true, what else is true road to arrive at this future where A to A is established and exceptionally valuable and powering all kinds of next generation collaboration, only to arrive at the conclusion that we need to start where we started. and it is a very Lindsey DeFalco: Thank you. Yep. Bob Moore: logical progression of these things where you can't just fast forward into building in an A to A protocol and framework in a world where that's the app store and there are no killer apps. and the only person that might even build on it is us. And I think like that's the definition of over-engineering. and interestingly, in in you know, before we talk to real humans about this, a lot of the conversations we had were with Claude. And Lindsey DeFalco: Mm-hmm. Bob Moore: in the spirit of AI psychosis, Claude loves this idea. Claude loves A to A, loves the idea these protocols, offers to make videos and and slide decks. And I don't I I think you know, its ability to fast forward to the future or see things as obvious when it relates to this stuff is very, very high. and I think our job as operators is to draw the Connective lines where we can do really good work product that's going to matter to people, validate ideas through it, and then use that to take us to the next step. and I think if anything, in this exercise, we've been able to find this really cool domino effect of what the next four or five steps likely look like. while creating a really nice build, measure, learn, loop-de-loop along the way at every single one of these steps and provide room for the market to keep pace and also provide room for us to screw up. and be able to make sure that all these foundations are really solid before we start building the next layer on top of it. Lindsey DeFalco: Yeah. And I think that whole pattern is actually just a really good metaphor for how I think we should be interacting with AI right now anyway. It's like if you get too hung up in just talking to Claude, yeah, it will it made some killer pitch videos. I was like, I feel like I could Bob Moore: Ha ha ha. Lindsey DeFalco: I could raise money on this idea. and I was, you know, so pumped up on it, so high on it. and then it's so important to just go talk to humans. because at the end of the day, like if If AI isn't making us better or making our lives better, then like what's the point? And so Bob Moore: Okay. Lindsey DeFalco: the you know, the approach being, okay, what can we do to make the people that are in these roles at these companies even better at what they're doing across the board? and it was really, really enlightening going and talking to a bunch of people because I got very much grounded back into like where we are right now, not where, you know, a frontier model will tell me that we're gonna be when, you know, it doesn't know for certain anyway. So just Bob Moore: Yeah. Lindsey DeFalco: in general I feel like that's a good pattern for like don't sit in a corner and talk to Claude for a week and make sure you you come back to life and and actually have conversations with humans about this stuff too. Bob Moore: Yeah, it is very grounding to get to get into the reality. even with people at, companies that are more often than not on the, on the innovative edge of the curve. like the gap between reality and an understanding of what's possible is very vast right now. However, the speed at which reality is moving is also very vast. and we're in a moment of kind of tail chasing. The nice thing is we may We may look at ourselves six months from now and be there. in in what might otherwise have sounded like an idea that was five years ahead of its time. So the question is, how do you make the most of those those six months? Because if you wait until everything's here, then you'll have missed it. Right. And I and I think it's just participating in the conversation and building against what the market is able to absorb, making sure that what you're doing is building these layers of knowledge and and in some cases technology and frankly customer training and customer readiness. So that as you move to that next step, everybody's ready to to come along. Lindsey DeFalco: Mm-hmm. Yeah. And that is an interesting thing right now, I think, for builders is it's a very we're in this moment of, you know, massive technological advancements, which are absolutely insane. But like if you don't work at a frontier model company, then like they're they're going to they're most likely going to get there first, right? Or you're going to create, you know, open Claude and then get acquired by them a week later. but most of them are going to get there first. And so it's this this time period where we're all just sprinting to keep up and, you know, try and get ahead, but is it realistic that, you know, anyone that's not one of these model companies is really gonna launch something super transformative that they don't get to first? I don't know. Maybe maybe I'm too much of a skeptic there, but like it is this weird you say like chasing our tails. it we're like sprinting to try and keep up. understand what they're coming out with on a daily and weekly basis and make sure that we're building towards the future. and then, you know, hoping that any innovative ideas we have they haven't already built and are launching in a week. Bob Moore: Yeah. this is amazing. I think while this may be one of our shorter episodes, it is probably a really, really important distillation of the journey we've been on to figure out that next best thing to build for our customers and what those next sequential things are gonna be. And we are releasing at the highest velocity that we ever have as a company. So We're excited to have anybody who's listening to this and who have have tuned in this long along on that ride with us. And if you want early access to any of those steps along the way, where to find us. we're we're always excited to work with people who are excited to geek out on this stuff with us. Lindsey DeFalco: Yeah, big plus one. Cool. Bob Moore: Amazing. Well until next time, keep on fighting the good fight and we'll see you out there in the AI ecosystem. Lindsey DeFalco: Mm bye bye.