Why Enterprise AI Agents Struggle to Reach Production with Open Source Pioneer Ben Sabrin

0:00Ben Sabrin: You wouldn't think that you'd have to tell an agent not to delete a company's source code if they're a developer of software. But that's actually happened. Yeah. So in this instance, you actually have to bring, you have to codify human behavior on a decision a human would have to make. Like somehow, how do you code in there that somebody's got to have a shaky finger before they press delete all versus the agent world? They don't have that experience of shaking. Are we sure we want to delete all? Welcome to the Revenue Builders podcast, a weekly show featuring B2B sales leaders and executives. Hosted by five-time CRO John McMahon and Force Management co-founder John Kaplan, the show takes guests in the barrel, behind the scenes with the people who've been there, done that, and seen the results.

0:53Ben Sabrin: Revenue Builders covers best practices for scaling and growing your business while sharing the pitfalls to avoid. Welcome back to another episode of the Revenue Builders podcast. I'm John Kaplan with John McMahon. Today's guest is Ben Sabrin, a go-to-market leader at Arcade, a company solving one of the most overlooked problems in enterprise AI, agent authorization and security. Something like 70 % of the agent work companies want to do today, they can't because it isn't secure enough. That's the problem Ben's company exists to fix. Ben spent his career building go-to-market before the category even existed.

1:38Ben Sabrin: He was part of the original crew at JBoss, one of the first successful open source companies, then became the original VP of sales at MongoDB, building that sales org from zero to 40 people in less than two years. A lot of the biggest customers his team signed back then are still MongoDB's biggest customers today. We get into what it actually takes to sell platform instead of a product. How he thinks about repeatability at the earliest stage of a company and how he's running his own team today using AI, not as a buzzword, but as the operating system. Let's take a listen.

2:24Ben Sabrin: so ben uh i we had a chance to speak a few weeks ago on a little prep session and i uh told you that i mean for for the audience i read a very compelling article in the wall street journal about the raise of 60 million dollars and the article was like the raise arcade raises 60 million to secure AI agents and I remember I was on my fat bike I call it my fat bike I ride a little recumbent bike in the morning

2:58Revenue Builders Hosts: it's actually not a fat technically not a fat bike

3:02Ben Sabrin: it's not a fat bike but I call it a fat bike you know what a fat bike is

3:06Revenue Builders Hosts: with the big fat tires

3:08Ben Sabrin: exactly are you done correcting me bro

3:12Revenue Builders Hosts: I'm just trying to help the audience so they get a picture of you on your little recumbent bike and not outside on a fat bike.

3:19Ben Sabrin: No, but now if they know me, they understand why I called a fat bike. So the thing that really struck me when I read the article, and I want you to kind of educate us, I don't get how the world even exists, the AI world even exists right now without this concept of securing AI agents. I was just working with something earlier today and it wanting access to my personal credentials and to these other applications. And would you do us a favor and just kind of get us level set on what Arcade is, why it's so groundbreaking, and then we can talk about what's led up to that and what some of the nuances are.

4:04Revenue Builders Hosts: Thanks, John. Good question. So look, Arcade fundamentally answers the question, can this agent take this action across these resources at this time? And it doesn't just answer the question. It actually serves the tools that the agent is trying to serve to the large language model, right? And we remove the ability for a credential to ever be passed to the large language model. And we are providing a set of deterministic rules that say this agent can, let's say, send an email, but it can't delete an email. Or it can update a database, but it can't delete a database. And you have to get down to the individual action that the agent is trying to take to have the confidence to let agents take action.

4:58Revenue Builders Hosts: And that's really where we need to get to as an industry. If we're going to start paying off the debt of the investments that people have been making in AI, we can't just have AI doing research. These agents have to start taking action. But if you're a regulated industry and you can't answer that fundamental question of who authorized this agent to take this action at this time across these resources, the agent's not going into production.

5:26Ben Sabrin: But even outside of regulated, like the reason why this struck me was I can't think of a scenario where this isn't a relevant conversation. I just I can't think of one. And even in my simple world, it comes up almost every day.

5:42Revenue Builders Hosts: Yeah, it's a very relevant conversation. And what's cascaded the conversation is we've seen a standard emerge faster than probably most in history, which is MCP, right? So MCP, which is the model context protocol that was started by Anthropic.

5:59Ben Sabrin: Give us a little tutorial on there.

6:02Revenue Builders Hosts: Yeah, so MCP was originally developed by Anthropic, and it was developed as a protocol that allowed a model to communicate with tools, right? And that tool could be Salesforce. It could be Gmail. It could be Outlook. It could be your Snowflake database, your MongoDB Atlas, et cetera, et cetera, et cetera, right? And that standard, initially Anthropic built it themselves, but it was always open source from the beginning. And then eventually they donated it to the Linux Foundation. And it has grown faster than I think anything I've ever seen, right? And now it's pretty much become the predominant software delivery mechanism.

6:54Revenue Builders Hosts: So if you're a software company in today's world of AI and you're not exposing your software via MCP to make it easy for somebody working inside of ChatGPT or Claude or Cowork or Copilot or Cursor, et cetera, et cetera, then you're likely losing share or minimum mind share for your users. because that's the way increasingly people want to work. Now, I could be living in my own little kind of echo chamber that is Silicon Valley and the circle in which I operate, right? But it's becoming increasingly common that people want to start living in a world outside of the browser. And they're using their LLM client as their main interface Because that actually is the biggest innovation that's happening in the world of AI right now, in Gen AI, is the interface to computing is changing, right?

7:58Revenue Builders Hosts: Gone are the days that a user has to live within inside the confines of the interface that you dictate, right? I can express intention and my intention can act across all of the systems that I have access to, right? Now, if you have a really good access graph for who should be able to do what and what actions people can take, you can truly enable that. And that is really what Arcade does. We are an enabler of that reality, of how do you change the way people work across lots of different systems. So do you go into a company then, Ben, and do an analysis of where they're utilizing agents, what systems they're talking to, and try to put in specific settings that you talked about earlier to try to essentially regulate the agent?

8:54Revenue Builders Hosts: Yeah, so we spend a lot of time. We have multiple different entry points, which is a blessing and a curse, right? You know, as you know, I know, John, you guys are great at helping people dial in message and focus, right? And when you have multiple different potential buyers inside an account, in some cases, it creates a little added complexity, especially when you're an early stage company and resources are somewhat finite. But it could be security, right? So security folks care about this problem in some ways, but typically it's security folks that view their job as they need to pave roads for their constituents versus put roadblocks in place.

9:46Revenue Builders Hosts: The ones that are putting roadblocks in place are going to try to solve the problem that Arcade solves likely at the network level. They're just going to block everything, right? And that'll give them the confidence. Or there's the platform teams, right? So historically, if you think about most enterprise software, even early Mongo, we were going and selling to enterprise architects, the enterprise data architects, people that drove that platform shift and that change. And what we're seeing is a lot of companies have an AI platform team, especially the ones that are a little further ahead. And this is a team that's responsible for how do I enable the organization to start to leverage AI in a predictable manner across the business?

10:35Revenue Builders Hosts: And a lot of times in that initial engagement, there is a bunch of discovery that we need to do in terms of like, what are they using today? Are they building custom agents? Are they just, you know, leveraging connectors inside of Copilot or Claude or something along those lines? In a funny way, it's almost like the old RBAC for an agent. Yeah, 100%.

11:02Ben Sabrin: What do you mean by that, Johnny?

11:03Revenue Builders Hosts: Role-based access control. I mean, that's what you typically did with humans. You'd say, okay, John Kaplan can do these types of things because of his role in the company. And these are specific systems, as Ben said. You know, what actions can Kaplan take across which systems? And Ben's saying the same thing for the agent. You know, which actions can they take at this time across these systems? And we sort of - That's all RBAC was really doing. But everybody wanted RBAC almost for any application that you had because depending upon the information in the company, you only wanted certain people in certain roles to access certain information at specific times.

11:47Revenue Builders Hosts: And now, like, add in the fact that this quote-unquote user, right, is this non-deterministic, non-human, very intelligent resource that doesn't like guardrails, right? As we've seen, you know, stories that continuously come out over the last several weeks, right? Every day. You know, oh, our agents broke the guardrails again. That's because you're putting the control. When you put the control inside of the agent, right, they're not going to necessarily obey, right, those controls because they want to win. They want to satisfy the intention that was asked of them, right? And they're very, very smart.

12:42Revenue Builders Hosts: And they're getting smarter by the day. But it is interesting, you know, John, you made a really good point around RBAC, right? So I think one of the other keys to anytime you have a platform shift, right, there's always a, there's never a clean break from the past, right? You always have to have a bridge to how things happen, especially in the enterprise. We all know enterprise, they don't just say, okay, we're doing everything different starting tomorrow. It just doesn't happen that way. And one of the things that we made sure to be able to do was be able to meet people where they are. So in a lot of cases, most companies have an entitlement system.

13:31Revenue Builders Hosts: They have a source of truth, which is feeding multiple different systems within the organization. It could be a database entitlement system that's saying who can do what within different data structures. There's your application entitlement system that could live in a sale point or another identity governance system or ENTRA, etc. Arcade is capable of pulling all of these things in and being a policy enforcement point inside of the agent loop. And we did that very, very intentionally because we believe that if we walked into an enterprise and we said, in order for agents to do anything meaningful, you've got to take all of this work that you've done around governance and entitlements and you've got to completely rebuild it in the name of an agent, it would have been dead on arrival.

14:21Revenue Builders Hosts: Right. Right. Yeah. So but what you know, and we shied away from from like having our own distinct policy language. But we're being pulled in that direction because the reality is, is people don't fully understand all the policies they need for agents. And the way they want to express those policies are much more in natural language than how they may have expressed them in the past.

14:54Ben Sabrin: You wouldn't think that you'd have to tell an agent not to delete a company's source code if they're a developer of software. But that's actually happened. Yeah. So in this instance, you actually have to bring, you have to codify human behavior on a decision a human would have to make. Like somehow, how do you code in there that somebody's got to have a shaky finger before they press delete all versus the agent world? They don't have that experience of shaking. Are we sure we want to delete all? So when you're doing that inside of companies and you're trying to figure out how to meet people where they are, are you doing this with like forward deployed engineers?

15:38Ben Sabrin: or tell us a little bit about that because we're hearing a lot about, you know, AI companies are now using these principles of forward deployed engineer. When Johnny and I were laughing a little bit in talking about the old days, what that used to mean. What did you call it, Johnny? What did you mean? Your software doesn't work. What it used to mean. You're essentially going in and writing code on the front

16:02Revenue Builders Hosts: and making sure that, you know, when you do do a POV, it actually works. but you know it's different it's different times these days so it's codifying not only that

16:12Ben Sabrin: selling software that mostly works these days versus even 10 15 years ago we were selling software that didn't work but the difference is now that people are like they're expecting it like i don't want it to work off the shelf like i want it to work in my environment yeah i wanted to make decisions like we make decisions so talk a little bit about how you how you do that in this world today.

16:35Revenue Builders Hosts: Yeah, no, look, I think you both make great points. And I think the reality is that we have a forward deployed team, and they build stuff on a very regular basis. And a lot of their purpose and their mantra is, you know, because, I mean, Arcade, Arcade is a middleware platform, right? So by itself, and when you sell like a middleware platform or a database, right, Your product by itself does nothing unless somebody builds something on top of it. And so the forward deployed motion really lends itself very well in that capacity because we're providing building blocks. And because the market, there's so much white space because at the end of the day, I think we are collectively experimenting as an industry.

17:29Revenue Builders Hosts: right and there's so much people see they see the potential and they see the possibility and so there's this thirst for like how far can we push it um and so we spend a lot of time and you know and yeah there's times where it may be a scenario where our software doesn't work the way we want it to um and there's some stuff that needs to be cleaned up but but in a lot of cases it's really more of, well, it doesn't do this in the way that the customer wants it to, right? And it's not that our software doesn't work. It just, it needs to be customized for their environment. And that could be because somebody has a homegrown Kerberos-based entitlement system, right, that we need to, that we need to hook into.

18:16Revenue Builders Hosts: But what I really hear you saying in here, correct me if I'm wrong, is like there's so much opportunity, especially for a company like yours and many other AI native companies that when you go into an account, you have to figure out where can I give the customer the biggest bang for the buck and where can I show the greatest value for my product? Yeah. Right? Yeah. And a lot of companies like, you know, at this stage, you know, you want your early customers to truly be partners. And the more you show the ability to build with and bring them in to that creative process, I mean, you show amazing relationships that tend to last for a really long period of time.

19:03Revenue Builders Hosts: And our FTEs are empowered to, we operate our engineering org out of a single monorepo. All of our FTEs are empowered to write code that we merge into the monorepo, provided they follow our SDLC guidelines. And they've all, everyone on the team has so far built code that's now in our core product. So let me just go back to that whole RBAC discussion. I haven't thought this entirely through, but you're basically doing RBAC for an agent, but I also may have an application where I still need RBAC for personnel, right? Yeah. Yeah. And actually, the way we see it is it's actually a Venn diagram, right?

19:52Revenue Builders Hosts: So the tokens that Arcade holds is every combination of user, agent, and action. Now, we don't mint those tokens. They're minted from either the upstream SaaS service or the IDP. But we hold them and we enforce those permissions at runtime. So when that agent is trying to go take that action against n number of systems, the first thing we're doing is we're doing a check to say, should this agent initially even be able to see those tools that it's trying to act on then should it be able to take the action that it's trying to take and and then there's a bunch of other policies that you can plug in along the way i mean do you also work with application providers also so if i'm building an application today where i want to make sure that my agents can do certain things, but not other things?

20:53Do you work with application?

20:55Revenue Builders Hosts: Yeah. Yeah, so we definitely have some companies that are customers that are software companies. And even some of our big enterprise customers, like some big banks, I mean, a lot of them are software companies just disguised as a bank. They have thousands of applications, yes.

21:11Ben Sabrin: It's that concept of the MCP though, right? Because you want to make sure that you're hanging on to your IP and your critical information and only giving out what needs to get out to whoever needs it. Like for me, in the simplest terms, when I read that Wall Street Journal article, I said this fundamentally is the number one reason why the adoption of AI and everybody squawking about the return on investment for AI is so low is because this hasn't been solved, is kind of what I was thinking in my simple brain.

21:47Revenue Builders Hosts: No, I think you're spot on. And it's a problem that we are aggressively solving. And I think the stakes are very high and there's a lot of legacy incumbents that are trying to tell a story from multiple angles.

22:02Ben Sabrin: I mean, as I'm trying to get the is Anthropic trying to get it back.

22:07Revenue Builders Hosts: I mean, Anthropic definitely is is doing some things, but I think we've all seen plenty of evidence that there's very few enterprises that are completely standardized on one LLM. And if you build control inside of one client, you're going to have to build it inside of every client. And we allow you to just do it from a single place and plug the control into wherever your users are working. But you get it. I mean, we get the competition from the identity guys and the API gateways and the network security guys and the hyperscalers because this is a problem that needs to be solved. Right. And, you know, I think first principles thinking really matter.

22:58Right.

22:58Revenue Builders Hosts: And we've designed this from the very beginning. I mean, we Arcade are, you know, probably our biggest claim to fame is we invented the, how you do OAuth inside of the agent loop. So we, the URL based elicitation that's part of MCP is something that we wrote. And we were solving the problem actually before the MCP protocol came out. And this is a fun story. um so when the first version of mcp came out it was november of 24 um it was pre me being arcade but arcade they were already building right um i kind of started consulting with alex and sam february of 24 or 25 um and right about that time arcade was getting ready to or contemplating publishing the protocol that was underneath Arcade as a competitor to MCP.

23:53Revenue Builders Hosts: And we had Langchain at the table that was all in on wanting to do it with us. And we even had OpenAI at the table. And OpenAI only agreed to talk because they made us promise that we wouldn't support MCP. That's probably the funniest part of the story. And so we had a blog post. We were getting ready to publish this protocol. I think it was a Tuesday and Monday, Anthropic dropped the latest rev of MCP because the first version was a desktop-only protocol. It worked on standard I.O. It had no real authorization or authentication story. When they dropped the newest release in March of 25, it finally worked over HTTP, and so it could work remotely.

24:40Revenue Builders Hosts: and it had a very, you know, a little bit more of an authentication story, still really didn't have an authorization story. And we were already solving those problems, right? And so they dropped their latest rep on Monday. We kind of hit pause on the post that we were going to make. And then Thursday, the company who made us promise that we wouldn't support MCP by talking to us came out and basically said they were going to support MCP. And at that point, right, Right. We believe that our brand wasn't big enough to go fight that fight. And so we decided to jump in and become active members with an MCP.

25:19Ben Sabrin: How great of a story that is, because, you know, when you just look at and I don't want to make any political statement or anything about these companies, but there seems to be so much distrust of the intention of those large language model companies that it kind of it's kind of like perfect timing for you guys. I mean, it it really is.

25:43Revenue Builders Hosts: Yeah, it's good timing. Go ahead, John. No, along those lines, many security companies have now added AI and LLM infiltration to their security assessments of your product, especially if you're an application provider. So I was wondering, maybe this is an ignorant question or a good question. Can Arcade help you, if you're an application provider, limit through policies what information Claude can access? Yes. Yeah, great question. So we hook into DLP systems. We hook into PII detection systems. Any place that policy exists, and so a good example could be, let's say you're somebody in RevOps, right, or somebody within the revenue organization, and you're running, you're expressing an intention that you want to go pull a bunch of account information on a certain subset of customers and you want to email it, right?

26:54Revenue Builders Hosts: And maybe there's something in a policy that says, you know, you cannot share any customer information externally. And one of the people on that email chain is somebody with an email address that is not at arcade.dev, right? If you had the right policy in place, we would block that, right? So we would basically block the ability to send. It would say something like, you know, this contains customer information, and therefore it cannot be sent externally, right? And so, yes, it is absolutely something. And it's all because we are sitting at runtime. So we are sitting inside of the agent loop and intercepting that call before it goes to the model.

27:43Ben Sabrin: It's out in a while.

27:44Revenue Builders Hosts: Yeah, the actual tool execution is happening inside of Arcade.

27:52Ben Sabrin: So you guys are geeking out a little bit. I got my satisfaction out of this conversation because it really is kind of validating what I thought it was. So thank you for that. Now, let's talk about go-to-market implications. Because now when I look at you and I'm like, okay, so Ben has, I mean, he's got, you got an incredible experience in just open source in itself. So that's kind of rare. Like you're one of the early open source, one of the early salespeople in open source, JBoss and Mongo, one of the early people at Mongo. and now the implications of Arcade and dealing with open source. What are the implications of what we're talking about on building a go-to-market engine?

28:51Ben Sabrin: So you've been in these category creator. You've been in these companies that are creating categories. There's no category yet. Or it's an early category. What are some of the implications on go-to-market? Yeah.

29:07Revenue Builders Hosts: I mean, so look, I think it goes to the persona that you want to hire, right? You know, as you start to scale the revenue organization, you need a different type of seller than you would if you were selling, you know, something that was a well-defined widget, right? Right. So when you anytime, especially you're selling infrastructure software, when, you know, like I said earlier, when the product by itself does nothing unless somebody builds something on top of it. Right. You need reps that are incredibly curious. They're not afraid of the technology. They take a keen interest in understanding the customer's business.

29:51Right.

29:52Revenue Builders Hosts: And how what you're selling can enable their business in some positive way. Right. And, you know, the good news is, is like on the research front, it's never been easier. Like there's there is zero excuse at this point to not be insanely informed as a seller. Right. Like a good chunk, like there's even a good chunk of your discovery that you shouldn't waste the time to ask because you already know you should be more validating. Right. What you know versus asking it like you've never understood it in the past. and so I think to that end, right, I am always looking, you know, one of the questions I ask every seller that I'm trying to hire right now is how they use AI and, you know, the more kind of tinkering and the more curious they are, the more interesting it is for me and why is that, right?

30:59Revenue Builders Hosts: specifically for us is because we enable that reality, right? And so I want my folks playing around and building and being excited when they never opened their CRM for the whole week because everything they needed either in or out, they were able to facilitate by asking their favorite model to do things, right? Right. And so, you know, that, you know, people that say, oh, I use it all over the place. Well, they're usually the ones that aren't doing anything. Right. Like I want I want details.

31:40Ben Sabrin: Well, you're right about curiosity.

31:42Revenue Builders Hosts: I mean, that's the number one thing. And then especially since what you mentioned earlier, you can go into an account and probably have 100 different potential opportunities. But which one's the best one for arcade and the customer at this time? curiosity and investigation comes into play

32:02Ben Sabrin: isn't that a go to market dilemma also Johnny? it's a go to market dilemma

32:09Revenue Builders Hosts: Ben's trying to get the thing really off the ground in the right way so it is and it isn't because eventually he'll find some opportunities that he can repeat over and over again I like to say where's the repeatability How do I get that? Because I want to try to, I mean, he's, I think you're in like, essentially product market fit stage. We are, we're about, we're about there. Yeah. And then I always think we move from there to we find some consistency and repeatability. And then you're like in this deal stage, hey, we're getting deals. And now we get deals consistently. And now you say, we got them so consistently and more and more reps are hitting the number.

32:50Revenue Builders Hosts: Okay, now it's time to start to scale. So, and that's what he's looking for now. He's looking for this consistency, repeatability. So it is definitely an issue, but it isn't too. Yeah, I'm a big fan of some of the research that Mark Leslie published years ago, right? The whole sales learning curve concept. And I actually have a little AI baked dashboard that I have. My GrokBot runs for me every morning on how. Really? Yeah. Yeah, on how my reps that I've hired are tracking on the sales learning curve, right? Because, you know, so where we are in terms of growth, right? So I hired the first rep in, so pre us raising the Series A, I hired the first person underneath me.

33:41Revenue Builders Hosts: And the only reason I came, I actually rolled my sleeves up and became the initial seller is because the relationship that Alex and I had, right? There was just tremendous trust. And AI had gotten so good that I'm like, yeah, you know what? I'll roll my sleeves up and get it done. I'm incredibly comfortable in my own skin. And I love selling at the end of the day. I love, especially when it comes to disrupting markets and creating a category, it's a lot of fun. And I also find that when you eventually get into the build stage, right? The fact that all those original logos or a good chunk of those original logos are stuff that you have your direct signature on, right?

34:23The ability to get people really marching behind you is a little easier, right?

34:32Revenue Builders Hosts: Because they know that you're not just, you know, spouting things from a playbook. You wrote the book and you know exactly how to get it done. But then we hired our second, like our first real cohort in June, right? So I hired four reps between June and July. And where we are right now is, I believe I'm going to get all of them on the board with our first enterprise deal before the end of October. That's kind of where we're currently tracking. And once we prove that, then it's, we step on the gas, right? Because at that point, I've proven that we can bring on new reps. We can get them productive in a fairly short period of time.

35:15Revenue Builders Hosts: there's more pipeline to be had from that kind of population of companies where we now have good stories to tell. And let's go do some world-class enablement, maybe a little command of the message, and get everybody in line and hire a couple amazing, you know, frontline managers to go, you know, really drive the team in the right way. and then just don't stop once we, you know, once we're confident in the repeatability. And that's what I call, I mean, this is the stage where I call it going from art to science, right? The earliest stages are, it's a lot of art. You're throwing a bunch of shit against the wall.

35:59Revenue Builders Hosts: You're seeing what sticks. You're not afraid of experiments. You know, it's like I write my plan in pencil and I keep my eraser handy because the chances of being wrong are very high. Right. But once you get to that, like, oh, OK, this looks the same. Right. And I'm starting to see it look the same over, you know, not just like a handful of deals, but like, OK, these 15 opportunities all look the same. Right. The persona is the same. The pain is almost identical. right with obviously some nuances the metrics are very similar right obviously also with some

36:44Ben Sabrin: nuances then it's like okay let's go where where you guys are also like i'm fascinated by this johnny you hinted on it when you we just kind of brushed over well yeah they'll get some they'll get some metrics in place they'll get some but that's just because it's native to you like Finding a Ben, or if you're out there wondering if you are a Ben, and if you've come from, and you've mentioned it, a playbook or your own playbook, there's art and science colliding here big time. Because what I'm thinking about is, how do you not get in the wrong deals? You've got to have some experience.

37:26Revenue Builders Hosts: It's hard, especially when the pipeline isn't super deep. It's one of the biggest challenges. is, you know.

37:33Ben Sabrin: And what is a qualified pipeline? I mean, if you get, like, I know you guys do, you got some cool stories with, like, Goldman, and, you know, that obviously had to, they became a collaborator and a partner, but then everybody's going to want to do that, and some people are going to want to do it. Hey, come bring your Ford engineers in, play around, let me kick the tires with it, and then I want you to do this. Then they think, because of the stage that you're at, that I can tell you, if I'm going to give you a certain amount of money, that you'll go build it the way my vision is. And that takes experience and moxie to be able to deal with a customer like that.

38:13Ben Sabrin: Yes?

38:14Revenue Builders Hosts: Yeah. I'm thankful. I'm very grateful. So Alex, our CEO and founder, is a great partner in that respect. It's very rare that you have somebody, Alex has a computer science degree, but started his career in sales at IBM and went through one of IBM's last great sales training programs. But he's been a founder and he's a product guy, really. And he loves to say no, but he also is not afraid to say yes. And really, when there's the right pull from the market, he will lean in in a pretty big way. And I think that's critical. And I've seen both sides of it, right? So if I think back to early Mongo, you know, one of the things that put us on the right path was we had gotten in deal cycles with Goldman and Citi and MetLife and some other big banks fairly early, right?

39:14Revenue Builders Hosts: and um at that point in time and this is what 2011 right uh maybe late 2012 is when some of the stuff really started manifesting but we really got into it in 2011 if you didn't support support Kerberos for authentication from a database you weren't selling your database to the bank full stop right and i remember like having to drag elliot into that kicking and screaming um but he did like he wound up spending a bunch of time in goldman's offices and and you know they became an unbelievable design partner and we did that because it was clearly repeatable Right. It was this was, you know, while Kerberos was on its way out kind of as a as a technology, it was still entrenched enough that if you didn't play in that game, the banks and some of the big telcos were never going to use your product.

40:19Revenue Builders Hosts: Now, Ben, do you partner with any other companies? Arcade? Arcade. Yeah. So we partner very closely with Langchain. We also partner with Mostra. So they're also an agent builder. From an SI perspective, Wipro is a partner of ours. They're also an investor. And our newest investor and partner we actually announced today is Cisco. okay and we're also just curious why would it make sense maybe i'm again i'm on the wrong track to partner with the cursors the cognitions of the world where they're going in and modernizing applications like in big banks and automotive companies and all that and and ai into those applications as they modernize them why wouldn't you want to partner with them because they're probably going in and almost doing some of the homework for you in the sense that the bank is how if it was a bank they're picking the the app that they really need to modernize because it's costing them a fortune and they want to add ai in there i don't know why wouldn't you try to work with those guys yeah uh we we have um made several overtures it's the mafia of the llms yeah it's uh we're still we're still little for those guys in some way shape or form but if they're listening we'd love to partner specifically I love what Cursor's doing I love GrokBot I've got all of my tools wired up into GrokBot and it does a bunch of amazing things for me and our engineers love we go back and forth there was a time we were all in on Cursor and then there was a time we were all in on Claude and ClaudeCode and now everybody is really loving Codex But part of the beauty of Arcade is we make that optionality easy because your connectivity to all your tools is just a single URL with all of the identity controls and governance and policy, etc., etc.

42:35Ben Sabrin: How do you manage from a CRO perspective the expectations? Like I have a lot of energy and desire and not a lot of applicable understanding of the market. And I'm sure you probably have people that invest and wind up on boards or advisors and they want you to go faster. They want you to do X. How do you manage those expectations being a CRO in a company like this?

43:13Revenue Builders Hosts: Yeah, that's a great question. I mean, I think scaling prematurely, especially from a go-to-market perspective, before you have a semblance of repeatability is just lighting money on fire, right? And so if they're in the business of lighting… It's also setting you as a CRO up for failure. Yeah, totally. Like, I mean, if they're in the business of lighting money on fire and, you know, they're going to give me a stay of execution…

43:40Ben Sabrin: You never get that, though, right? You never get the stay of execution.

43:43Revenue Builders Hosts: No, of course not. It's the easy, you know, it's, you know, this, I always joke with people. I mean, this is one of the loneliest places you could be when you're the CRO of an early stage company. It's the easiest place to point the finger is at you, right? It's never a product problem. It's never a delivery problem. It's always a sales problem. Man on an island, man on an island. Right. But I've got, I've got broad shoulders. I'm used to it. Right. Well, it's a phenomenon in any company I've been in. And everyone thinks they can do the sales job, but most of the sales guys don't think they can do any of the other jobs.

44:19Yeah, I would agree.

44:22Revenue Builders Hosts: Well, in today's day, I could write some code, but no, you don't want me building product. Tell us a little bit about, you made reference to it.

44:33Ben Sabrin: I'm intrigued by the GrokBot and just in general, like what problem were you trying to solve? You wanted to apply the learning curve to your, and if you could just give some people some spirit around, you know, a lot of times when you're doing interviews, 12 months ago, people would give me answers that, well, my company doesn't allow me to use. That's why I say that those answers are out the door now. So regardless of what your company allows you to do, people are still using AI. So tell us a little bit about how you, in your position, you had a gap, whether it be research-based or information-based or productivity-based or what have you, and then you solved it with what was available to you.

45:19Ben Sabrin: Just walk us through that a little bit to encourage folks.

45:22Revenue Builders Hosts: Yeah. So there's a bunch of different ways that I use the machines for me. I think like if we rewind, right. And if you think about how we needed to do things as a leader in the past, now I'm even going back, you know, probably pre-call recording, right. You know, in order to really coach reps, you had to be in a lot of calls, right. You had to, you had to physically be there sitting, sitting, you know, in the room and doing your best to, to shut up. And observe, which is not always the easiest thing, especially if somebody is veering off course. But with really good – I mean, we record any call we possibly can.

46:12Revenue Builders Hosts: And I have a skill that I then use and I've given it to my reps so they can also coach themselves. So they know exactly the things that I'm looking for. and it's basically analyzing their calls on a weekly basis and it's looking for across several different things. Number one is adherence to the message. Are they delivering the message the way we want them to deliver the message? There's still a lot we can do in order to crystallize and make that better. The second is I love MedPic But I also love from a behavioral perspective, the concepts that came out with spin. You know, I'm an old dude. I was very fortunate.

47:04Revenue Builders Hosts: My professor in college is the guy who hired Neil Rackham to write spin. He was at Xerox at the time. And then he became a professor.

47:13Ben Sabrin: Oh, well, I know the story well.

47:15Revenue Builders Hosts: Yeah. So Dick Canada was my professor at Indiana where I went to school. And so I'm looking for seeking behavior versus speaking behavior, right? And that's one of the things I'm consistently tracking talk time as a metric. And then adherence to MedPIC. Like, are they, are they following, when somebody opens the door to be able to easily figure out who the economic buyer is, are my reps walking through it, or are they, you know, not recognizing the opportunity? Are they not getting to metrics or even the beginnings of starting to map the value equation, right? And so I have that skill. It runs.

48:00Revenue Builders Hosts: I use it. I get a report that kind of gives me the high-level themes, and it scores them across each of those different metrics. And they can see it, and they can run it themselves, like encourage them to. I'm like, don't wait until our biweekly one-on-one to try to get better. I've delivered this to you. go be your own best critics. That's one area. The other is on the kind of like how we're tracking in terms of moving deals across stages and building pipeline. I have that that runs every morning. I get a good snapshot that's looking at how many new conversations, how many stage two is converted to stage three.

48:49Revenue Builders Hosts: Because one of the key metrics that we use for how we're tracking along the sales learning curve is how many of the new reps have active deals in tech evaluation, right? So we've moved out of qualification and we're truly in like an actual deal, right? and how many are they stacking, right, in that. So I have that that runs every day. It's looking at pipeline across all different stages. And then I have another one that's looking at intent, right? So I want to make sure that we are effectively prosecuting hot opportunities. And so we use a bunch of different tools under the covers because we have a product that somewhat caters to developers.

49:38Revenue Builders Hosts: There's a really good signal-based product called Rio.dev that's pretty powerful. Also, Sumble is a really good product that has good buying group information, plus my call recorder. And so I have four different systems. And then we also have the Swarm on top of it, which helps you map where's my best point into an account. And so every day, I get a full view of, like, here's the intent snapshot. Like here are the 20 kind of warmest opportunities based on a multitude of different factors. And here's our path in. And then each individual rep gets their specific version of that that they can, you know, from a single button then act on.

50:25Right.

50:26Revenue Builders Hosts: So they can go launch a campaign in Lemlist directly from that kind of intent report or go fire something off on LinkedIn, et cetera, et cetera. So those are some of the ways that we – another great example, like, John, you guys will both probably completely relate to this. So who likes, like, building a new pricing model? I mean, it's never, never a ton of fun if you do it well, right? And so like a couple of weeks ago, I was at a point, and this also speaks to where we are, like we're starting to get into more repeatable patterns, right? We're starting to enable some resellers, right? When you enable a reseller, you need a price sheet that is, you know, very understandable and predictable.

51:18Revenue Builders Hosts: And so historically, if I wanted to kick off a project like that, I would bring my product marketing or marketing leader, somebody from product, one of my sales engineers or my lead sales engineer and rep or two. We'd get around the table. They're like, okay, here's my thoughts. Here's what I want to go do. You own this part. You own this part. You own this part. You own this part. Go do your research. Let's come back. Right. Yet Claude and I, I spent about, I don't know, five hours and we built out a model, had about a seven tab spreadsheet, got down to margin, a calculator for reps to be able to use, a one pager, something that I could push out to partners.

52:07Revenue Builders Hosts: and that took me like a day. And part of it was like, it went through all of our call recordings, all of our existing deals, our pricing that we publish, our basic SaaS pricing, any proposal that we've sent out, it then went and did a competitive look out in the market to pull out anything they can from somebody who looks similar, right? And now we have like a really good enterprise pricing model. And like, that's the microcosm of the kinds of things that you can really do with AI.

52:49Ben Sabrin: You know what I love about that journey that you just described? And if I can call a day a journey, so congratulations, that's amazing. But if you didn't know the problem you were trying to solve, this is why I really encourage people to not be afraid. you know the problem you're trying to solve. You know that complexity is a problem in pricing. You know that you need some type of calculator that's easy and definitive. You know. And so what I try to encourage people is to use the technology as a thought partner. Totally. Not as a thought replacer, if that makes sense. Yeah. And once I did that, I'm just being completely transparent.

53:34Ben Sabrin: The intimidation factor for me a couple of years ago or whenever it started to go away was when I used it as a thought partner, it was never as smart as me on the stuff that I'm talking about, on what I'm trying to automate. And I think that what you just went through there is a great example of, solving a problem that helps. And Johnny and I are really, really big on this. I think when you're, because everything you just did, you're solving it also for sellers. And there's just not enough of the technology in the stack today that is really built to do that. It's to keep an eye on a seller. It's to keep an eye on, you know, and it's run by people that haven't sold and they're looking at it and they're saying this is a quality deal or not a quality deal.

54:25Ben Sabrin: I just think we're in the best time to be alive is that to use that, you know, codify your knowledge with a thought partner and create less friction for your sellers is the name of the game. Yeah.

54:41Revenue Builders Hosts: I mean, it's like I kind of said this earlier, right? We're at a point where if you can clearly express your intention. Yeah. Right. And you can, you can provide good direction. Right. And the prompt doesn't even have to be that good anymore because the models have gotten really good. Right. It can truly be a stream of conscious intention, not structured as a very well-written prompt. Right. And then I'll always in that, you know, basically put in something like, what other questions do you have for me in order to succeed? And, you know, then it gave me like this list of seven, you know, here's seven things I want you to answer before I can go research more.

55:28Revenue Builders Hosts: Right. And you can, you can get some pretty good outputs, but we've got, I think the other thing that you made a good point, John, there's, there's a lot of folks that the biggest thing that could create, you know, I would say negative return in the world of AI is when people aren't kind of doing what we just talked about, right? And they are just relying on the AI. And they're not reading what the AI writes. Like we've got an internal use policy that, you know, if you're going to share something with somebody internally that AI wrote and you didn't read it, nobody else is going to read it. Right.

56:17Revenue Builders Hosts: Why should, why should we waste our time? Right. When you didn't, you just, I could have just generated that same thing myself. Yeah. Well, it's when you understand the domain really well and you ask the LLM's a question that a lot of times the answer comes back and you're like, that's not right. Yeah. Now, 90 % of the time it's right, but many times we're like, that's not right. Yeah, you're missing. That's the domain that I understand. I can't imagine the answers that I get back and the domains that I don't understand, and then I have to assume that 90 % of it's right. Yeah.

56:58Ben Sabrin: Then take us out as we land the plane here, take us out five years, not only for to the best of your ability to kind of like where you think what you're doing and what Arcade's going to be doing and what's going to be normal in the enterprise environment. Tell us also what you think the sales industry might look like in five years. There's no right or wrong answer. We're just, we love to ask the question.

57:31Revenue Builders Hosts: Good question. Look, I think five years feels like an eternity from now, but you know, I've learned that time just keeps moving faster and faster as we get older and it's mathematic, right? You know, when I'm one, a year is a hundred percent of my life. When I'm old like me, it's like a 2 % or something. So it just keeps moving faster and faster. But look, I think in five years, the browser's dead, right? I don't think people work predominantly through a browser. I think we're basically, apps are built on the fly based on an interaction. MCP apps are kind of an emerging thing that's coming in that respect.

58:22Revenue Builders Hosts: I think that there's any work that is repeatable, like process that is just very repeatable is done by machines, right? Without, you know, human in the loop. When it comes to sales, I still believe that people buy from people. And as long as people in some way, shape or form are a key part of the buying process, then I think people will still be involved in the sales process. Now, I think that the yield curve could look very different. Right. I think we should be able to theoretically. Right. If we can remove all of the busy work that we put our reps through. Right. And most of what reps are doing is in they're in deal mode.

59:21Revenue Builders Hosts: And because they don't have, the agents are kind of doing all the busy work behind the scenes, whether it's writing their proposal based on all the conversations that you've recorded and the notes that you've given it based on the frameworks that you have and putting things through CPQ because it's got all the specific terms and sending things to legal and all of the stuff that occupies a lot of cycles. we theoretically should be able to get a higher yield per head, right? If we can make them more productive because we move,

1:00:03Revenue Builders Hosts: we remove a lot of that burden. But I still think it's a very human thing, at least in enterprise sales, right? If you're selling something small and the agent is going to be the person, the thing buying it, well then do you really need a human on the other side of that? Probably not.

1:00:21Ben Sabrin: The buyer will always dictate that.

1:00:23Revenue Builders Hosts: Correct.

1:00:24Ben Sabrin: And the buyer might be an agent in the future on some of that lower end stuff.

1:00:29Revenue Builders Hosts: Any of the commodity stuff, yeah. But if it's brand new enterprise software, no one's ever seen it before. To Ben's point, I think you need a human involved. And, you know, humans are unpredictable. For sure. You know what Ben's going through now where he's starting a company, he's trying to search through different organizations, try to find the right persona to gain repeatability. some of that AI will help with, but you need a human involved to talk to people.

1:00:57Ben Sabrin: Totally.

1:00:59Revenue Builders Hosts: Amazing. Sales has always been part science and part art. Yes. AI will never, I don't, I won't say that. The science is going to get better. You won't pick up the art part. I think the science part, I mean, like, modeling, you know, the calculation and the modeling side will just get very easy and automated, right? But, you know, the stuff that where there's still a decent amount of subjectivity, right, which is a lot of sales, right? Yes. I think is going to remain very human.

1:01:38Ben Sabrin: Yes. It's funny, though. If you took that concept of a forward-deployed engineer and plopped it into a go-to-market at organization and looked at what people are actually doing and struggling with, you'd think you were looking in a cave because it's going backwards.

1:01:55Revenue Builders Hosts: Yeah. You know, it's funny too. And you guys probably see this. I, there's a lot of these new AI native or AI, you know, modern CRMs. None of them are being built by salespeople.

1:02:11Ben Sabrin: Right. Big red flag.

1:02:13Revenue Builders Hosts: When you look up the founders, it's always a couple of tech guys with no sales experience that are building sales tools, not only CRM, sales tools. Yeah, across the board. Yeah, I agree. It's fascinating.

1:02:26Ben Sabrin: So, Ben, I know you got a tight schedule. Thank you for responding to my email to you. I was intrigued by what I read in the journal. I'm now even more intrigued. and thanks for sharing some of the secret sauce. How can people follow you? Can they follow the company? Can they follow opportunities at the company? What's the best way to do that?

1:02:53Revenue Builders Hosts: Yeah, so we've got a very active LinkedIn company page. I'm fairly active on LinkedIn and I'm pretty easy to find. I'm the only Ben Zabrin you'll find.

1:03:05Revenue Builders Hosts: And we are hiring and we will continue to hire in a very calculated way until we have that true repeatability and then we'll still be calculated, but a little more aggressive. Yeah. Ben Sabrin, thank you so much. John, guys, I appreciate you having me. It's a long-time listener, first-time caller. So it was -

1:03:28Ben Sabrin: You killed it, buddy. You killed it. Thank you.

1:03:31Revenue Builders Hosts: It's not incredible. Thank you too. And thanks to everyone for listening to another episode of the Revenue Builders Podcast.

1:03:41Ben Sabrin: Thanks for listening to today's episode. If you enjoy the content, please subscribe, rate, and review the show to help us reach more people. This show is brought to you by Force Management, where we help companies improve sales performance, executing the growth strategy at the point of sale. Check out forcemanagement.com for more information.