speaker-0: He's like, Yeah, SAS is just a crud database. speaker-1: S managed services are a a drag on gross margin. Let's ditch the managed services. speaker-0: I'm telling you it's worth zero. speaker-1: Most teams, most companies try to overinflate what we need to do with customers. Alright, welcome back to another episode of the Chief Customer Officer Podcast. This is coming to you on Thursday, November 5th. Jay, how's it going? speaker-0: Hey Jeff, it's going well. How are you? speaker-1: I am I'm optimistic. It's it feels like, you know, we're we're in November. Feels like the end of the year end of the year is like so close, but I don't know. The one thing that gives me a lot of hope is that like the weather right now in Charleston is just so supreme. Like a little a little brisk in the morning, drinking nice, you know, hot coffee at night, you know. I'm sitting on the porch watching a little T V, doing some work on our like screened in porch. It's my favorite time of year. So speaker-0: Yeah, no no doubt. I wanted to give a shout out to our old friend Adam Robinson. Can you believe the news there? Very sad. speaker-1: Yeah, that was one yesterday. speaker-0: very sad. And if if for those who don't know, Adam Robinson's an entrepreneur, Austin based entrepreneur, very outspoken, very fun to interact with and listen to. didn't we have him on the GR podcast? speaker-1: I'm not sure if we ever had him on podcast, but speaker-0: I feel like we had some calls with him though. Well 'cause he was Yeah. I think he was trying to get us to buy R B and B R R B to B? Is that what it was? At at the time. R two D two. But speaker-1: Yeah. speaker-0: I was I was just I was just reading one of his posts last week and I was chuckling 'cause he's just he was just so outspoken. He was sort of railing on Apollo and I just thought it was I just I remember last week I just clicked like on that post. It was so funny and then to hear that news just incredible. Yeah. So Thoughts and prayers of his his family and his wife's family and the family of the pilots that were involved in that in that crash. So speaker-1: Yeah. Seems very very tragic for sure. well, there's a couple of things that I think are are interesting or top of mind for us as we spent some some time over the last week. I think let me start with one that I'm working on 'cause I think we can knock it out and then would love to jump into a couple of things for you. But you know, just it's it's November getting towards the end of the year and so the proverbial, you know, how do we start planning for next year? what does it look like, you know? Team wise, you know, what are the kind of overarching revenue goals we have to go meet, right? You're just kind of doing some of the tops down, at least right now it's a little bit of the tops down modeling. and we'll probably do some of the bottoms up modeling here soon. But it got me thinking, and I made a post this morning and I I'm curious just because I feel like you are very thoughtful about these things too. you know, I've I've kind of got a I'll call it a standard, a standard Excel sheet right now. Claude actually messes with it though. I've just had it build it, you know, in in Google Drive. But You I I've probably enhanced it from what it used to be, right? I think everybody used to build the same one, which was like, here's how many hours I think we spend on accounts, and you know, here's how many of each account I think I'm going to get, and just some percentages of like, okay, cool, this is this means I I need this many CSMs in this segment, this segment, this segment. so I've used Cloud to enhance, I think, some of that, which is you know, we can actually look at things like how many how many calls are we on with them, how much time are we spending with them in Slack? we can actually probably like break down the time that it actually takes us to serve an account maybe much more accurately. And so there's been some enhancements and things like that. But I guess maybe the overarching question that I've I've started to dig into is really I don't think it's maybe a new question, but maybe just because of the advent of AI, it feels n maybe maybe like a larger question or maybe a question that desire derives more accuracy, which is you know, really how do start to plan for next year knowing that, you know, many processes that you have today, maybe many systems or tools or technology are going to probably be a step function different just because of of AI tooling that we have and the capabilities, you know, different skills that we can develop, different tools that we can build ourselves, how we can kind of reinvent workflows. And so I think there's maybe two questions that come out of that for me. One is I'm curious, I would assume the answer is going to be yes, but like I assume people are starting to use I guess like dollars that we're spending on AI credits and usage as like kind of like Baked into how much an employee costs, right? So it does the cost to serve, does the cost to serve actually go up because of our token usage and can we apply that to, you know, CSMs and whatnot? So I feel like that that's gotta be a cost that is like starting to become into planning more and more. So that's like the first question that I'd have. And I think the second question is really just around, I guess, the the age-old question of efficiency, right? If I'm going to introduce more technology and systems and I think there's going to be some version of efficiency, you know, how do you really bake that in? and take that into account. So I don't know. Those are the two questions that are on my mind and I just I I figured we riff on them for a minutes 'cause usually when that happens I I walk away with some ideas I can go ask Claude to help me execute on. speaker-0: Yeah. Well, I think so there's two couple of things going through my head based on what you just outlined. Number one is I think that no matter what kind of team you run, no matter how big it is, no matter what stage you are, it's probably worthwhile right now to try we know, we always talk about this this good strategy, bad strategy book. It's it's one of my favorites. What is the problem you're actually trying to solve in your customer base? I know your team is scaling. You guys are growing really, really fast. And I imagine that that that's creating some strain on the people in your business, probably the processes or, you know, the product. it's strains everything when you're when you're, you know, doubling or tripling or quadrupling every year. So, but I think the the first thing to answer is what problem are you actually trying to solve? And you've shared a little bit with me. So I think you've already done this, right? Like where are the constraints in in the business for what you need to go for what your anticipated growth is? And by the way, like if you're In a in a slower growing business where you're just you're primarily worried about retaining the customers that you have, that's a that's a that's a goal too, right? That's an objective. What are the constraints with that? Like can you identify what's preventing you from doing that? And then get laser focused on what's the action that you need to take as a company to go resolve that. And then once you know what that action is, start strategy first, right? Outside in. Then you say, okay, well, what role are my teams or my people No matter whether they're CSMs or services or support, we gotta start blurring those lines for a minute. what roles do those people need to go play in solving that problem? And then specifically, what does that look like? Like how many hours? How many how many interactions with the client? Where are those interactions happening? Can any of them be automated? What are we gonna actually do? So I still think if put all the tools aside, we still have to be good leaders and good pro speaker-1: Problem. speaker-0: root cause analysis identifiers, right? The five Y's for our teams, like what's what's the constraint? And then what are we going to do about it? And then you build a plan from there, right? And that plan's going to change. Like this idea of annual planning to me, especially if you're growing fast, is you got to be careful with that because you probably need to be changing every quarter, if not every month, right? Your your your strategic focus of where you're headed. So anyway, that's my first answer, but is that too is that too high level? speaker-1: no, it's I mean well so it's helpful in clarifying a couple things, which is one, to your last point, we're I think we're doing that in practice. Like we are you know, we have a weekly executive meeting. and I would say like you can see pretty rapidly how we're like we're planning some stuff quarterly, but that that stuff quarterly is really just around kind of like overarching goals that we need to go after. It's not really getting into like I guess like we're we're not we're not forcing ourselves into constraints like, okay, we're, we're only going to do this this quarter. We're only going to do that, right? We're not bit backing ourselves into a corner. We kind of say, hey, here's the two, here's the two major objectives for this quarter, and how we go do them is a weekly evolving conversation. Like, yes, we have some soft commits and we have some longer things that we're working on, but you know, we're trying to, while we're still nimble and early, right? We're trying to allow ourselves like the space to be, you know what, our assumption hypothesis was wrong. We're actually going to shift over here and we're going to then drive on this for a week or two and see if it's the right thing to go do. So I think that to me is something that we're already doing. And I think to your point, it just yeah, I think like your your point is just echoing the like we still need to like some of the context matters, right? Even yes, we're going to set some like yearly targets, right? We're going to say, hey, at the end of next year, what does it look like? That's going to evolve and change like a bit, but like the context like that we have now matters and we sh we just need to continually I guess be it's almost like the context needs to be something that you're bringing in weekly, biweekly, monthly, right? Like okay, I have new context on our customers, I have new context on like things okay, how long does it take us to upsell cross sell a customer? you know, do what what is the biggest upsell like opportunity? Like what is the the product or the skew that is upselling the most and like Does that become easier? Can we make that easier to do? How is onboarding, right? Have we really made it onboarding as efficient as possible? But I think all those context pieces, I think to your point, matter in that that understanding still feeds the model. And so you can't ignore, you can't ignore that. And like that's where I think I've tried to lean on Claude even more, Claude and and OpenAI even more, since it's connected to all of our tools. again, it it it has done a great job of I guess, using prompts and then using some identifiers for me, you know, segmentation. different fields that we're using, but go find this type of data that's going to help bring the context for us to build a model that is effective for us, as we go. speaker-0: Yeah. And I think y especially if if you have access to I don't know, I tend to lean on I talk about call recordings like it's the only thing that exists on the planet. But I mean that to me, like sort of so much context actually happens with your with your clients that if you can have it scan those thousands and thousands and thousands of words that are spoken back and forth with actual customers on the front line every day, it's a really it's a really invaluable tool. fact, we're we're building go to market agents. Customer feedback loop agents now that that really lean on that kind of stuff heavily for defining our go-to-market, for defining even down to the copy that's going on our website. How are we going to position the work that we do? Where are we going to focus? What are we going to build? Like it's everything, right? Yeah. you asked about cost to serve as well. I think my sense is that, and this sort of dovetails into another topic that I think is relevant here. I don't think Most organizations are using enough tokens in the customer success part of the organization for it to really matter. But cost to serve, like your your software license kind of in your subscription kind of stuff tends to follow the teams that use it. That's how most finance teams will do it. They'll just allocate the portion according to headcount of those licenses. Now, last week we talked about, you know, this phenomenon where you know somebody on your team racks up a $30,000 token bill, right? Which is pretty funny. But so My my thesis is you and me, I'm I'll put myself in this bucket too, and it virtually every every organization that I've seen is not using enough tokens to make for it to even matter, right? Because you have to start using enough tokens for it to be like equ equate not equate to one to two percent of your FTE headcount, but of like twenty to thirty to forty percent of your headcount before it actually is really gonna start A, making a difference and B mattering to finance. So but yeah, so I'll come back. I got another like a an an a podcast that I listened to this week that I'm gonna walk through on that topic. But does that does that make sense? Yeah. speaker-1: It does. Yeah. and I I think that's probably true for like where we are at as well, right? I I think also like at this stage of growth and and everything, right? Like we're yeah. I guess like we're not we're also not afraid to be investing in these things right now. It's like the the time speaker-0: Yeah. speaker-1: like we have the need, right? we have the and we have the we've got the dollars, we have the need, we have like the use cases and stuff. So see that makes sense. speaker-0: So so what I here here's my so here's my thesis. What once you have defined the problem that you want to s go solve in your customer base, whatever that is for you, whatever the constraint is to growth, to retention, to expansion sales, to adoption, whatever it is, right? Then once you've identified that, then what I actually think we need to do is flip this whole thing inside out. When it comes to agents and humans, flip it inside out. What we've I think been doing, a lot of companies are doing, almost everybody is doing today, is they're saying, here's my team, here's the work that my team does. Now which tasks can AI Go automate? Sure. What I think and and what the way we're starting to think about this is how do you flip it inside out and say, The agent's gonna do it all and it's gonna delegate to the human what it needs what it needs human touch for. Okay. So there's this company called Spawn. Have you heard of this company? I don't I'm not a gamer, okay? So I'm not like real familiar with this. But the there was a a podcast that I just happened upon. It's called How the Most AI Pilled Company in the World Works. and it's about this company called Spawn. I can't remember if there was actually, I don't think there was a guest on there. It was a guy just sort of Pulling apart this article that was written about it. And what Spawn has done is they have I don't know, 10, 20, 30 different team members, right, on their team, but they have taken an AI first approach to releasing updates to their gaming platform. They release something like a hundred code changes an hour through these these agents. They spend three million dollars a month on tokens. at that was at the last as of when this, you know, article is written and this podcast was recorded. What they've done is gone from, you know, they first said, okay, this is an agent that's going to take care of this particular part of the game and this part of the platform. And then they had a one of their team members sort of owned that agent and was assigned to it. As they've evolved, they've let those agents continue to to evolve and and and drive the growth and the build of that platform. And what the people that work for the company now do This is where it gets really interesting. They actually play the game, right? And they tell the agents how they want so they keep the agents going the right direction because agents can drift, right? As if they're left to their own devices. So they keep them on track. They the the other thing they do is they parachute in when there are problems, right? And by the way, they're not assigned one-to-one to an agent anymore. They're the the people that work in this company now are more like a forward-deployed engineering support team. Yeah. Looking across the system. If there's an issue, it's surfaced, they jump in. If a judgment call is needed. Somebody will jump in, but it's l less of a specialist role. And then what's the third thing I was going mention? But anyway, the the point is that the they are they are they have built a true agentic system that is self evolving. the third thing is like the agents are not going to make step function changes in the platform, right? Things that are make it a hundred times better. The agents are just sort of incrementally making it one percent better every day. Right, taking user feedback, taking player feedback, looking at metrics, making it 1% better every day. So that the other thing that humans do is say, hey, here's a new direction we're taking this the platform into, and they'll sort of push it down that path. So I this whole concept just blew my mind as I was really thinking about it and thinking about what I see most teams doing with AI now. It's nothing like that. Right? Nothing like that. So, but when you when you get an engineering organization that's actually truly building that way, it sets the bar at a different level. So the question is, for you and for me, how can I take that job that needs to be done with with the customer and just make it hundred percent agenc start with and then have it tag the human in where only a human will do, right? Because there's gonna be plenty of that. There's gonna be plenty of that work to be done still, I promise. You're not getting rid of your humans, especially when you're growing like the way you are. So yeah, just how we pause. Yeah. So anyway, I'll pause there. speaker-1: that's super interesting. So there is a podcast that I have teed up in my YouTube right now that I was gonna listen to after this if I have time. or like at some point today. But it's you know that I it's my ha how I AI, it's Clairvo. You know that I have like a you know an speaker-0: I know which one you're gonna say. I know which one you're gonna say already. Is it about the wealth management platform? speaker-1: No, just that one. Okay, sorry. I'll tell you about that. speaker-0: So the speaker-1: He she just had a a a product manager from Figma on, and I saw a clip that piqued my interest, which he basically also there's two things that caught my interest. One is he's essentially created a product manager operating system that every product manager now uses. And it's like a collection of skills and other things, right? But it's all to your point, right? Now everyone's using kind of this operating system. It's making itself better. Everyone's using the same skills or updating those skills, right? But it's like this whole now everyone's following the same process, just it in largely it feels like, Claude or whatever tool they're using, harness, is, you know, essentially at the core. But the other thing that he said that I saw a clip of that I thought was interesting is he said, let me let me like give you a real specific example. I want to optimize a certain feature on a page, like a button click or something else. He's like, I'm no longer doing that. I have an agent that goes and looks and says, Okay, what's the most likely thing on this page that I could test? Great, let me test this. It creates the variance itself. It runs through the variants, it deploys the variants to testing, then it looks at the data and then it basically comes back and basically says, Hey, do you want to merge this? Hey, I did this t hey basically it's it comes back to him as the human in the loop and says, to your point, like, hey, I went and optimized the page. Here's the thing I think that you know, I think w we were testing that is correct. I you know, here's the data that's statistically relevant. Do you want to be able to push this change and merge it? And can I just basically make this the new version? So he says, like. Now that is just running in a loop versus him having to always come up with these himself, always be looking at the data, right? Like he's essentially developed this. That's right. and and now, like you said, okay, he's the human in the loop, but it's like reverse, right? Okay, I'm the human in the loop, making sure that the decision is really the one that I want because I'm gonna have to be ultimately responsible for it. but all of the underlying work was done by the agents. speaker-0: That's right. And you're you essentially it took someone with a product manager's brain to define the the job that it wanted that agent to do for it, but then it goes, right? It runs on its own. Yeah. So in that situation, I would imagine that always on continuously running agent is burning a lot more tokens back to your cost to serve question, right? Yeah. So what is that for your organization? That's really good. Let me share this other one with you and maybe we could put all these in the in the show notes. Let's see. Hopefully this doesn't share audio. But this is I just again I this is the Late in Space podcast. I've never really listened to this, but it's really for AI engineers. But it's about it it the guy on this podcast, his name is Amon Singh. He comes from a company called ARCA, ARCA. And what they are is the you'll hear him say in this podcast, they are not an an AI company. They're not a software company, they're not a SaaS company. They are a wealth management firm. And Then he goes on to describe the architecture and the the technology that they use to build this. So right now what I'm gonna do is give Gainsight some kudos because I actually think they're doing what every software company should do. Every single software company. If I get another software pitch with feature functions, I'm gonna throw up in my mouth because it I'm telling you, it's worth zero. Whatever you're building in software and trying to sell is worth zero. Because anybody can build it. If you don't have a data moat, don't do it, right? What you have to do instead is sell the work. You have to sell wealth management. That's a service that you provide. What Gainsight is doing, which I I think takes some real guts if you're a software company, but it's the only way through if your investors are willing to support you in it, you have to sell the work. And so what they're trying to do is sell renewals as a service, right? They are saying, no, we're just gonna do the renewal for you. It's a completely different game. Then they build the agent architecture underneath that to to be able to do that at a reasonable cost and gross margin profile so that it's actually a sustainable business for them. Right? So what is the is anyway, so yes, but it's less like speaker-1: Well like just I guess to clarify, it's le so to me though, that's not a that's not a the software is not valuable. It's to me that is like I guess understanding what part of your software is most valuable or what part of the data that you own is the most valuable, right? Which is which for Gainsight is we can help you retain your customers and grow them. And so I think like you're alluding to it doesn't matter which features I can turn on to your team, right? Yes, like I okay, like hey, if if I go to Jeff and pitch him, hey, wouldn't it be great if you just retain more customers and you Expand them and wouldn't it be cool if you also did that with an infrastructure or a layer, you know, that's basically all built on technology and optimizes for your team to basically be the human in the loop only when needed. Yes, I want that. But speaker-0: I don't even care about those words, right? You said technology. What do I care? You're selling me a renewal. That's all I care about. And in in the interaction I get to have with your team, it's a managed service. It's a managed service. speaker-1: yeah. Yes. speaker-0: But let me give you one more go ahead, finish your thought. No, finish your thoughts. speaker-1: So yes, it's a managed service, but at the end of the day, like it's both, right? I think it to me, I guess it's both. It's not just the managed service that you're after because like they're going to deploy some sort of technology to me, a dashboard, an interface, something that humans can still log into. So like, but to your point, I think it's sure maybe maybe the way to maybe like what I er the interesting thing is maybe that like the boom of software companies has been the fact of like Managed services are a drag on gross margin. Let's ditch the managed services, right? Let's actually get rid of services as much as possible. Let's totally optimize on giving the customer software that they can use that will that they can then go do whatever they want with, right? And we don't care. It's managed in the cloud. That's what has been built over the last like 10 years, 15 years. And I think you're starting to see that flip where it's like, shit. They can actually just go build the software already. So what's the most valuable part is that. We've been doing this for thousands of companies. We actually know the motions and we can build the underlying technology better than they can. But now it's now it's actually we should sell them the actual service and the outcome versus the software. The software's like becoming the least valuable part of the equation. speaker-0: It's totally a commodity. I mean, did you see service now? speaker-1: I was just about to say yeah. speaker-0: What 35, 39% last week? Yeah. So, you know, my GV investment is now in question again. But but it's really interesting. So, all right, I'll I'll give you one more example of how we're thinking about this. And maybe they're listening. I hope I hope our our our vendor here is listening because maybe it'll give them some ideas for how to work with us. But we use a PSA tool, professional services automation, right? We get our we got our bill for it. quarterly bill and every time I look at it, I'm just like, what the the hell are we paying for here? Right? It's a little clunky. it gets the job done, but guess what? We actually spend a bunch of tokens in Claude to get the data out out in a format that we need to get it out in. So what we spend on that platform is not what we spend on that platform, if it makes if that makes sense to you. But between my VP of delivery and tokens we spend like there's a lot of additional spend that goes around it. So But we do time tracking in there, we do project financials in there, profitability analysis, at least track the data for it, right? All we're doing is using it as a data tracking tool. Yeah. I can clone that with literally a handful of prompts. Okay. It would take a weekend to clone it. And I'm not saying that I'm not saying that it would be perfect. And I'm not saying that's the best thing to do. Okay. I'm just telling you it can be done. And it probably will be done. by me and my team at some point when it becomes the top priority for us. So how does that company pivot away from being a software company? I mean, they are just a a CRUD database. This is back to Satya Nadella, Microsoft CEO, said this a year ago and people were like, no, no, no, no, no. He's like, yeah, SAS is just a CRUD database. Create, read, update, delete, right? That's a the developer term. That's all this is, right? Because we're doing all the analysis in Claude now. Yeah. So how does a company like that pivot from being a software company To selling me outcomes. Well, they need to sell a service that says, of course you can use our tools to track time and we have to make those usable and friendly. And you know, all the, you know, the resource planning capabilities are all in there. But guess what? We're just gonna do all that for you. We're gonna produce your your project profitability reports. We are going to approve time for you every month. We're gonna build an agent, right? That that that learns your business and approves time the way your VP of delivery would approve time. She's gonna have to be involved and she's gonna have to help us train it, but we're gonna do that. They deploy Ford Deployed Engineer into our organization to make sure that it's implemented that way. And they're bringing back feedback into their agent platform to be able to do that for us. This is the future of where software companies have to go. They we have we are going to have to think of them as verticalized services or even horizontal services that get rendered on an outcome basis. And so I've heard those words for. I don't know, two years now, and they just bounce off my head. But until I really thought through that example and forced myself to say, okay, well, what would I what would I buy from my PSA software company if they offered it? That's what I would buy. And I would pay 10 grand more a year for that than I pay for the tool that I have now that I have to go figure out how to deal with, right? Because it's designed however it's designed. So there's my thesis. If you don't have a data moat or an action moat, I would say that you have an action moat. Junction has an action mode because you're actually doing something physical on the back end to make something happen in the real world, right? So you are doing this already. You're providing your what you're providing is a service. speaker-1: Yeah. speaker-0: Already. So you're already in the good you're already in a good spot, right? From from a go to market offering standpoint. Most software companies are just not. speaker-1: Yeah, no, I I'm starting to I'm starting to, drink the Kool-Aid. I'm I'm kinda getting what's that. speaker-0: As I rant with my hands, you're you're drinking schooling. speaker-1: It I also think it it makes agencies more valuable too, right? Like hundred percent. You can have an agency that comes in and helps you like an AI agency that comes in to help you automate some of these areas and and understand these things. Like there's a a value in that too. all right, one quick share. Since you j you were just bringing up, have you seen this website? Can I vibe code it? Can I vibe code I've speaker-0: No. speaker-1: actually used this quite a bit. So to your point, you can plug any tool in here and it will tell you. whether it thinks it so it goes and tests across all these different models and basically says, can it happen? So for instance, let's just say I'm gonna use Monday.com as project management tool. It basically says like can it happen? Yes, it's kind of a weekend project. You could save 144 bucks a year. Here's the prompt it would use. and then here's like, you know, other things that it it like pulls in. But it's like it's cool now. It's kind of funny. It it I think I did this the other day with resend, which is like an email tool or not resend speaker-0: Yeah, yeah. speaker-1: there's You can type in email octopus, all these. So anyways, and then it has a a death list of which ones are which ones are like ripe for it. now you have to subscribe, of course. But I think it you know, a bunch of of like social tools, social posting tools and stuff are in here. So like I don't know, go go use it. It's it's been pretty funny for me. I've I've tried it a couple of times and I I've actually used the prompts and it comes back with like, cool, like this is you know, like you said, give it some underlying databases and some like data that you want to capture and otherwise it it it works pretty well. speaker-0: Think about so the template of an agent should include an open source structure for the data related to that agent. So if I have a a professional services automation agent, right, or a or a VP of delivery agent, then it should come with a database that you know is is already well known and structured appropriately for time tracking and for tracking projects and accounts and project profitability, all that kind of stuff. Because that's domain knowledge that's valuable, right? Yeah. If you have a if you have a data model embedded in the the definition, that's where I think can I vibecode it.com would would be even more powerful. And you and I have talked about creating open source versions of certain plat types of platforms, right? That could be a starter kit for going to build your own thing. So I think open source has a big role to play here because everybody will eventually have software that's really tailored to their organization. Agents will maintain it and evolve it, and there will be engineers inside of those organizations who manage and and and evolve the agents internally, or it'll be done as a service by a third party. But these won't look like software companies anymore. speaker-1: What's interesting that you I don't know why, or I this is just where my mind's going right now, is that I'm starting so based on your comment right about basically like flip it, where it's like, okay, the agent is doing all the work and it involves the human. Now I'm and now I'm just picturing, okay, now I have a customer journey map, like with what needs to happen with like what the customer is trying to achieve. Like that's when I when I say customer journey, I want like, okay, they're trying to get live, which means like they're trying to get to their first production order. Okay. So for that, that's called onboarding. And Onboarding, what does that look like if it was an agent running onboarding? how does it interact with the customer? What are the things that it does, right? And then when does it involve the human? so I don't know. It's an interesting exercise now to start thinking about okay. Again, I I think I told you this before the call, right? Like I think most teams, most companies try to overinflate what we need to do with customers. We need 10 to 20 journey activities a year with a customer. I need them every single week to be on a call. And I'm just trying to like get rid of that and boil it down into like there's actually really probably five to seven things that like Across the entire year are most important for us to do right with our customers. And like what are those? And if I was to think of them as agent first, what does that look like? Okay. Agent is going to tell me when the appropriate moment is to go do an upsell. Okay, what's it going to look for? It's gonna look for these signals. Okay, what does it need to what does it need in order to prepare that? It needs to access HubSpot records, it needs to access our Slack, our emails, and it needs to probably pull in a bunch of context. Great. Then it needs to create a template of you know, a proposal and put that in front of the CSM. And then the human moment right there is put it in front of the CSM. And basically say, Here's the signal that we heard, here's the proposal that we put together, like based on what you know about this customer and like what you need you know, you have to go drive this with the customer and have a conversation. So like you need to start putting your hands on this like speaker-0: And below a certain tier, I'm just gonna go do that for you because you can't afford to go do that, right? I'm gonna s I'm gonna send out a proposal. Maybe I'll let you review it first. Maybe not, right? Once once the agent gets better and better. I mean, I think that's why you always need a team internally to manage the agents. That's what the ops team will become, right? It's like here's the goal, here's what we're trying to achieve, like let's continue to evolve this thing so it's more and more efficient with both our human time and the token spend. But yeah, this whole inside out, we've gotta flip this whole thing inside out. That that's where my head is. Yeah. And that's that's what we're trying to figure out for our for our company too. speaker-1: all right. Well I think we we didn't get to all your topics. we have three minutes. Is there something you something you want to hit real quick or no? speaker-0: I I think, y you've been talking about AI, like having having teams that are that are sort of AI curious and AI interested. One of the things that we're doing, I just thought I'd share this with you, it is we are gonna become we are right now working on becoming an anthropic certified partner. So we're getting you know, you have to have ten people trained and certified on Claude, which is not gonna be that hard because we use Claude every day. We build agents every day in Claude. but what I what I realized is that A, if you're if you really want to train people on AI, just teach them a tool. Right. And then if you if you teach them Claude, then a lot of those skills are transferable to other agent platforms, whether it's Glean or DevRev or Copilot or Gemini, whatever it is, right? The the concepts are still the same, right? And so so again, that's just an idea that I had. For you is maybe people need more advanced training on on the agent of choice that you're using internally. And it will help them move faster with it and emphasize the importance of it. speaker-1: Yeah. yeah, I think it's a good I think it's a good call. we I know we have one minute, but I would say like one thing that we did last week I think was just a a better at least for a virtual team, like we split up into so there's eight of us, four teams of two, and we just went for literally I blocked off all of Friday. We had some customer meetings here and there, but I I tried to give my team time and space and basically said we met at nine a.m. We picked four things that we were each team was gonna go do by the end of the day. We had a demo time at two thirty or three o'clock and then followed that up with our Monday team meeting yesterday, which was like, cool, we saw the first demo. What are you guys gonna do to drive that through this week and like get it done? but again, I think like splitting into teams of two, giving people these small team like I just sat in a I sat in a slack huddle Sam on my team and we went through and, you know, mapped out a couple of things that we wanted to do for some AI work and you know did that together like on a slack huddle and we did that for a couple hours and then we went away and did some individual work ourselves. We came back for a couple more hours and tweaked a few things and then we demoed it. And I think I don't know this that that kind of style working like works out really well. Yeah. Yeah. I like it. All right. I'll let you go. this is cool. I appreciate your insights. Later. speaker-0: Yeah, you too, man. All right. See you next week.