Jeff Breunsbach: All right, welcome back to another episode of Chief Customer Officer. The Chief Customer Officer. Jay, what's going on?

Jay Nathan: Podcast. Hey man, how are you?

Jeff Breunsbach: I'm doing all right. You know, just it's another gray day here in in Charleston. I feel like it's been you know, been an unusually hot or raining pretty much all summer. No no in between.

Jay Nathan: Yeah, really. It has been raining a lot, which is I'm I'm grateful because there are some places that don't get any rain. But today it was seventy three when I got up to go swimming this morning. So that's nice for us. Yeah.

Jeff Breunsbach: that's nice. Yeah. I think humidity yesterday when I checked was like high fifties, low sixties, which is like again like a reprieve for us here in Charleston. We're like it felt nice.

Jay Nathan: Yeah, that's a good it's nice break.

Jeff Breunsbach: Yeah, it felt a little crisp yesterday morning.

Jay Nathan: Mm-hmm.

Jeff Breunsbach: all right. Well, I think what an interesting place. I just was showing you offline a a personal tool project. You know, I'm in a NFL survivor pool and decided to use Claude to help me build a simple one off little tool that I could use for the eighteen week NFL season. And I think it reminded you of an article that you were reading about, kind of a concept of of I think it was Benedict Evans. Is that who it was?

Jay Nathan: Yeah, Benedict Evans, and I I actually I can't even quote the article. I'll I might pull it up while we're while we're talking here. But the the thing that I he was sort of talking about, you know, this whole idea of I need to pull up the article again to be honest with you, but the the thread that I took off of it was what we would have historically done. You probably would have built a spreadsheet to solve the problem that you solved for your fantasy football league, right? you might have used a no-code platform would be an evolution of that. Then, you know, somebody might have actually seen what you did, or you might have done this yourself, built a little SaaS point solution tool for that. And then somebody else may have bundled that point solution in with like five or six other apps for fantasy football enthusiasts, and it would be a fantasy football platform. And I think that's like the evolution of how software has sort of been built over time.

Jeff Breunsbach: Yeah.

Jay Nathan: But now, and some of the, I mean, there's there's some obvious advantages that you showed me in this app, that sort of pulled in outside data sets that you didn't even know existed, right? The LLMs just went and found it. so you sort of skipped the spreadsheet and no code. You built the the point solution for yourself. It's a personal point solution. But it got me thinking, it's sort of tied tied to that article in my mind because you know, what is going to happen? There's still I think we're still in like the first 25% of the AI boom, right? It feels like, man, if you're not in now, you're late. You're not. I don't think so. Like you could you could pick this stuff up at any time. And in fact, like three months from now, it'll be totally different and new. So like doesn't matter if you're along for the ride right now or not. I wouldn't I'm not discouraging people from getting involved with it. But the point is the the proliferation of software is going to be profound because it can be generated so quickly. And what you were just talking about was interesting in response to that. And maybe I think you have a couple of examples to share of what you've built. So what what this means, and you you said this, I'll try to paraphrase what you said. What we've done historically in our companies is we've taken software off the shelf, right? These point solutions or these platforms that were built to generally solve a problem. We've put them into our organization. And then we've sort of bent our business processes to those solutions, which is was a good thing in some cases because some companies didn't have process, right? Or have the appropriate process. They were doing things their own way. may not have been the best way, but they were the only way they knew how. And the software sort of brought those best practices with it. But now what you can do, I I think what we will see is you'll have many more options for those solutions now, right? You'll have A bunch of software companies that can provide those solutions tailored to little smaller niches within that category, right? So we were talking about customer success software, right? But maybe there's customer success for high-touch organizations, customer success platform for low touch organizations, CS platform for hybrid organizations. That that could actually happen. There's a question there as to how many could be viable in a category. I don't actually have data on that. But anyway, your your little, you know, fantasy football tool got me thinking about that thread and and like what does it look like to to s sort of not build in house but have a lot more software vendors who can actually build quality technology now.

Jeff Breunsbach: Well it makes you I think it I guess there's two two things that come out for me. I the first question that I would probably pose to you is like, do you think then that most of these platforms or point solutions I guess need to find ways to more to allow their platforms to be more customized? Like in the in the advent of AI, do they essentially need to I guess open up the aperture for customers to configure and customize the platform? So that again, like the the benefit to the customer, right, is I have a a platform that has been kind of tried and true. It's tested, it's got data and privacy and you know, has all these workflows and things built in. And there's a lot of goodness there. But like you said, maybe I've had to kind of round the edges of my business to fit into what this is. And I wonder if if now these platforms in order in order to compete, right? Like, yes, you still have to solve the problem of the majority, but I guess to keep the minority, like you're you might, you might have to start, you know. figuring and maybe I mean we've always kind of done this, right? Like there's been, I don't know, there's been the top 10 customers that you have that generally have a little bit more availability on the product roadmap and the product team's a little more attentive to like what they want. But I wonder now if that that truly becomes almost like another step function up where it's like, you know, yeah, we used to make sure and appease some of these top 10 customers, but we actually now need to like almost, you know, a allow them kind of like a a s a sandbox so to speak of like, hey, they should be able to kind of c customize and configure these parts of the business or these parts of the platform because otherwise they're gonna either go build a solution themselves or they're gonna go find a vendor that will, you know, because of of the advent of AI, how how easy it is to push out code and be able to, you know, make changes, it it feels like that's almost like a necess it's going to become a necessary, it feels like.

Jay Nathan: Well, isn't I mean, isn't that the whole idea of it being agentic in nature is that it's like naturally customizable. I needed to make decisions based on the way we make decisions, right? And and it's

Jeff Breunsbach: Yeah.

Jay Nathan: doing that based on input and history as opposed to like some rigid workflow that always works the same way 'cause it's been configured that way for a certain type of, you know, company at a certain scale.

Jeff Breunsbach: Yeah.

Jay Nathan: So I I mean I think the answer to your question is an absolute yes. Right, but it that that's what agentix software is in my mind, is like a hundred percent customizable technology. Hundred percent, yeah, maybe a stretch, but

Jeff Breunsbach: Yeah. The and I think the second thing that this just leads me to is again, I I think the power is in the hands of a business person that can also understand like what to do with with AI and technology. So some examples that we've done recently. our our sales team, you know, we essentially a part of our sales process is you know, customers might be inquiring about our pricing that we get with the labs. and so that would require us to understand what tests they might want to price out. and again, you think about this in the past, right? Okay, we get a list of what we call markers and IDs from the customer. salesperson takes those in, it's in a spreadsheet, it's an email. we go and probably ask, you know, somebody on the product team or somebody on the engineering team, like, hey, can you can you bump this up against our pricing database? And then, you know. They get to it when they get to it and it kind of comes back. And like there's just this long loop of like, okay, it's it's not real time, it's not helpful, right? It's it's just kind of like we're all just passing things along in order for it to get back to the customer. and so one of our our one of our salespeople went, took took a kind of Vercell solution or Vercel app that they built. currently it points to a spreadsheet that's updated with all of our pricing. this is before, you know, we're we're kind of getting the blessing of our engineering team to hook it up to the database. But but what they proved is in a very small MVP solution is cool. Salesperson can now come in and input a bunch of marker IDs. It goes and validates those marker IDs, looks up those labs, looks up those tests and those prices, and comes back with a nicely formatted, branded little PDF that you can basically download, send back to the customer and say, Hey, here's some, here's some of the pricing that you're inquiring about, and here's what our price is. And so, you know, now the last part is hooking that up to the database. So it's truly, you know, kind of live to like what our true pricing is. But I guess to like to your point, right? Like that little point solution now has saved multiple back and forths with our internal team and internal steps, gets a gets a faster answer to the customer and puts you know more power back into the hands of the the AE and the sales team and the CSMs are going to use this, right? Like this is a a tool that, but again, I think. I guess where I'm going with this in my mind is it feels like there's a lot of internal processes that have always been on the back burner, that have always been afterthoughts, right? like,

Jay Nathan: Yes. Yep.

Jeff Breunsbach: the the engineering team will get to that when they get to it. But it's more important for us to focus on what's truly in the platform for our customers. even though that's actually a a large friction point, right? Is like our customers want to know pricing, our our prospects want to know pricing. We've just never thought of it as a, you know, as a large enough problem. It's never been a large enough problem for our engineering team to spend cycles on. And so that's like an example. Cool. We can actually build this solution ourselves. the second one is a very similar use case, but we have to go price out what's called test kits. this is something that can be sent to your home in order to to do a blood draw. so you know, there's a tasso kit, which is something that goes on your arm, you press a button, it blood draws the blood for you. It's meant to be better than pricking your finger and doing ADX. anyway, so this, you know, has this has logistics involved, it's got shipping back and forth. there's a component, right, that is just variable in nature to this. and so we've again historically managed this through spreadsheets and pricing and different spreadsheets. And then we have external versions, internal versions. You know, you've got all this stuff. and so our one of our operations leaders who heads you know heads up our test kits went and built again, a very similar pricing solution. It's a little different than the labs piece. this one is more specific to test kits. And now you can see in this application that she built, you know, pretty clearly, okay, here's the here's the markers that they're trying to test for, here's the prices of those. Then it gets into the logistics. Here's the shipping speeds back and forth, here's the cost of that shipping speed back and forth. but you can basically see this all kind of laid out in a way internally that's really helpful for us to understand. Okay, here's how this basically got built. And then you've got kind of sharing internally, sharing external, externally buttons. it's it like kind of downloads it, keeps a log of this stuff. you input a customer, like there's just, and again, this is kind of MVP early innings, like you said. But to me, I guess like the power of being able to customize solutions for our business because we have these unique aspects of it, is like pretty cool. And I think like we're again starting to see some of the early pieces. I think what's interesting right now, and I think something that I'm Preaching is that these and and what we're doing is is building a platform that we can access all of these tools from a single place. Right. Like what the challenge it right now they kind of live in little Vercell apps, but we're pushing them

Jay Nathan: Yeah.

Jeff Breunsbach: all towards a internal solution so that again, I log in and now I've got access to a suite of tools that everybody in the business theoretically has access to to do the same things. and that way again, we're kind of pushing everyone back into a single place versus okay, let me which which website do I need to remember to go to this pricing tool or this test kit thing? and then

Jay Nathan: Yeah.

Jeff Breunsbach: I'm switching context eight times a day, right? Like, no, we should figure out ways that this, you know, essentially all comes back together and says, Okay, this is this is internal processes and this is like an internal app that or internal platform that we're building for all of our processes.

Jay Nathan: Yeah, okay. So lot lot to un unpack there. you said the power is in the hands of the business person who's knows what AI can do. And that is absolutely true. I was just having a conversation with one of our FDEs this morning. So we had a this is all moving toward a different conversation I wanna have with you about like what we're building next and for our own for our own purposes and for our clients now. But I was just having a conversation with him this morning about you know, the limitation here is not going to be the technology is like not the limitation at all, right? We know that, but the the real limitation is the how creative can you get with the questions that you want to try to answer. we have a hospitality client and we work on their publicly facing website and bookings engine and loyalty. program. I'll just leave it at that. And one of the things we were doing over the weekend is sort of playing with agents and figuring out like You know, can we go book what does agentic booking look like on their website? Is their website actually optimized? Not for a human, but for an agent to go book? And what would we do to their website if it wasn't? And all we did was just ask the agent, like, hey, tell us what would make this easier for you. by the way, could you actually find this booking website? And how did it stack up against the others in this particular area where we serve this client? And so like, God, like we're gonna build agents for this customer that give the optimization recommendations and then also track competitors that you know sort of build insights on a weekly basis for these teams. And and we're gonna do that. That like that's the creative part of

Jeff Breunsbach: Yeah.

Jay Nathan: what needs to be done here, right? Like the things you would never have done before because they would have been manual, you will now do right to your point. Now the here's where I would take what you said and and go one step further with it. So I've been playing with Instinct and Town. Have you heard of these tools yet?

Jeff Breunsbach: I think I've heard of town, which is essentially you can create it's all like a town you create a townie, which is like your basically your agent. Yeah. Okay.

Jay Nathan: Yes. Yeah. It's basically it's a personal agent. It lives in your inbox. Okay. Instinct is interesting, similar concept, but instinct, you start, you engage it through iMessage, like a text, and then you engage with it there. That's where where you're working with it. So instinct feels a little bit more like my personal life and bre I almost called it Bex because my Tangie is name her name's Bex. But town is more of like the the business one for me at least, because that's where all my email is. So it lives in your inbox and you sort of communicate back and forth with it. So what I think is the most important thing that we could be doing for our clients right now, and as we build agents for ourselves, is communicating with them in the environment that we already spend the most time in. Because there are only so many applications that That we can handle going to. And why should we have to go to more applications? So I think where this is all going is you will have agents, like the the two tools that you just mentioned, the lab pricing and test kit pricing, those are those are either tasks within an agent that serves your entire company, like a team based agent. One that has an email address, right? It's like, I don't know, call it junction at junction.com, right? Or whatever give it a name. And it can answer all these questions based on all the tools that you've built in the background. Sure, you could go to the website too, the the actual Vercel app that you've built if you need to review data, if you need to do something unique. But there should be an API on top of that as well that's connected into your agent and you can communicate with that piece back and forth without having to go to another tool. You just do it from your inbox. You just do it from Slack or Teams or Discord or whatever you use. So that's the kick I'm on right now. It's like we're not going to more places. We're going to fewer.

Jeff Breunsbach: Well, so this is so this is interesting too. So I talked about the wiki that we're building, like our internal knowledge base. and the thing that my solutions engineer and I keep talking about is that we are building this for agents first. And it's really hard because I look at the interface that we've built and I'm like, we should we should optimize the look and feel, we should change this and that. And in my mind, I'm I'm like almost like reverting back to like, my team's gonna be in here searching for things. And it's like, no,

Jay Nathan: Yeah.

Jeff Breunsbach: like. This th it's really a that's really a secondary part. Like, yes, we want them to be able to access it if they want to go curiously look at something. I want them to be able to do that. But the whole reason we built this as a wiki that lives in a GitHub repo that's got MD files and all this stuff, right? Is it is like readable from AI and agents first. And that is the whole purpose, right? Like you want them to be able to go find the information as quickly, efficiently, and as effectively as possible. And so like I guess putting all this additional time into Let me optimize the look and feel of this thing. Let me like over, you know, over engineer kind of like the visual aspects. You know, we've essentially like had to almost like catch ourselves every time being like, no, no, no, this is like minimalist. This is like bare minimum that the team could go search for something if they want to. It's gotta be thinking, Okay, yeah, is this is this still optimized for the agents? And like is this structure still there? Is that you know, do we have all the kind of write permissioning and files and instructions up front for each of the agents and and like AI to understand on the way in.

Jay Nathan: Yeah, I th I think system systems of record will be headless. And what you're talking about, your wiki is a s is a system of record. Right. that is really cool. so yeah, so what do you guys so are do you have like APIs built into those of your cell apps? Or are you putting MCP servers in front of the like how are you? connecting them. We're right now is it just a collection of apps that are gonna live in the same environment.

Jeff Breunsbach: Right now it's a collection of apps, but I think like that is the next. I think to your point, before we keep stacking apps on top of each other, I think we need to go through. We're kind of doing it on this wiki, but like we need to go do it on the other apps as well, is to is to go think through, okay, like what's the access point for agents and AI to interact with this. the thing that is interesting that you mentioned is I keep thinking about agents as as almost like specialized doing you know, like an agent that's great at renewals and that means that, you know, again, they're they're doing the right things for research and putting together the proposal and negotiation tactics and right, like that's I guess like I'm envisioning the agent doing that with a a a lot of information that we're able in context that we're able to give it. But what's also interesting that you just mentioned, right, is like that still feels like there still has to almost be another entry point for the team, right? Like I shouldn't have to remember that this is the renewal agent or call the renewal agent specifically. It's like I should go call my townie or my my, you know, my app and say, Hey, I got an early renewal opportunity, just got off this call, you know, here's the here's the notes. Can you go spin this up with the other agent and have it, you know, coordinate for you? but yeah, this whole I guess like the whole notion, 'cause you think about GrocBot as like another one that just came out that people have been enamored with. and it again, I think this all kind of continues to go, I guess, to evolve back into the same discussion of like shared knowledge and context, making sure people have access to the right like making sure that you have access to the right things. And so I'm curious, I guess, like, do you see like the towny example, the grock bots, those still all feel very, very personal? Like, okay, I'm gonna hook it up to

Jay Nathan: Yeah.

Jeff Breunsbach: my I'm gonna hook it up to my personal email or I'm gonna, you know, but I guess like when I think about Businesses adopting these agents. like this reminds me of is it Spotify that has river? Have you heard of this?

Jay Nathan: I've heard of some Spotify something, but no.

Jeff Breunsbach: So they they have like they they've I you know I guess Toby from Spotify is is very early on the agent, you know, an AI train. So I think this is now running for seven or eight months. They have built an AI agent called River that can be called in any Slack channel. and essentially now is like almost like the creme de la creme. but I guess they're it it looks like what they're doing is solving it. almost like everyone has access to river and river can go do all these things versus each individual person having their own individual agents. And so I wonder like if that's I guess like I guess where do you see it going? Do you see the okay, each of my team needs to have their own agents. 'cause again, now that I'm th saying this out loud, then I start thinking about okay, if I've I've got, you know, ten team members on my team, that means I have ten agents that we have to kind of monitor and upkeep, make sure that's got the right context and instructions and files. Or is it one agent that you essentially have that's central but can coordinate with all the other agents, like for everybody.

Jay Nathan: Yeah, okay. So this is a really interesting question. and I will say this. I was listening to a podcast just this morning. I'm not even done with it yet. On 20 BC, the CEO of town, the founder of town, was on there talking about this very thing. And I think what'll happen is so the reason you would have different agents is because the Context for any given task is is is different, requires different guardrails, requires different inputs. And if one agent had to have all of those inputs, it would be probably wildly inefficient, right? It's like using an LLM that understands the entire universe to do every task. You don't actually need to do that. so I think there is one almost like meta-agent. In fact, like in our system that we've built, we actually call it meta agent, but it's it's the agent that gets That gets called and then delegates to other agents in the system when there's a renewal discussion or when there's a coaching discussion for our team or when there's a plan. Different agents have different purposes and they have different contexts. Not to sort of compare them to humans, but sort of the same way that you would you would have people that know different things, right? Have different skill sets that have access to different data sets within your organization. Same thing. Now, I also think there are individual Contexts and I didn't say agents there, but there are individual contexts that could be loaded up in a person's agent. So, like if the central agent in this case, like River, or in our case, we call it Hal, like if it didn't have the context to answer the question, it should have enough context to say, hey, Jeff knows about this because I've seen. this discussion in his inbox or in his Slack channels. So I'm gonna go query him. I'm gonna go communicate with with the Jeff agent, right?

Jeff Breunsbach: Yeah.

Jay Nathan: That I've spun up with Jeff's context and am able to ask him questions. That way I don't have to keep all that context in my memory. I could just go delegate that off to Jeff and get see if I can find an answer. Here's what Jeff knew. So I do think that will begin to happen. And I think some of these, you know, like like town as town has got a very viral component to it because what you're gonna want to do is you're like, okay, Jay just sent me an email from town. I and I work in the same company. I need to go get town because that thing actually sounded really good. It it it was

Jeff Breunsbach: Yeah.

Jay Nathan: he answered the question perfectly. I could tell he didn't write it himself, but I got what I needed, right? So now I'm gonna go do the same. And then if I've got people within my domain, within my company That are all using the same thing, then I can start to build that up naturally. So you've got an agent, I've got an agent. We've also got a company level agent that's because I'll just say one more thing. My towny in town, that's what they could like to your point, what they're called, is Bex at town.com. That's a single email address. There are probably tens of thousands of people that have Bex at Town.com as their the communication point for their agent. Right. So there is one central agent. Bex is one central agent. And as long as Bex can can keep my scope limited to just me or keep my company's scope limited to just people in my domain on my account, then you've got your answer, right? Bex is the equivalent of the meta agent. And then you and I have agents that are able to answer specific questions to us, and you could probably go create more agents as well that that have different contexts. So that was

Jeff Breunsbach: So

Jay Nathan: long winded, but

Jeff Breunsbach: no, it's helpful. and I like the way I like to hear how you're thinking about this. I misspoke earlier. It wasn't Spotify, it's Shopify. I keep getting those mixed up. But but Toby,

Jay Nathan: Close enough.

Jeff Breunsbach: Toby from Shopify is the one who's been like early on this train. So he wrote an article. This is actually now back in I mean, this is wildly outdated. It's in May. So in May, it actually reminds me, so he wrote this article called Learning on the Shop Floor. And it reminds me of management by wandering around with the Hewlett Packard. I think it's yep,

Jay Nathan: Yeah, HP guys. Yeah.

Jeff Breunsbach: who it was. And so I'll read this real there's two quick snippets, but I sent you the article. I think you should read it because I I it's actually kind of interesting. So he so it's called River. River is an AI agent that lives in our company Slack. You talk to her the same way you talk to a teammate by mentioning River in a Slack channel you can read she can read code, run tests, write code, open pull requests, query database, blah, blah, blah. In the last 30 days, so this is this was authored In May, May 9th. So in April and May, 5,9838 Shopify employees worked with River across 4,450 different Slack channels. It opened 1,870 pull requests in the last week alone in our main mono repo. about one in eight pull requests merged into our code base last week was authored by River, reviewed by us, and she only works in the open. You cannot call her in a private Slack channel. And so the other reason I thought this was interesting is he goes on in this article to say, so I work with River in a channel called Toby underscore River. And many people have followed my pattern, but every conversation is therefore searchable. Anyone at Shopify could jump into my own channel. and there are a hundred over a hundred people who react to threads, add color, add context, pick up the torch, help with reviews, all from like a channel that I'm working on with River. but they this also reminds me, I think they're the ones that also have a they're they have a vendetta about private Slack channels. They have a scoreboard where they're trying

Jay Nathan: Yeah, that's right.

Jeff Breunsbach: to get, you know, people to because again, they they see it all as context and knowledge, right? Like if it, if it's out in the open and searchable. So, anyways, so yeah, I I think this the whole, you know, shift into agents is going to be really interesting to figure out. Like you said, it it feels it it feels like that's the next, I guess, layer of this AI cake of like, okay, you know, how do we get our teams working in the the spaces that they're usually they're used to. but making sure it's got all the right context, all the right permissioning, all the right, you know, kind of everything, the durability that comes with like an enterprise solution, making sure all that stuff exists so that the agent's as effective as possible.

Jay Nathan: So so ha okay, you got Shopify doing this. Two questions for you. Number one is how how do you think it actually changes the way that work looks inside of Shopify? And then second question is, how many companies do you actually think are working like this out of all the companies in the world?

Jeff Breunsbach: Well, so this is so the okay, I'm gonna answer the second or the second one first, I think is more interesting because because right, I I think there's always the cliche that stuff comes from the top down, right? That like basically your team models the behavior that you have. And I think like wow, it's pretty powerful. If you actually go read, I'm he he's one of the top five follow-s me. Like I almost like go to his Twitter account every morning because I think it's it's pretty interesting. but I mean he for I think he's been beating this drum of AI for years. And I think Him being it like I think a former engineer himself like jumping into it and now he's involved with Omachi and like

Jay Nathan: Yeah.

Jeff Breunsbach: you know like I think it's pretty interesting that you have maybe somebody that is so ingrained in like modeling the behavior that like it's almost like it forces the entire again, think about he said five over five thousand employees have engaged with River, and this is back in May. I mean, just think about like I guess think about the change management of 5,000 people and how, you know, people are hemming and hawing about like, okay, can I give ri you know, can I give AI access to my email or something? You know, you have some you have some companies

Jay Nathan: Yeah.

Jeff Breunsbach: that are over there thinking about, okay, what permissioning can we give it, whatever, right? Like, like clearly he's been pushing this so far that he he's got, I I would imagine there's probably a a team of people that are you know that essentially are charged with like, how do we spin this up and move as quickly as possible on these types of internal solutions and If he's driving that, you know, like that's a I don't know, that's a pretty big bias for action when the CEO's there, you know, on Twitter talking about AI every single day, using it himself, coding himself. I think like if you go look at his GitHub, I think people have been talking about how he's like this is the most active he's been on like actual coding things himself in like years. So I think like all that modeling behavior is pretty interesting to your second question.

Jay Nathan: What is his what is his Twitter handle?

Jeff Breunsbach: Twitter handle is

Jay Nathan: Can't find him for some reason. Maybe I should just Google it.

Jeff Breunsbach: just at Toby, T O B I.

Jay Nathan: that's killer. I found it. Okay.

Jeff Breunsbach: So that's so I think like that's an interesting one. You know, I guess like to your point about like how many companies are doing this. I don't think there's that many. I mean I think but I do think of the ones I guess the ones that are like biased to act to biased to action and like around AI and stuff seem to be the ones where you have a you have a executive or a leadership team that is like is highly convicted of it, and highly convinced of it and they're, you know, putting the action behind it and that's

Jay Nathan: Yeah.

Jeff Breunsbach: the first mover advantage almost. The and then to your other, you know, what how does work change? Like, I think this is the this is the more interesting part, right? Is now you start thinking about, okay, my work, even as a customer success leader, is going to change, right? And you have to be okay with understanding that that your job is going to change and that you need to figure out what is the new way of of working look like. so I mean, I I think engineering, it's maybe been more clear about how it's changed, right? It it feels I saw somebody the other day put it as like tactical programming seems to be like something that AI should be doing almost exclusively. And then strategic programming should be something that like the engineers are worried about. So I guess they they split that up as like the tactical programming is like actually writing the code. The strategic programming is more about the architecture and the layout and like the like what is the machine and the system look like, and spending more time on that. So it feels like engineering, maybe that's been more clear than maybe other roles about how like, okay, maybe I'm I'm coding less but reviewing more. I'm thinking more about the overall architecture, the bigger picture. How does this stuff fit in if I if we start expanding on things? and so it'll be interesting to see how that plays, I think, with other other departments, other roles.

Jay Nathan: Well, I think one of the things you've already seen in engineering teams is that the middle manager is going the way of the dodo because like if I've got if I've got agents giving me more real time feedback, I've got leaders sort of at the top of the food chain who can see more about what's going on and have it have it sort of summarized for them. Like, why do I need so many middle managers? And I I I think that's a thing in engineering. It certainly is in product. Everybody's a builder now. You can't be hands-off like pushing paper, so to speak. I think you know, we'll we're starting to see that in in other teams as well. I mean, we see customer success teams getting bigger and bigger, honestly. Well, I should say individual teams getting bigger and bigger. Customer success as a as a function is probably getting smaller. But that is also

Jeff Breunsbach: Yeah.

Jay Nathan: an interesting point, right? It's like, no, the expectation is we're not going to get a 10x return on this ARR anymore. We're going to get a 4x return on it. So you can't have, you know, named coverage for 50 accounts. You can have named coverage for 150 accounts per CSM, and they're going to be expected to know what's going on with those 150 accounts. It's the same phenomenon, right? It's move up the stack, use your tools, do more with more.

Jeff Breunsbach: Yeah. well cool. I think it's good good kind of ending point for for today's episode. But I think this this evolution of AI agents. so one one other thing I didn't mention is personally, I think I've talked about this a while ago. I set up a little home chef agent for my wife

Jay Nathan: Mm-hmm.

Jeff Breunsbach: and and we've been using it, it's been you know slightly working, but again I think the the limitation has been I have to go query claud and like get it to run it. It's not something my wife can text and easily do today. And so I've seen these agents pop up more, but I w I'm I might go start as like a first use case of like I forget what's the one other one you called out that you can iMessage or text message? Instinct.

Jay Nathan: Instinct. Instinct dot co. Yeah, there's a wait list to get on it, but it wasn't long for me. So

Jeff Breunsbach: Yeah. Cool.

Jay Nathan: actually I can invite you now. I think that now that I have a have an account. So

Jeff Breunsbach: okay. What a viral component.

Jay Nathan: yeah, yeah, exactly. So all right, cool.

Jeff Breunsbach: Cool. all right. Well we'll wrap it up for this week and we'll we'll we'll be here again next week, you know, talking more AI and agents.

Jay Nathan: All right. Have a good week. See ya.

Jeff Breunsbach: All right, see you man.