Sumner Vanderhoof, CEO of Propensity: Autonomous ABM Still Needs a Human

Sumner Vanderhoof 0:00
Autonomy and context are the two words that I'm kind of like, you know, and these things are changing fast, but those are the ones I think I'm latching into for the next 12 to 18 months. Um, it's not agents, to be honest with you. Um, I think agents are a task machines, and I think those are fine, but fully autonomous processes, that's where it's headed.

Kevin Kerner 0:19
Um, hey everyone, this is Kevin Kerner, Tech Marketing Rewired. I just got off a podcast recording with uh Sumner Bander Hoof, who is the CEO and uh I would call chief technical officer at Propensity. He's becoming one of my favorite people in uh B2B marketing these days. We're very like-minded in so many ways. We're in a bunch of the AI code always. We're trying to manage organizations to push them as fast as we can to be AI native, and we love the tools. Uh we're we're just geeking out on uh all the things uh ABM that's going on in this space right now. But I think this is a really great podcast to hear. If you don't know a lot about autonomous marketing and the things that are happening with the recursive loops, the idea that you can take an agent and have it work on a problem for you and take it out of a human's hands and actually have it optimized using an agent. It's pretty incredible. So his platform is able to do that today, and he has a really strong vision for what we're gonna see in the future with uh autonomous agents and context, context engines. So it was an amazing discussion. I think, I think you'll really enjoy it. Before we get going, though, this podcast is sponsored by my company, Mighty True. We are a growth advisory for tech brands, and we have a uh workshop or blueprint process where we take brands through the primary, we look for the primary growth gaps that they have in their company. And increasingly, that is with systems of context, digital workers, uh connecting all these systems in ways that drive a higher growth and more efficiency for companies. So if you're interested in that at all, go to Mightyentry.com/slash blueprint and we'll get something scheduled for you today. All right, I really can't wait for you to hear this one with Sumner. Let's get to it. This is Tech Marketing Rewired.

Sumner Vanderhoof 2:05
Sumner, welcome to the show. Yeah, thanks for having me, Kim. Good to talk to you.

Kevin Kerner 2:09
Yeah, it is so great to have you here. And as I told you on, I've told you three times now, you win the background award on uh tech marketing rewired.

Sumner Vanderhoof 2:16
Yeah, at least for my part. It's not fake. It's a real background. Oh, that's not like I thought that it was a virtual background. No, not a virtual. There's actually, you know, and they get played pretty regularly. So my kids are the piano players, but those are my twists.

Kevin Kerner 2:30
I love it. I love it. I will not take the uh listeners into like prog rock world and all the stuff that you and I like, but we'll do that on another another episode. Uh so let's just let's just kick right in here. You and I did a a pre-call where I think it was just like how amazingly like-minded we were in so many ways. And one of the things that you said was that you you spend more time in ChatGPT, it seems like, than you spend time with your team team, which I can totally relate to. Now I know that's not entirely the thing, but just explain to me as a CEO, like why you're why are you in it so much? Like what is what's your workday like that you're leveraging it as much as let's say I do.

Sumner Vanderhoof 3:10
Yeah. I mean, I and you know, I I don't know that it's a good thing. I think I think it's a necessary thing at the moment we're in that I'm spending so much time with it, but I don't know that it's the the long term. I actually have been telling the team at November 1st, but you know, it's 10 eight or 10 days from now, I'll be back into like regular communication patterns with everybody. So yeah, my take on it right now is like so much is happening so fast that like, and and first of all, I'm so I feel like I'm I think I told you this the other day. I feel like I was grown in a lap to be able to use AI really effectively for what we're doing here. So we've built, we started with building our own internal tooling on AI. So autonomous engineering is what we call it. So our entire engineering team now does all of our code work through our own MCPs. So we've, you know, all the you know, the warnings about AI thrift and AI slop, we've just like carved out all of that stuff. So now AI has a very clear run through our engineering practice, so much so that we're allowing our customers to vote on ideas that they want to have implemented. And then our own engineering MCP will pick that up and do the development based on the customer scoping if it meets the certain stage gates that we have. So we have a lot of trust for our own engineering practices now, like not quite to production deployment levels automatically, but through like staging deployment, certainly. So that's it's just been a ton of fun. So I think that was like you know, down that rabbit hole. And then on the kind of the more customer-facing of how we use it to grow our business side of things, we built our own autonomous marketing platform within Propensity. So that is what Propensity has become and what it's what AI has enabled it to do. So we take context from CRM, marketing automation, website traffic, all the ad platforms, put it into Propensity, and then use that to inform the best audiences, the best campaigns. And it just executes and learns, and every campaign is smarter than the last one. So I think by nature, I'm moving away from user interfaces and into AI interfaces because behind the scenes is all the real meat of the software, the servers, and the execution. But on the front end is you tell it you want uh you know, $150,000 campaign for Q1 of 2027 that executes competitor conquest and it knows what that means and it goes through it. So we've been we built autonomous engineering in order to build autonomous marketing, and now I'm fully in the kind of the AI, uh AI world at this point, kind of just talking to AI and telling it what we want to build, and it just goes and does it for us.

Kevin Kerner 5:58
Yeah, it's incredible when you have the background and context. And I guess your your like product vision there primarily, right? So you're you're needing to be in it a lot to understand what it can do. And the autonomous stuff for us has been a game changer in terms of what we're trying to do with loop. We do have this thing we call loops, where it does sort of a similar thing. It's an agent, it's a digital worker that's looking at target assets or sites or whatever, and it's recommending changes. And on our own site, we run loops on our blog posts, and it's incredible to watch with the A, how hard the AI, the digital worker goes after changes to optimize the thing. Um, now I don't, you know, we have to watch it. You know, there's always got to be a human in the loop, but I can see a world where that autonomous looping marketing is the way that things get done. And it's just like digital workers happening. I mean, you've been doing this for a while, so I wonder how fast that transition is that you're you're seeing it, because you have customers that are doing this stuff. And then what does it take? And I don't even know if you would know the answer to this yet, but what do you think it takes to move beyond a think a marketing team that has a human doing all the stuff in your platform versus giving them agentic workers that are doing things on their behalf?

Sumner Vanderhoof 7:20
I I think it's it's both. So the humans are the guides now for the AI. And it's you know, we we uh we're kind of straddling the line a little bit because we've been around long enough to have a really healthy customer base. So the question, and my my colleagues ask me this all the time, it's like, how fast are we going and are we outpacing like what we've shipped before? We're still in startup mode. So we still have this agility that allows us to move really quickly. But the customers we're bringing on now are being brought on fully into that like autonomous ABM, autonomous marketing motion. And that means the onboarding's easy, it means the implementation is super easy, it means that they're interfacing with the learning system from day one, and the context they're adding to the system is just providing that continuous loop uh from the onset. So I think you know, my my involvement with it is product development, probably interesting. Part of it is Kevin, too, like the it's the product I've wanted for the past two years, I'm able to build and then some, right? Like I think we're able to stretch beyond what I even thought was possible. I used to say I wanted to clone myself to be able to be in multiple places at once, and now that's not even a thing I I like first of all, probably wouldn't have liked that very much. But secondly, I don't think it would have been as it's certainly not as effective as what the tooling we have now. So my my new framework is is basically like you know, bring on everybody into the new the new world as soon as they're ready. That might mean they get pieces of it now, and they're just using for analytics and insights where we have other customers that if you ever you know if you use Claude, there's the always ask and then always answer answer for me, and then there's the red button that's like never ask anything, presume everything, full access, bypass permissions.

Kevin Kerner 9:17
Yeah.

Sumner Vanderhoof 9:17
We we have customers that want that bypass permissions, and then we have others that want the governance and the control point. So the people and our customers that are involved here cover that gamut. Some of them are like, just do it for me. Uh we want ABM to be autonomous. We don't want to build like the in-house expertise on ABM. We want a tool to do it for us, and then we have other customers that really want to shepherd it, guide it, um, as a it's a core strategy for their business. And they don't put it to the you know, unlocked permissions level. They they put it more on the guarded side of things.

Kevin Kerner 9:53
Yeah, it's so interesting. I was just talking to someone this morning about how we when we're, you know, we're in it every day, we're doing all the things, and you kind of think that everyone has this mindset of where you know we are, or maybe they're a little behind, but they, you know, they know all the things. But we got to realize there's like, I don't know, maybe it's 90%, I don't know, 80% of people in marketing, they're still doing it kind of the, and this is not no knock on them, it's just they haven't gotten there yet. But there's a lot of people that haven't gotten to all the way to autonomous looping and recursive loops and that type of stuff. And that's okay as a product. I think you guys are really well positioned as a product because you are giving that option for both, and you have autonomous engineering that's able to build these features really fast. You don't have the technical debt of some of these other platforms, which is, you know, a huge opportunity. I wonder how much pushing you need to do on the teams or the organizations or helping these marketing organizations sort of bring them along. I know you're saying that some people are ready and others aren't, but you know, how many people are at that advanced end of the scale and then how many are just using the platform as it would have been used two years ago?

Sumner Vanderhoof 11:05
You know, it's I was just talking to one of our board members about this yesterday, that the pain is still the same pain that they're coming to us to solve. And it's just how aggressive are they in solving it with different tactics or different tools. And the pain is a lot of the times it anchors into like you're looking for predictable, repeatable growth. And a part, a big part of that is marketing being able to hand leads off to sales that sales wants and can work and turn into pipeline. A lot of people still sales yells at marketing because they're not giving them valuable leads, and marketing says sales isn't even calling on the leads that I'm giving them. It's the same old problem. We've always seen you're looking for predictable, repeatable pipeline, you're looking for marketing sales alignment. And what we have now is an opportunity to take this like the the friction that existed between those teams and in that process, and then remove it. So basically now, you know, the marketing motion uh will it is hands off to sales, and we have a function where sales either accepts or rejects the leads as part of this. And you know, that's that type of tool has been around before, but sales has always been like, yeah, I you know, what's what's what are you gonna do with the data? I accept it, I reject it, and like that just goes into the pit of like nobody learns anything from it. Well, now it goes straight into the model. So every input, whether it's an impression, a click, a qualification moment, uh, an exception, accept from sales, a reject from sales, pipeline be built, all that immediately informs the model. So now that engine that was built with friction in it smooths right out, and then it learns from itself. So every every contribution from a marketer, from a sales rep, from the customer makes the engine smarter as we go through it. So we have customers that come to us, they're still trying to solve that engine, they're still trying to solve that process. It's just how far do they adopt it? Do they adopt it? Do they believe that, like, you know, and it's not AI, do you believe they can smooth the process out by inject like pattern matching through the newest tools we have, which is AI. It's just it's not the it's not AI that's the answer. It's that process that becomes automated, that process that becomes frictionless now. And then you get numbers. It's like, oh, well, it can be predictable, it can be repeatable. If you put this in the engine, it gets us out the other side. And if we make these changes, then it gets tighter and better. So in some ways, we're say solving the same old problems, the forever problems, with just tools that actually make it frictionless now between marketing and sales, which is uh a huge gift for uh for everybody involved.

Kevin Kerner 13:50
Yeah, that's the type of automation I think everyone would like stand up and applause for. Because if you really think about the CMO challenge or the marketing team challenge, you know, it's you have to you have to drive effective campaigns, you got to get people to fill out forms, you got to qualify them, you got to get them. But if you got to get them to the doorstep of sales, then something needs to happen. And if you could smooth out that friction, or at least have both sides agree in an easier way, I think that's a that's solving a really big problem. And and those are typically disconnected systems, you know, very disconnected and hard to make work together. Is propensity building some of that capability inside of your platform? And or does it interface with like a Salesforce or a HubSpot sales or something? Or does it do both, or is it in or using another source system of context to make that happen?

Sumner Vanderhoof 14:40
That's that's the word of the probably the word of the decade going forward is context, right? So that it's wherever we can get it in wherever systems we can be in. So we integrate fully with, so first of all, for sales, be where they're at. We don't need to be tool for them, but so Salesforce, HubSpot, Gong, Outreach.io, Apollo, any of the tools that they use. Marketing typically uses propensity more regularly, but sales does use it occasionally if you want sales inside of Propensity. But they accept, reject, I mean, wherever we can get the input from is really what we're going for. So we can pull data a context. We call it the context matrix. It's a a map of all the GTM data you have. Some companies aggregate signals and that's what they call it. But it's essentially that's what it is. It's a signal aggregation tool to inform that discrete like marketing the sales process for B2B companies. So we we have a context matrix and it's it's sophisticated enough to tell you like how much progress have you made in developing your c your signals and how high quality are the signals. And then you find from a like a RevOps and marketing ops perspective where the gaps are in your signals. And then that in that's the first layer. So a signal is you know uh sales engagement, sales touches, sales call volume, and connection rates, but another signal is uh technographic data, job postings. And we have all of that built into propensity where the more signals we aggregate, the smarter the model can be. And we've built tooling, uh, there's also false positives, right? So you want to exclude certain signals that aren't meaningful. So there's ways to like like from the onset, even if your data is dirty, as I I think that's a a misnomer that like with if you bad data in, bad data out. Well, with AI, it doesn't really have to be that anymore. You can actually sub you can subdivide the good data from the bad data without doing big database migration projects, put it into a context engine, and then feed the loop around the quality stuff that actually informs the best patterns that you can have on the other side of it. So we've just built that like sifter, the context engine, the context matrix. It takes all those signals, puts it in, and leaves sales where they're comfortable, and it gives the engine the intelligence it needs to inform ideally the perpetually uh evolving campaigns that go and become better and better and better.

Kevin Kerner 17:15
Sure. I I I think that's really cool. The idea of context is something I've been like running around in my head because we have some clients and I have some friends that are implementing kind of master, they're not really CDPs, they're just master context context engines, and they're building platforms to do it. And they're either doing it through Google Enterprise or Copilot, or I someone mentioned Glean the other day as a tool that's that's you know, that content, and then they're putting they're putting agents on top of these things, and then they're able to, you know, pipe in other on a other connector to them. And you you have a context engine that runs your stuff, but what's your thought about more of master company context and how it relates to a platform like yourself or a marketing use marketing and sales use case for that stuff?

Sumner Vanderhoof 18:07
Yeah. We we have customers that have already built their context engines in Snowflake, for example. Yeah, yeah, that's a good way. They aggregate all of it, that makes our lives a lot easier because then we don't have to piece it together and have 10 different connections. Now we have one connection to their context engine, and now the context feeds the marketing and sales engine, the go-to-market motion. So if they have it already, which is a rare treat, then we leverage that, but it still goes through the sifter because not everything you have in Snowflake is typically like cleansed, let's say, qual high quality data. We look for quality, we look for recency and frequency, we look for like how recently the data's been updated. Um, we look for like robustness, does it have enough data points? So typically, even your best databases are missing something across the board there. So we enrich, we add uh add the layers that are missing within propensity to just give it more more either structured data is actually what we work with in because it allows the model to be more intelligent. But even if it's unstructured, we can work with that and pass it through the sifter. And the end of so whether it's uh a sale snowflake, uh, Clickhouse is a great one. I love that one. I love Clickhouse as a data store right now. Um, but wherever it's coming from, we absorb it, put it through the engine, and now marketing has got a tool and an AI tool at that that they can just ask questions against the context. So one of my favorite is like again, uh you know, just talk to my phone while I'm at the gym or something, and I'll say, give me, you know, I want to build my Q1 campaign. Uh, and here's the parameters I want to do for it. It'll go build the campaign. It'll know based on our history what the best channels are, where the mark where our our market engages the most. It'll build out the draft. I'll approve the draft on my phone from voice chat typically, and it'll actually build out the campaign and launch it for me if I kind of approve everything downstream. And then it'll feed automatically hand off to sales because we've already got that pipe set up into Salesforce. So that whole engine is now built on the context that we've already given it, and then it just feeds itself. And it's now I'm the guy and I'm the idea. Somebody said the other day, uh, as a barker drive, you mean I get to market? And that's like I don't have to do the administrative overhead anymore. I get to actually do the the marketing work.

Kevin Kerner 20:35
Well, yeah, that's the change. And those are all little digital workers inside that thing. I am so glad someone else is in the gym talking to an agent of some type. I I bet I'm looking around and I'm betting I wonder how many of these, I wonder how many of these other people in here are actually like talking to the the thing. I'm glad to know someone else does that. It's sad. It's uh it's embarrassing.

Sumner Vanderhoof 20:55
But then everybody's got their headphones on, so I'm like, I'm fine.

Kevin Kerner 20:58
Right. And they're waiting for the like the uh shoulder press, and you're doing the thing and you're like, shoot, sorry, I was uh you know talking to the talking to the agent. What do you think before I get into the benchmark questions, where do you think this is going from your product perspective? Maybe the AB ABM side, but you're they're getting even broader than that. Where do you think this is going? Can you like where is your head at in terms of next level of what you're doing? What is what is it?

Sumner Vanderhoof 21:25
So uh autonomy and context are the two words that I'm kind of like, you know, and these things are changing fast, but those are the ones I think I'm latching into for the next 12 to 18 months. Um it's not agents, to be honest with you. Um I think agents are attached machines, and I think those are fine. That's a good place to start. But fully autonomous processes, that's where it's headed. One thing we I refuse to do at Propensity is install any MCPs other than we have two. We have our engineering MCP that we built, and we have the propensity MCP. I just won't, I don't the the last thing I want is to wake up at 12 months and have 50 different MCPs connected with agents all over the place and I don't know what's going on. So I just feel the pull towards chaos if it's an agentic first approach. So instead we're taking like a process first approach and automating processes in its entirety. So if we don't see it as a full end in process that we can move through in its entirety, then we won't add an agent to support it. It's a long-term kind of a play for us where we think that's better scale and freedom than building a bunch of agents that are just tacked on. You know, it's the you know, you end up with 200 different SaaS licenses a few years ago, and you're gonna end up with 500 different agents running all different conflicting instructions and things like that. So I think it's you know, train the context on the process uh and like automate the process. And then within that process, that's where you create the constraints, the control. So we have one MCP that has all the agents embedded in it to facilitate our engineering process. We have one propensity MCP with all the agents in it and all the context to facilitate the go-to-market process. I think that's like the goalpost that hasn't yet totally materialized for a lot of companies yet. And as they the bigger companies use governance as a way to get there. They're like, well, if we don't have governance, we don't have control. If we don't have control, then chaos ensues. So by nature, they're moving towards process first approach. And I think everybody is gonna kind of follow suit. And the companies that are building those context engines are gonna find that the ones that thrive are gonna, because the context facilitates a process, an end-to-end flow, and not just a repository data. It actually has a purpose, uh, an output on the process.

Kevin Kerner 23:56
You could really you could you could really end up with a lot of disconnected individual agents doing all kinds of crazy stuff. It's kind of like when I first got N8N access and I started building my first like workflows years ago, and I had a bunch of little tiny uh workflows going that didn't connect to each other, and then I noticed that this this workflow should actually should have connected to this workflow. And then I did the same thing when I got in Clogg Code and I built a bunch of Node apps and cron jobs, and they were all disconnected. It takes a while to build think process first. One of the things that helped us was that we the other way I flipped my mind on this stuff is I think about a person's job description, like a role, and then I try to figure out what that type of person would do, like a data analyst or a uh content marketer or whatever. And and then I think about the process on top of that, and I I organize my agent thinking around the types of digital workers that you we want to create, and then it's a digital worker folder that also connects to all these other different individual digital workers and has human loops. So it's just but we're early times on all this stuff. I don't know. I don't know how I think I might have mentioned to you like managing agents outside of a platform like yours. And that it's not we do it in GitHub, but it's really not the optimal place to have marketers in GitHub maintaining stuff in a dev environment. Someone needs to build something that's easier to manage these things and maintain them.

Sumner Vanderhoof 25:35
He's got like five different licenses running at full bore, doing uh a bunch, like dozens of agents simultaneously all week long. And it's just the newest form of technical debt that's gonna come out of it because everybody's empowered to do it. He's actually really sophisticated as an engineer as well, so I hope he's he's got his his bearings. But like you said, it's early and we're I'm building for 2028. Like I'm building for where the world is headed. I that's at least trying to as best I can. Yeah, I'm trying not to build for the current state. Um, I think there's I see announcements sometimes, and I'm like, you guys, you're the that is a flash pan moment type of like AI announcement that isn't gonna bring the customer or the user or the marketer forward into the next. Like they're not leading the edge, you're not on the leading edge of anything with it. You're just adding another agent to the pool. And I think for us, I'm just my goal is it's so it gets back to what we said in the beginning. It's like it's still about predictable, repeatable pipeline. It's about reducing the friction between marketing and sales, and it's about maintaining that process. So our job isn't to just introduce a bunch of agents across that, it's to make sure that entire engine is sophisticated and working. And you asked me a minute ago what's the vision. Like right now, we're in the like autonomous ABM, is the where we're at right now. Like I way faster than I thought we would be. I thought I was aiming for like December, January, February. It's October, and we're there, which is like the engineering speed is insane. On the other side of it, our goal is fully autonomous marketing. So beyond just ABM, there's so many other areas that you can automate to still create that predictability and repeatability and pipeline. And then the end goal, I think, for the entire process is fully autonomous go-to-market, which combines like marketing, sales, success, renewals, the entire lifecycle of everything where you can turn on a context engine and it loops through the entire and it makes your whole business smarter as it goes. And I I don't think it replaces um the need for people. I think it really empowers everybody. My, you know, what we found is our team is so much more effective when they can like, you know, they can do a lot more faster and contribute at a much higher rate. And they're rewarded by everybody, I think, right now, isn't as enthusiastic as I am about what we're building and and kind of what what this offers up is from like a just a day-to-day enjoyment of the work perspective. Um, so I don't know, unless you like the grind of like uh, you know, in in the weeds.

Kevin Kerner 28:21
No one likes the grind. I don't think no one likes the grind. No one likes the grind of that type stuff. It's so it's so true. And uh you just you just need a team, and we we we're the the same boat with team. You gotta have a team that is creative enough to see the opportunity and then a little bit fearless in terms of what they want to accomplish and has some grit because it takes a while to get this stuff built and built in the right way for scale. And uh, but I I I love the vision, like the the process part is something that's missing from most companies in general, before AI ever happened. You know, no one many companies don't build for scale and process. And I think the AI stuff is getting us all to think about that in a way. And then I think the platform thinking, like what propensity is doing, will move people, it'll help them into a very systems-driven thinking, which is needed to run, you gotta run it, you gotta have systems thinking to run a successful business, successful marketing department, successful family. You know, it's all gotta be, it's all gotta be systems. Okay, I have three benchmark questions I want to ask you. I asked these to every guest. We're about the end of this season, so we'll put out the report at the end pretty soon here, probably in the next month or so. And you maybe you can answer these from Propency's point of view as the CEO or let's say chief marketer product person. So the first one is what's one workflow you put in production this this year that moved a real number? Something that you were like, this workflow killed it.

Sumner Vanderhoof 29:49
I don't think there's been a number that hasn't been like insanely positively influenced. Uh I think the the biggest one is uh we added this is actually this will be an interesting point, uh Kevin, because it's like the inverse of everything we just talked about. So we added a human to the loop intentionally, um and that we did it at point of like qualification. So we added a team, and it's part of our product as well, where if we don't have enough signal and we don't have enough context to suggest that sales should be working this deal or this prospect, then we pass it to a market research team. So we added a market research workflow both inside a propensity and with a team. And their sole job is to ask benchmark questions. What are you working on at your connect with people, ask the benchmark questions? Is there interest? Is there a need? And they have reduced the they've first of all they've reduced the false positives that sales gets by a tremendous number because sales only works validated leads from the market research team or that are like high enough scoring that they skip the market research effort. So that workflow has been like we've gone from you know the the the quintess, like the normal, I think ABM outbound numbers. I think when we're a marketing or a market research team validates a lead, sales converts like 50% of them to pipeline. So it's an insane like handoff motion that we've been able to add. Um and we did it, we I just called it like human-driven intent data. So instead of buying intent data, digital intent data, I'm buying it from our market research team, and it's just human-verified intent. And that's been a huge driver for us over the last six months or so.

Kevin Kerner 31:40
That was the way it was done in the 90s. It worked back then too. But yeah, so it literally had to be.

Sumner Vanderhoof 31:45
We just had this whole AI vision, and then that's what it, yeah. Sorry, yeah.

Kevin Kerner 31:48
Yeah, but no, some things are just they just work, but that's really interesting. So you're then the research team is doing some level of qualification and research on that actual like account slash opportunity that comes over before it gets to the BDR salesperson.

Sumner Vanderhoof 32:04
If so, there's high-quality leads that just auto-stamp go to sales, and then there's a whole swath of them that are like, we think maybe they're ready for sales, but we don't want to give sales a bunch of stuff that's not ready for. So they just work those and they just ask, and they typically call into non-leadership roles, people that pick up the phone a lot more, and they just say, Hey, what are you working on? You know, it's more tight than that. But it's like really, this is you know, Corey with propensity. We're doing market research. We have a benchmark report we're putting out there. We want to ask you a few questions about what you're doing right now. And then as the call they do not book demos, they're not looking for handoffs, they're not looking for like introductions, they're just looking for qualification validation. And is there something here that's worth a conversation at some point down the road?

Kevin Kerner 32:52
And it works great. So smart. That's wild. Yeah, because I worked at Hard Hanks back in the day, a long time ago, and they had the CI technology database. And that's how we used to get intent data. You they would call people up. It was a research project. They'd say, Hey, I'm calling uh research, trying to figure out how much hardware, blah, blah, blah, is getting sold. And they would run them through a research report, a research study, and it would be it was great. It was really good data. And that was intent data back then. Of course, now it's a lot different. It's really good. Way to go. That's uh that's that's a good one. That's really good. Silly one.

Sumner Vanderhoof 33:26
I like it. You gotta come out of it from all angles, I guess. We got the human that's really good. AI.

Kevin Kerner 33:32
No, some people are just yeah, not not ready to even don't don't want to hear AI anymore. So it's great to hear like an old, like a it's being done the old way, which is really cool. And it's working. Second question is what's still stuck in pilot? Something you expected to be running by now that isn't.

Sumner Vanderhoof 33:48
So I measured this the other day. Like, because I still play the role of like CTO at the company, and I get my my hands dirty on code every once in a while. I you know used to do it like nights and weekends, but I'm in like the refounding moment of the next gen of of AI and propensity. And I measured it. I used to contribute like 60 touches on tickets. Um, and that would be like a a pull request or a merger, whatever it is, right? Uh a week, uh 60 a week, um, which is really efficient and effective. I'm doing 6,000 a week now, which isn't in I'm like literally 100x more effective. So there's the the the hardest thing that's stuck, if there is one, is data migration projects because it's still the piping of the internet. There's nothing AI can do to make moving, you know, 500 gigs of data faster and more reliable, and then do the delta switch over. So data migration projects. I'm an old school uh cloud, before it was RevOps, it was cloud consulting. So I have a big background in like moving Excel files to the cloud, is where I got started my career. And I feel like those projects are still you gotta have a person doing them, you gotta know how to do the cutovers. You got so those are the ones that are slowing us down if we're slowed down at all. But every software-related project is uh 10, 20, 100x, what we were doing six months ago, even it's an insane.

Kevin Kerner 35:20
That's very interesting to hear. Yeah, because now data architecture is database architecture and re-arc re-architecting tables, all that kind of stuff, happens like no problem. Oh, oh, so I made a I made a mistake on this table. Just go and fix it. It's fixed. And that used to be something that would take weeks if you made a mistake in some schema. It's incredible. It's an amazing time. This last question is what's one thing your team won't hand to AI and why? Like you're saying, we won't do this. This is not a it's not something we're doing.

Sumner Vanderhoof 35:52
Yeah, we I don't give it at our bank accounts. So like I stopped short of that. I've been tempted a few times. Uh those ding investor updates are a pain in the butt to write, but I I still kind of I don't want to give it. I probably should at some point some access. Yeah.

Kevin Kerner 36:11
Well, I told you, I think on our pre-call, I was telling you I was using Muse, and I've and I'm the guy who will just hit like you know, I'll hit accept, accept, I'll and I did have it buy some tickets for me on on Muse, and it worked great. It did it through Stripe, you know, the Stripe integration or Apple Pay, but it's really it's really cool. So it got me thinking about, you know, I think I would probably continue to do that. Claude just is launching duet, so that I'm really interested to see how that that runs versus and that I'm that's one other thing I'm really interested in that would be a great thing to have in some of these products is a muse-like, you know, dot-like thing that is you know, maybe you call it something, you know, it's named uh, you know, whatever, and you can talk like you are talking to yours right now. But uh that's been really helpful for me. Yeah, just a companion. Yeah, it's been it's amazing. Uh it's has my running schedule in it, it's got you know, bought tickets, it's got some outreach I needed to do, it helps me follow up on emails. It's easier to use than dispatch, I think, because that's some something about dispatch is just a little disconnected from everything. But so I I I would have never guessed Facebook would have had the lead in that. But I don't know.

Sumner Vanderhoof 37:28
From the B2B space, we've gone, we've started with the functional AI that just did the thing. You had to prompt it just right, and it would do just the thing you asked it to. And then we move to conversational AI. So now it talks, you can talk to it and like massage your output. And now we're at the point where we're doing comp I call it companion AI. Yeah. Where it is like uh, you know, it's got the context of everything. It doesn't have your credit cards for your business, yeah. But it has the ability to purchase ads on your behalf, it has the ability to run campaigns, so that's not that far apart. And you get you got to feel comfortable giving in budgets, and then you can just di deactivate the uh the launching protocol. So you can draft everything, but only a human can launch. So that's I mean, Kevin, you nailed it. That's exactly what propensity, the autonomous ABM, the autonomous marketing is built on, like a companion model that you can talk to at the and give it those details and just have it act on your behalf and remind you of what you should be doing.

Kevin Kerner 38:30
Yeah, I wonder if it ever gets to the point where it's more personal, like this. My my companion propensity app knows me. It knows our company, but it knows how I like to run and knows my voice. And because there's that would be really cool. If it knew that I had to, you know, could also tell me that I need to go to the gym or my running my whatever. That might be a little too much. Yeah, that's right. Okay. So I'm gonna do AI roulette with you. So I've I've had whisper flow here going along the side, listening to us, but it also has what we talked about in our pre-call. So who knows what it'll come up with. I'm gonna click send here on a prompt I had just to give me it. It'll be a question from the AI, it's not from me, so it'll read like the AI. So okay. ChatGPT scopes your work, Codex builds it, Claude checks it for mistakes. I noticed who got the grading job. I'm fine. I am fine. If one of them had to go tomorrow, which one gets fired? Notice who got the grading job. I'm fine.

Sumner Vanderhoof 39:28
So this is some previous conversation I think I had to. So I you know, and so basically the way I use it is I use voice a lot of the times to scope out things, campaigns I want to build, or in our engineering tool, I'll scope out like a new project with the team and then send it. We have a what we call our MCP Forge. So I'll send it to the Forge and it will uh package it up, whatever I scope, and immediately send it to Codex for development. And then Codex will finish that up. And then just for a second set of eyes, I send it to Claude. I I like Claude as a second set of eyes, but it's the thing I need the least. I actually I'm a bigger fan from an engineering perspective. I'm a bigger fan of Codex. Um but our all of our customers use Claude. So that's where we build on, we build tools for Claude, but we engineer inside of Codex. I think amazing.

Kevin Kerner 40:23
Like Claude has taken over. It's taken over. If you had to lose one, Claude or Codex.

Sumner Vanderhoof 40:29
Day-to-day, yeah. But from our customers, they you know, they wouldn't, they would not be happy about that.

Kevin Kerner 40:35
Yeah, it's just amazing how how Claude has taken over. And like if this time last year, I wouldn't, I guess we wouldn't have been saying as much about Claude as we did this year. And who's and that's why I'm so scared. I've got all these different LLMs that we use. I'm so scared to drop one because who knows what's gonna happen in the year. Google might come and overtake things with anti-gravity, but it's pretty it's a pretty amazing time. Well, Sumner, this has been awesome. I could go, I go a lot longer, as you know, but I'm not going to. I'm gonna save you the time, but it's fascinating. It's just amazing to hear someone that's in the product. It's really a next gen sort of approach that you're taking. You're just gonna, if you can pull this off, it will really, or as you're pulling this off, I should say, because the product's great. It's really a there's a lot of potential in the marketplace for for what you're doing. It's amazing. I'm a fan.

Sumner Vanderhoof 41:24
I appreciate that, Kevin.

Kevin Kerner 41:26
Thanks a lot. Is uh what should people do if they want to get a hold of you or learn more about what you guys are doing?

Sumner Vanderhoof 41:31
Uh yeah, there's there's only a few sumners in the world, so feel free to look me up on uh LinkedIn. Uh sumner on LinkedIn, propensity.com is great. Uh, and then we're gonna be at the London ABM Summit on November 4th. Uh, is a great next place to find us.

Kevin Kerner 41:47
That's great. Yeah. Yeah. And I know Kelly Greenwald is a is a friend of ours, and I'll mention her in the notes here, but she's fantastic. Your team's great. So I really appreciate you coming on. I know you're super busy, and I hope to uh see you in Austin at some point.

Sumner Vanderhoof 42:01
Absolutely. You got it, Ken.