The Data Stack Show - 150: How Salespeople Use Data, Salesforce vs. Snowflake, and How LLMs Are Transforming Sales with Brendan Short of Groundswell

Episode Date: August 9, 2023

Highlights from this week’s conversation include:Brendan’s background and journey to Groundswell (2:25)The impact of generative AI on sales reps and product building (5:38)Lead sourcing challenges (12:22)Salesforce as a data model (14:30)The need for guardrails in building applications around sales (24:37)The question of interfaces in the layers of Salesforce (26:11)A UI solution for sales and marketing (30:45)The future of logic and machine learning models (37:11)The battle for data ownership (39:36)Actioning data and the role of refineries (46:03)The potential for decentralized systems using generative AI (46:59)Product building for the future (57:47)Building trust in data tools (59:10)The era of innovation (1:09:20)Final thoughts and takeaways (1:10:43)The Data Stack Show is a weekly podcast powered by RudderStack, the CDP for developers. Each week we’ll talk to data engineers, analysts, and data scientists about their experience around building and maintaining data infrastructure, delivering data and data products, and driving better outcomes across their businesses with data.RudderStack helps businesses make the most out of their customer data while ensuring data privacy and security. To learn more about RudderStack visit rudderstack.com.

Welcome to the Data Stack Show. Each week we explore the world of data by talking to the people shaping its future. You'll learn about new data technology and trends and how data teams and processes are run at top companies. The Data Stack Show is brought to you by Rudderstack, the CDP for developers. You can learn more at rudderstack.com. Welcome back to the Data Stack Show. Costas, we have a really fun show. We talk with all kinds of people in the data world.

And today we're going to talk with Brendan Short. He's the founder of a company called Groundswell, but he spent time at Zoom. He's been a founder, has exited startups, many of which use data, but all sort of focused in the sales space. And he has some really strong opinions about data models and how they create moats for startups. And his perspective is really going to center on sort of Salesforce and sales tooling, which is interesting. But he's also the first guest that we have who is building a company entirely based on generative AI. So his startup is leveraging LLMs to generate actual materials that are sent on behalf of salespeople. Fascinating.

So a couple topics we haven't covered before, which I'm super excited about. I really want to learn about how he, as a go-to-market person, is thinking about building a company on generative AI. Because we've talked a lot about MLOps, we've talked a lot about LLMs, but he's taking a company to market. And so I want to hear from his perspective how he is thinking about that. So that's my burning question. How about you? Yeah, we are going to have a unique opportunity to talk about how data can be a mode, but

in a very surprising way. I think like, okay, everyone says that data can be a mode, right? But the first thing that everyone thinks when we are talking about data modes is having some unique secret data that no one else has access to, right? But with Brendan, we are going to talk about something a little bit different, a little bit more meta, but equally and even like maybe even more important and much stronger as a mode. And it has to do with Salesforce. I'm not going to say more right now, but I'm really looking forward to having this conversation with him about data modes.

All right. Well, let's dig in and talk with Brendan Short from Groundswell. Let's do it. Brendan, welcome to the DataSec show. So excited to have you. Thanks for having me. Appreciate it. All right.

Well, we've known each other a long time, actually, through many startups, trials, travails, and successes. And so because I know you, I'm actually going to ask for you to tell your history in sort of two phases and tell me if this doesn't make sense. But you have a lot of your career has been focused on go-to-market, which I'm super excited because we haven't had that many people on the show who focus on that. But you also recently founded a company that's sort of being built on AI, which is really interesting. So can you sort of tell your story and maybe those two phases

or give us the breakdown that makes the most sense? Yeah, absolutely. So yeah, I've spent the better part of a decade in B2B SaaS go-to market. So squarely in B2B SaaS and literally started off as an SDR 10 years ago or so. I didn't really know what that role was about a week in. I realized what it was and it was quite the shock. Cold calling, emailing, brutal. That's right. Exactly. So I didn't understand this role. And yeah, it basically is just knocking on doors digitally all day long and being told no 99 out of a hundred times. And that's what success looks like. And yeah, it was, you know, early SaaS days, I guess. And I was lucky enough to join a company, my second company out in San Francisco, which I was the first employee.

And we grew to about a hundred employees, went from basically zero in revenue when I joined to about 10 million in revenue in just shy of four years. And then I left, did some consulting, ended up joining, or actually I started a company, my first company, a SaaS company, which was acquired as a good base hit after a couple of years. Then I left in just trying to figure out like, do I want to start another company or not? And ended up joining Zoom in 2020. And so we doubled headcount that year. We can talk a little bit more about that. There's a bunch of challenges, a bunch of fun things, crazy things that were happening in 2020 at Zoom, of course. Lots of things were breaking, lots of problems to solve. And I was leading operations

and enablement for the BDR team there. The BDRs at Zoom rolled up the marketing. So it's people that are doing net new sales, but also doing upsells and cross-sells. So we can come back to that at some point as well. And then, yeah, most recently, you know, started a company, started my second company, SaaS company that leverages generative AI. And I'm definitely in the camp of like, I think generative AI is the biggest technology shift since the internet itself. I think it is actually kind of bigger than mobile. I think it'll touch basically every company and certainly will change SaaS. And I think it'll change a lot of the data world. So I'm sure we'll get into that, but I think it'll also just generally change product building and how people are building products and the interfaces of those, how people

are interacting with those. And from my lens as a go-to market rep historically, or I'm still, I'm still a rep as a CEO of a founder of a company today. I think also for sales reps, the way that they interact with software, with data is going to be very different as we continue to see the AI craze unfold in the coming years. So that's where I'm at today. Love it.

And what's the name of the company you founded? Just so everyone knows. Yes, it's called Groundswell. Thank you for helping me plug my company. Oh, yeah, totally. You're a humble guy. Okay, I am really sad that I am going to save dessert for Costas because I think that he has a bunch of questions about AI. I want to cover maybe two things. So first of all, I would actually like to go back to the days at Zoom during the pandemic when they were experiencing sort of ironically explosive growth in a time when many things were in a state of chaos or implosion. And you were leading a team and you mentioned operations in your title. And so at any company the size of Zoom, operations is going to require a lot of data to actually run

your organization. You described your experience as a data consumer. This is something that we don't often get to hear about on the show. You were a data consumer. You were trying to manage explosive growth and I'm sure running a large team that you had all sorts of data needs. What was that experience like? What were the limitations you faced? And then I think this is three questions,

but I'm bad about doing this. Did that influence your decision to start Groundswell? Yeah. Yeah. So I'll take the angle here of being the go-to market person that is frustrated with data engineers or with data people that own the data. I would say that as a data consumer, another way that I would put it is like black boxed. I think it was unclear to me what data we had, what data we could use, what data we didn't have, but we were able to use. I think a lot of these questions are

probably top of mind more for go-to market folks. I would also say that some of it falls into the category of like, as a go-to-market person, even in operations, like I don't know what I don't know to some extent. I need education, even within my own organization at Zoom, of what data we have. You know, in our world, we were using Salesforce at Zoom, as many companies do, Salesforce or HubSpot. That's like the primary way that go-to-market folks are consuming data in a lot of cases. That's the interface in which reps, you know, sales reps are actioning the data. And so if it's not in Salesforce, like it basically doesn't exist. Maybe there's third-party data sets that we're looking at, but we don't have access to this

magical place called the data warehouse. We don't have access to other tools, you know, like RutterSack and others that maybe product people have, the data engineers have. We just have, you know, what gets pushed to us into our interfaces. And so at Zoom, I think that was an interesting moment for me where when I joined Zoom from the outside, it's like the magic of Zoom is this free product, right? It's this product-led growth motion.

And when I joined and I started like researching like within the team doing interviews with, you know, reps there and with leaders and ops people, it was like, okay, you know, let's roll up the sleeves, like show me this treasure trove of data that is the free users. And it turned out it was actually basically not being used. It was a little bit, but it was very little of that data was being exposed to salespeople. Some of it was by design, but a lot of it was actually just unintentional and oversight. And so I think that was, for me, a very big aha moment that probably only coming from the outside

was it possible for me to see with kind of fresh eyes to say, like, we have to be using this data. This is actually a huge component of what makes Zoom magical and what makes any PLG company magical is being able to actually leverage that data and activate that data within the go-to market motion. For sure. idea of the data that you wanted, because I come from a marketing background, not necessarily a sales background. And so marketers are notorious for saying, well, I want this data, but it's a very ambiguous request. What's interesting to me about your experience at Zoom is you were probably using it every day and your reps were literally using the product you're selling every day. And so any friction point or any sort of experience, my guess would be that you actually had a ton of intelligence about the data that you wanted to use to empower your team with. Is that accurate?

I think so. Yeah, it's an interesting thing to think about. I mean, I think as a consumer of the product, dogfooding our own product definitely helped. There were also, though, a lot of products at Zoom that I didn't even know existed until I joined full-time. Other products within the Zoom suite that I had no idea existed. And so that's probably the case with a lot of companies. As a rep, you're selling something that you've never actually done the job. So you may not know what data is relevant. You're not actually using the product yourself. So I definitely think that that could be a challenge. I would say that this is a bit of a tangent, but as an ops person at Zoom, I was doing operations and enablement.

So basically at the core, I was trying to build out the playbook for the BDR team. And again, these are salespeople that are opening conversations primarily. So they're setting up meetings for account executives, for salespeople. I think because I had been a rep myself previously, I knew to some extent what would be helpful in terms of data. I think you also get into this interesting place that I've seen time and time again. And we definitely had this at Zoom where, you know, I see this often with lead scoring where somebody technical, maybe it's an analyst or some team, maybe it's even a marketing function that builds a lead score to help the sales team. What I've found is like, actually, salespeople do have fairly strong opinions on what they want to see in a lead score

on what a good lead looks like, but they don't know if that did exist, how to get that data, where to look for that data even versus just having a lead score sent to them. And it's like, okay, this person is an 88. I don't really even know what that means, but I don't know, it's some part off score or something. And marketing team threw it over to them and it's like, okay, this person is an 88. I don't really even know what that means, but I don't know, it's some Pardos score or something. And the marketing team threw it over to me and I don't have any other options. So I'm just going to go reach out to this company because apparently it's the highest propensity to buy lead in my book of business. Yeah. Yeah. Super interesting. Can we talk about the interface a little bit. So you said that if it's not in Salesforce, it doesn't exist. And that to me is a very weighty phrase on a number of levels. And, you know, of course,

I think anyone, you know, anyone who is at a company that uses Salesforce, especially our listeners who, you know, have to deal with Salesforce data, it's coming probably into the warehouse or they're dealing with a Salesforce integration, right? It's just sort of, it's the behemoth, right? Like any company that's dealing with leads or accounts is using Salesforce. Salesforce in many ways is like a pain point for so many data teams because it's kind of inflexible, right? Like getting that data into Salesforce is not actually easy, right? The data you were talking about, you know, okay, well, we want a lead score, right? And it shows up as a number in Salesforce without any context, right? And so from the data side, that's actually a challenge because it's like, well, it's actually

pretty hard for me to send you the context for that number because my only option is to send you a, you know, 88 and it's going into a field that's like a number field in Salesforce, right? How do you give that context? What's so interesting to me, though, is that almost the entire world runs off of the Salesforce data model, right? I mean, we can, Salesforce has done a lot of interesting things from a marketing standpoint and, you know, there are a lot of things there, but really at the core, at least my conviction, and I'd love to know, actually, this is a question for you, Brandon, and for you, Kostas, like, I think Salesforce's data model dominance is actually sort of the underlying foundation

of their success, right? So you sort of have leads, accounts, opportunities, you have phases. I mean, every company runs off of this and they can make it as Frankenstein as they want, but like it all relates back to the same like three or four objects in Salesforce that actually comprise what it means to run a business. Yeah, totally. I completely agree. I mean, I think that if you ask most sales leaders, frankly, Salesforce is kind of the necessary evil. And it's become this massive company, I would say, because of the ecosystem, which I think is very interesting, their data model then is used by all of the companies that plug into them.

This is like one of the biggest, you know, every company wants to become a platform. Salesforce is the one that blazed that trail for the last decade and a half, right? They are the platform in B2B that everybody looks to to try to replicate. And what that means then is all of the companies that plug into Salesforce and into that ecosystem, Groundswell, my company, for instance, has to do the same thing, right? They have to fit into the data model that is existing in Salesforce, right? Which is leads, accounts, and contacts, which we all know now is like not really actually a great data model. It is what it is. And it's, you know, it's, yeah,

the other phrase for Salesforce is like, it's the carpet. It's the first thing to be bought. It's the last thing to go at a company. Like it's there. It's not going anywhere. There is no alternative. Maybe HubSpot, hopefully.

Hopefully something else anyway. Probably an AI first CRM. But I do think that to your point, it's the blessing and the curse of Salesforce, right? Is they had such an opinionated data model that anyone can understand it. And when I leave one company and go to a new company, it's the exact same format. And I've personally spun up probably six or eight different Salesforce instances. And it's easy to spin up. It takes a couple of hours and you're basically working. You have a working instance. But I've talked to hundreds of companies and dug in very deeply into Salesforce

and sold companies that integrate with Salesforce and literally 10 out of 10 say our Salesforce is a nightmare. Our Salesforce is a mess. And so I think it's easy to get started, which is the blessing. And then the curse is like, it kind of breaks at a certain point. It's actually not good once you get to a certain scale and then it's your stuff, like you can't do anything about it. So yeah, that's my... Brandon, I have a question. Do you think that the reason that things break is because of the data model or because of how the data model is exposed out there? Yeah, that's a billion dollar question. I don't know the answer. I mean, I can tell

you like at Zoom, for instance, you know, we had a dozen different tools for the team that I was supporting and, you know, reps were supposed to work out of all of those different tools. And, you know, it was pretty standard B2B SaaS sales stack. All of them, including like a Tableau, for instance, is a good example. ZoomInfo, LinkedIn, Sales Navigator, whatever. These all technically integrate into Salesforce, right? So it's like, okay, we don't want reps working out of another interface. Salesforce is the source of truth. Great. There's a Salesforce integration. Okay. Let's buy that software. But actually from an operational perspective, when you look at it, that data is not flowing into the contact object or the account object. In most cases, it's just iframed in. It's just some image that's in Salesforce, but it's not actually on the contact level. And so then what happens is you're basically left with a bunch of different iframes inside of a login in Salesforce.

But reps are still looking at effectively multiple different products within this login of Salesforce. To me, that's like, you know, it was in part the data model or something is broken there where I can't just have it all rolling up to a single account object or a single contact object. It actually doesn't really work that way once you kind of get under the hood. Yeah, yeah. No, it makes sense. I mean, for me, because, okay, Eric, like you also asked me about the data model. And for me, it is like a chicken egg kind of question. I don't know like if the data model is,

let's say what's called the success or the success of Salesforce made the data model dominant, right? Like, I don't know. But what I know is that the data model right now is a kind of mold for Salesforce. Right? Like whoever decides to go and start like a new CRM, one way or another, they will create something very similar at the beginning at least.

And that was like, that was very obvious, like even like 10 years ago, when I was working at BlendO and started integrating with other CRMs, pretty much all CRMs were following the same data model. At the end, it was more of like, okay, how we can build, let's say, a different user interface and whatever experience we can deliver on top of that, right, convert to Salesforce because Salesforce had, and still has like its own, let's say, rough edges when it comes like to the user interface, but at the end, yeah, we still have Salesforce, right, I don't think like the rest of like this long tail of CRM's might not like to do something crazy. So I don't know what caused the other,

but I do know that if you manage to reach a point where your data model, let's say, pretty much defines a whole category out there, yeah, you've done something great. I mean, this is going to be super, super helpful for you. That's my feeling. I don't know. I don't disagree, Costas, but I think that the... Okay, so here's the next question. Actually, this is great. We're turning this into a panel for you and Brendan. Yep.

This is my next question. So I agree. I mean, I don't think anyone disagrees that Salesforce accomplished something pretty incredible with their data model. The problem is that everyone's Salesforce is a nightmare because they are creating sort of Frankenstein iterations and custom objects and all this sort of stuff in Salesforce, which is such a nightmare for data teams and actually go to market teams. I mean, all the crazy customization is happening. This is what's interesting. So the data warehouse is sort of becoming the central source of truth, right? And in many ways, actually, I think the data warehouse, what they need to displace is Salesforce. That may sound like a pretty direct statement, but Salesforce is the source of truth at so many companies.

And so in some ways, the data warehouse is competing with Salesforce. The problem with the data warehouse competing with Salesforce, at least in the context that we're talking about, is that you have unlimited options and data models, right? And that means that there is no opinion and no opinion means that every company is going to do something custom and that becomes very costly at scale, right? And so, of course, Salesforce's data model being opinionated has been a part of their success. It's reached its limit in terms of its utility, in terms of how do we serve context? We talked about the lead score of 88, Brendan, right? How can we serve that in Salesforce with context that's not a crappy iframe that includes all

of this other metadata that's really important. That's so difficult in Salesforce. And like, there's such a large company, can they overcome that, right? But the alternative is like, okay, well, you can build any data model that you want in the warehouse, but then you have fragmentation, you have, you know, people building whatever they want. That doesn't necessarily, like, it's hard for a company or even a business to sort of wrangle that, right? Like, unlimited customization isn't good either. So I guess the question is, if the data model is what made Salesforce dominant, but the warehouse opens up unlimited opportunity, like, what does the future look like from that standpoint?

I kind of disagree with what you're saying, Eric. And I'll explain why. I don't think the problem that Salesforce has is that it's not flexible enough. Actually, you can go and create like whatever custom objects you want there. There is like a database behind the scenes, right? Is this database, let's say, exposed in the same way that like a Postgres database where you connect like directly with with a SQL client is? No.

And for a good reason, because I mean, they're not selling like a database that you can do whatever, right? Like they are selling a platform where you build applications around sales, right? So there has to be some kind of guardrails there. Like, yeah, like, and the guardrails in a way is the core data model, right? Like there has to be some connection there, like to connect it with a context of sales. You can't avoid that. Like, even if you go and build this thing, like on the data warehouse, like you're going to replicate like something similar.

Right? Now the, so I don't think that like the problem with Salesforce is necessarily, let's say the lack of flexibility, like adding new tables or like new columns or like whatever right you can do that let's say what do you think like for example that's like a question back to you Eric

let's say magically we could put like the data warehouse as the back end for snowflake do you think that this would have changed anything outside of like making snowflake like the data warehouse as the backend for Snowflake, right? Do you think that this would have changed anything outside of like making Snowflake extremely happy, I guess? They might be trying to do that.

Maybe. That's a great question. I think, yeah, that's a great point. Maybe it's an interface question. Maybe it's more of an interface question, right? And to add something here, when I'm using the term interface, I'm not talking necessarily about the user interface, right? Interface is like, there are many different layers of interfaces on top of

something like Salesforce that end up giving you at the end an iframe, right? So I don't know if like the UI is the problem. I don't think so, to be honest. But I think we are lost, like in all these different layers of interfaces there between like the different systems and like how we are building like platforms at the end on top of it. While at the same time, we keep control over our market, right? I guess maybe what I would say is, I think that's a really good point.

I think the challenge is that for most companies, when you think about a data team interfacing with a go-to-market team and trying to get stuff into Salesforce, Salesforce certainly has unlimited possibilities for customization. The problem is it's extremely expensive. It's very difficult to maintain. And there tend to be just a few people who can understand these custom things that are built on top of it. And so I think that's why the warehouse is very appealing in terms of creating more flexibility where you don't have the limitations necessary. So I guess that's what's interesting to me about the warehouse and the data models. Basically, if you confine yourself to what's possible with the Salesforce objects, everyone ends up defaulting to the main.

Even if you build something really wild, you still end up defaulting to the main because that doesn't actually work. Yeah. I want to quickly add something and give the microphone to Brenton because he's the guest, by the way. He should be talking more. But I would like to add something, and it is inspired from something that Breton said. Like Breton mentioned at some point that if it doesn't exist in Salesforce, it doesn't exist at all. And I totally agree with him. And although I'm coming from the data infrastructure world,

I'll wear my product hat now, and I'll say something important here. I understand the pain of the data team, okay? Like I totally understand. But when we are talking about building something that's going to be used by salespeople, we primarily have to serve the salespeople. It's the same also like with the marketing people, right?

Now, one thing is like how we can build tooling for the data teams to achieve what they need without like cursing and hating their lives that's one thing right but the fact that we are building for like sales people to me at least means that like we can't replace what the sales people are used to work with with with whatever we think is better, although we're coming from a completely different world. Right? And I'll finish here with saying something that I noticed in the conversation with Bretta. She used terms like SDR, BDR, AEs, I'm pretty sure that if you go to the people that listen to our podcast, and I'm not saying something bad about them, they won't know what these roles are. And that's just

an indication of what a complex domain sales is. And we have to respect that. I'll stop here and give the microphone back to Brendan. Brendan Duggan- Yeah, let me just say, I mean, I think from my perspective, this is a, you know, I would say before I started Groundswell, but certainly through the experience of starting Groundswell, I think that salespeople don't care too much how they interface with this data, but they do need a source of truth that they can trust. And so from my non-technical, I would say, perspective, Salesforce is underlying database, fundamentally the database. And then on top, there's a UI where you can build workflows on top of it, right? I think that where we're seeing kind of this source of truth

shifting from Salesforce to the data warehouse is, okay, so maybe there's more, especially in the PLG world. For instance, Zoom, every signup did not go into Salesforce. I know that's the case with, you case with companies that have millions of signups. Without naming names, you can think of these PLG companies, they don't have them all automatically created as leads or as contacts in Salesforce, right? Because it's expensive

and because there's only so much you can do. And a lot of those salespeople aren't actually going to do anything with them. However, those are like very interesting things for the marketing team and for the sales team to know about. And so I think that the question then becomes like, okay, if again, if Salesforce is a database, and then let's just call it a UI and workflows that you can build on top of that. If the source of truth, the database does shift to Snowflake or a data warehouse. What does that then mean for the UI on top of it? Can there be this thing that emerges that is a UI on top of the data warehouse where salespeople and marketers interface with the data without being technical people?

And I think that's like literally a UI, like how am I as a sales rep able to understand who are the people I'm supposed to go after today, but then also building out automated workflows on top of that data. And I think that is an interesting question. I mean, just to go back to it, Eric, I think like it's confusing to me why Snowflake wouldn't try to go after Salesforce. I don't know. They have some partnership. It's above my pay grade, but like to me, it's very obvious. why Snowflake wouldn't try to go after Salesforce. I don't know. They have some partnership. It's above my pay grade. But like, to me, it's very obvious. Like, that's a very big opportunity.

If I were Snowflake, like, I would just build a UI on top of my database. And like, I don't know why you need Salesforce then. Like, if all the billing data is going back to the data warehouse, if all of the, you know, whether it's a customer or a not customer, you know, basically every, you know, previous activity, marketing activities, that's all going back to Snowflake. I'm kind of now left with like this clunky thing called Salesforce that serves basically no purpose. And so I think that, I don't know, maybe I'm just not smart enough, but to me, like I would

be very threatened if I was Salesforce, I would be threatened by Snowflake and other data warehouses. Because at the end of the day, it's easy to build a UI on top of the database. I think the database and where the center of gravity is by far the most important question and by far the biggest moat for a company to build. And so I think that's the interesting thing for me is like, what does that look like? You know, five years out, if more and more companies are putting more and more data into the data warehouse, I think at a certain point, Salesforce gets squeezed. Yeah. And it's hard for me to imagine a world where that doesn't exist. And then,

yeah, we can talk about generative AI, but I think that also kind of breaks things. Yeah, let's talk about AI a little bit. So I'm going to continue with the panel thing here. So this is a concept that has interested me for a long time. If you think about go-to-market in general, right, that generally includes like sort of marketing and sales in whatever forms they exist, right? If you think about the warehouse becoming the source of truth for all this information,

right? And in many ways, like what removes the moat from Salesforce is the ability to pull their data model into the warehouse and then build things on top of it, right? I mean, that's, you know, ultimately sort of the big threat, right? Is that you have this, but we can pull it in and we can sort of build on top of it. What becomes interesting to me is that you have this central source of truth that is not dependent on the Salesforce data model. And then you can almost imagine SaaS tools as a set of endpoints or interfaces that are actually just consuming input or output from this very large data set and ideally models that are sort of making decisions

on a layer that someone's configured logic in, right? Like, that's really interesting, right? Like, even if you think about, okay, when does a rep need to reach out to someone? Or when does a, you know, a nurturing email need to be sent? There are entire publicly traded companies, huge fortune, you know, 500 companies built on like building email campaigns, right? When you think about all that data living in the warehouse, those companies actually in the future just become an API endpoint where they're being sent a signal that says, you need to send this user this message at this time, or that's going to a rep or something, right? How far away do you think we are from that?

Or do you think that's likely? I mean, I think that's sort of where things are going, where the delivery mechanism is an endpoint and the logic is mainly managed on sort of a layer on top of sort of the central repository. Yeah. I mean, I'll give my take really quickly on that. Like I'm seeing this already. Like there's companies, I think that like, you know, the guys over at Inflection, if you know them, they're building basically Marketo, but on top of the data warehouse. I think that's a very obvious trend that, again, you don't need Salesforce or HubSpot in that world. And so I think that's definitely going to happen. I also think that like where the logic lives is where the value will accrue.

And so I think that's actually a very big question that I don't hear a lot of people talking about, but where you actually build the logic. So as an operations person, I'm building logic in Salesforce. I'm building reports to send to my sales team. I'm building automations to trigger emails, whether that's in outreach and sales tools or Marketo and marketing tools. And I think that logic is the stickiness of the platform as well. And so I think that then the question becomes, okay, if I'm not building that logic in Salesforce, even if it's kicking off, you know, a workflow

that's in an external tool, let's say, you know, Marketo, for instance. Where is that logic going to be built in the future? And can that logic be built in, you know, something that is in the modern data stack somewhere that is outside of the Salesforce ecosystem. And I think that also in kind of an AI-first world, which I believe is an inevitable future, I think the other thing is like, okay, logic is kind of 1.0. The 2.0 version of that is machine learning models, right?

It's actually a feedback loop into, okay, we had logic, it kicked off a marketing email or it surfaced a lead to a sales rep. What happened? What was the outcome of that email? And then feeding that back into the model. And I think this is where there's gonna be the most value over the next five years is in those machine

learning models. And I think if Salesforce doesn't have that, it's going to be somewhere in the modern data stack, or maybe there's some third-party tool that can own those kind of machine learning models that again, are ingesting data from all of the different places and then kicking off workflows or kicking off actions in third-party tools. And I think that's where that logic lives and where those machine learning models are housed is, I think, up for grabs as I see it right now. I agree. Kostas? Yeah, it's interesting.

I mean, there are a couple of different things that are going on here. First of all, there is a very, how to say that, like strong force in this whole conversation. And that's the data itself. Like who owns the data? I think with all this craziness that's going on, like right now with AI and all the conversation about where like, SAP GPT was trained, what data was used,

like how we can use like data to, to do that stuff, like seeing like Reddit or Twitter, like suddenly being like, we're not going to be as open as we were because we're scared, right? Like we built all this data, like we created all this data at the end and yeah, now it's scary, it gets scary. Right. And I think we are entering like a phase, but probably we have entered this phase already where the whole, I think like the war, like the battle, like what's like

the best term to use there, like in the war, like, the battle, like, what's, like, the best term to use there, like, in the market right now will be around the data ownership, right? Now, Salesforce, it's kind of, it's a little bit sad, to be honest, because Salesforce had, like, way too many opportunities

to dominate in many different ways, and somehow the manas like to not do it, and that's not to say anything about, like, the people who run the manas like to not do it. And that's not to say anything about like the people who run the company, right? Like it's just like extremely hard like to do that. But if you think about the Heroku acquisition, for example, they had like an amazing opportunity to become like a cloud provider. Okay.

It didn't happen. Or like go and appeal appeal to developers and provide developers with the tooling to go and build on top of this platform instead of this chaos that it is right now. And then at some point they came out with Einstein or whatever the name of this thing was.

I don't know what it was doing, but in my mind it sits next to Watson from IBM. You know, it's like Einstein and Watson, they talk to each other and they pretty much say nothing, right? And they had a decade of crazy access to crazy amounts of data to go and build. And we can see what's the value out of this. See what Microsoft did with the data from GitHub, right?

So I think there's going to be a lot of fighting around who's going to own the data. I think there's going to be also a lot of like, how to say that,

it's not only market dynamics that are important here, it's also what the state is going to say that it's not only like market dynamics that are important here it's also what the state is going to say like what legal implications will be around that like we are missing you know like the legal frameworks around these things and things might change like a lot when this came out so one thing that is like super important and that's why, Brendan, what you said, why Snowflake is not building this UI on top of it. What I would say is that, obviously, Snowflake is after the data. That's what they want, right? They want the data to be hosted on them.

And actually what they say is, yeah, sure, let us host the data and then Salesforce operates on top of us. Marketo, the same thing, right? Now, I think that people in Salesforce right now, they realize that, no, we have to safeguard this data, right? If we want to survive in the long term. And the weapon that they have, the really powerful, there are the weapon that they have, like, they're really, like, how to say that, like, powerful.

There are two weapons that they have. One is, like, the platform. What you said, all these integrations with all these systems out there, we are, like, 99% of them, like, us, we don't even know about them. Like, all center software, like, crazy stuff that we don't even know that, like, markets exist around them, but they are huge. And the other one, of course, is, like're saying about the mode of the data model, which

is more of a cultural thing in a way. We have a whole army, a whole generation of salespeople out there being educated around that. Yeah. Man, that is fascinating. Okay. So, yes. So many thoughts there.

Brendan, do you have thoughts? Because I have another question, but I want you to respond to what Kostas said. I guess the question is like, this is maybe a dumb question, probably is. But by the way, I totally agree. I don't know what Einstein is, but it seems like it should have been the AI play for Salesforce to have a bunch of data. I guess technically, though, they have a bunch of data. They have customer data, for sure. But I don't know that they have access to third-party data flowing in the same way that Snowflake does into data tables sitting in the same place, multiple different data sources coming into the same data table. So I don't know technically that they could have trained models off of data that flows into Salesforce.

Is that right? Or technically, could they? I mean, I don't know, but I would assume that if they wanted to extend their platform in a way that it could accommodate that, they have the luxury of time and the position to go and do that. Right? So, yeah. Do they have it? Probably not. Should they have it?

Today, we say yes. Maybe five years ago we would have said something different. I don't know. It seems like afterwards to have an opinion on that stuff. But they were already a big part of the power that they

have is the platform itself. So why not extend it outside of the service also to the data? And there is like a lot of data that is generated in there, right? Like when you have, like I remember, for example, when we started syncing data from Salesforce, there were like all these hundreds of like tables that we couldn't figure out what they were. And they were actually tables that were generated by applications of like people who were like

in stopping. Like call center software, for example, right? the tables that were generated by applications, like people were like installing, like call center software, for example, right? The call center software was directly like adding data there. Now, obviously, it wasn't like the whole spectrum of data that could be generated there. But in a way, it was happening. And it was happening like to accommodate the interoperability needs of like the platform and make it like work better, right? So what technically could happen with training on top of this data?

I don't know. That's something that I would like to ask you now with the experience that you have with AI and what this data can do for the salespeople if you combine it with AI. Yeah, I mean, I think that's, I think, I don't know, what is the analogy here? can do for the salespeople if you combine it with AI. Yeah. I mean, I think that's, I think, I don't know, what is the analogy here? Like, you know, whatever, 10 years ago, five years ago, data is the new oil. I think that's like, we're at an inflection point where it's like more data is better for sure. But I think we're at an inflection point where it's like, actually, you don't just want to

get as much data as possible. You need to figure out how to action that data, right? And the people that make money on oil is, you know, the oil doesn't make the money. It's like the refineries that do something with the oil that actually make the money. And I think that's where we're at now. That's what we're trying to do with Groundswell is like, okay, there is all of this data in all of these disparate data sources, whether it's Snowflake, Salesforce, third-party tools like ZoomInfo, and LinkedIn Sales Navigator, many other tools that go-to-market folks are looking at,

marketing automation, et cetera. I think now is the question of, okay, how do we actually do something interesting with this data? And technically, it's quite difficult to do that. It's hard to make sense of all of this data that lives in these disparate siloed places. And so then the question is, okay, can you point these new tools right at the end of the day, like generative AI is just a tool, can we just use this tool to point to a bunch of these different currently siloed systems to not necessarily pull it all into a central place and action it, but can we point and say, hey, go collect data if it looks like this, or go kick off a workflow, kick off an email campaign if that certain action takes place in that siloed system.

And I think that's one of the big unlocks that I believe we'll see in the coming years. And I actually think that, interestingly, I think that there's maybe a world where you don't have to centralize all of the data into Snowflake. So this is kind of a little bit of a different point, but I think that there's a world where generative AI actually unlocks the ability for these decentralized systems to exist, but your ability to actually take action against them at scale using these new tools. So I think that's also an interesting trend where maybe you don't need to just be spending tons and tons of time and resources building out and dialing in data tables in your data warehouse and ingesting all of the data from all these

different places into a single place. I think maybe it's actually, you can just have these autonomous agents going out and fetching data from these decentralized sources. And then there becomes an interesting question from the business side of what is the value of those decentralized places? I think then it does come back to data. I think it is like, okay, the places with the most interesting data and data exhaust are going to be, the value is going to accrue there. And it's less about the interface of those softwares. Yeah. I have a question. It's like a little bit of like a fundamental question, I think. So at the end of the day, what is the needs of the salesperson, right? Because that's why you build all that stuff, right? It's not like for the sake of technology itself. Like it's to deliver value and like help sales perform better or like whatever.

But when you go, and I'm sure you're doing that, like go and talk like with BDRs, SDRs, AEs out there today, right? Like what is the pain that they have that can be solved by data and like all these technologies? Yeah. So the pain is as a salesperson, you're responsible for trying to bring on board customers, right? And so you have, say, 100 companies assigned to you and you're literally handed, you know, 100 logos and they say, okay, you've got the next year to go land as many of these as you can and your quota is a million dollars. The question then as a salesperson is like out of 100, who am I picking today? Who am I going to go reach out to and have a sales conversation with?

And what data unlocks is focus, right? Data unlocks the ability for you to say, I'm going to focus on these three companies this month for certain reasons, and I'm not going to focus on the other 97 companies because these three companies are the most likely to take a sales conversation with me and the most likely to buy because of literally hundreds of different reasons. There's lots of different reasons and those reasons sit in these siloed tools. And so the way that it works today is sales reps are just literally manually jumping into these different tools, researching, trying to figure out, you know, out of my hundred accounts, who just raised funding out of my hundred accounts, you know, who just got a new VP of it. And that's who I sell to. And maybe it's a good time to go reach out to them out of my accounts, you know, who used to be a customer

and went to a new company. And now they're not a customer yet, but I should go reach out to them. Out of my accounts, who used to be a customer and went to a new company, and now they're not a customer yet, but I should go reach out to them. There's all of these different kinds of sales plays that salespeople are running to try to figure out what are the highest propensity to buy leads in my book of business. And so I think that the technology, what technology enables is efficiency, right? It's the ability for you to get in front of the right customers at the right time with the right message so that you're not like as a salesperson having to do this mundane, monotonous, painful work of manually researching across a bunch of different tools. Yeah. I think that's one of the best descriptions that I have and explanations that I have heard about what is value

to sales from technology, to be honest. So thank you for that. That was awesome. And what kind of... Okay, just to help also our audience understand a little bit, what are these sources? Like all these different tools that you're talking about, right? Like, what are they? Like, what kind of data they are like offering?

Yeah, so it's a bunch of different data. I'd say there's a few different categories. Some of the data is sitting, so you can kind of categorize it in two different buckets. First party data, third party data. first-party data being data in your own system. So that's your CRM data, data warehouse data, marketing automation data, sales engagement tools. So these are like email sending tools specifically for salespeople, companies like Outreach and SalesLoft and Apollo. Then you have third-party data tools. So these are tools that have like contact information. ZoomInfo is the biggest company in that space. They're a publicly

traded company, multi-billion dollar company. They've been around for a decade plus. You know, other companies in that space, Apollo, there's many companies in that space. And to varying degrees, there's also companies, those companies and others that are tracking other signals that are relevant potential customers for me, which of those companies have recently purchased Marketo. It's a good time for me to go reach out to them and say, hey, my product integrates with Marketo. Would love to show you how we can make, you know, Marketo better for you. So you basically want to see all of these different type of, I call these signals, basically signals that indicate that it's a good time to go reach out to have a sales conversation.

Yeah, there's hundreds of these companies that are tracking all of these sorts of signals for sales teams. Yeah, okay, that's great. And how is PLG related to all that stuff? Like what's the difference when we're talking about like a product-led growth motion? Yeah, so product-led growth is really, I think, the tip of the iceberg or the tip of the spear around, you know, where the source of truth is moving. At the highest level, like at Zoom, every signup did not go into Salesforce.

So if somebody signs up and starts using zoom, they might be at a fortune 500 company. That signup actually may not literally the sales rep that owns that account is responsible for reaching out to that account may not know that person even signed up much less what they've done in the product or that multiple people have signed up at that same company, right? If 25 people at the same company signed up for a Zoom instance in the last week, right? That's probably an interesting company for you to go reach out to and have a sales conversation and try to serve that customer and see if there are potential products that they may be interested in using or buying.

And so because it's just such a high volume game for these PLG companies, all of the users are not going into the quote unquote source of truth for the sales team, which is a big problem. And then the second level of that is what they're actually doing in the product, right? So are they using the product, right? So are they using the product more this week versus last week? Are they adding more users? Are they connecting to a data source? Or are they doing something interesting? What are the events that are happening within that usage? And those are sitting in, you know, kind of traditional

product analytics tools, right? Whether it's Amplitude, Pendo, Heap, whatever. But again, salespeople don't have access to those tools either. And that's kind of overkill anyway for what a salesperson needs. So I think that again, like the PLG motion just means that there's a lot more data, a lot more users, and that it's just too expensive for all of that to be housed in Salesforce, frankly. And it just breaks the architecture of Salesforce. At a certain point, you need things like time series events.

You need, at Zoom, we cared about week over week, minutes spent on Zoom. That's a good indicator that they're trending up. It's a pretty difficult thing to actually build into Salesforce. And then by the time you get it built into Salesforce, you're like, ah, actually I think I want month over month usage, right? And you have to go back and do it all over again. So I think that's where it really, the problem is exacerbated in the PLG world as it relates to Salesforce being the source of truth.

So what I hear from you is that there's another huge source of more signal for salespeople out there. So in my next question, in my last one, before I give the microphone back to Eric, how is AI helping with the fusion of all these signals, right? Yeah. AI helping with the fusion of all these signals, right? Yeah, so I think there's two primary ways where it's helping, as far as I can tell so far. And again, I think we're very early days here. So in five years, I think there's going to be a lot of things that were not clear right

now. I think that the first way is what I talked about a little bit earlier, which is kind of these autonomous agents that are going out, scanning different data sources and coming back with information, right? So it's basically what humans are doing, you know, the sales development role, for instance, a lot of what they're doing is researching text, reasoning through that text, and then writing a message based on that text. That motion is fundamentally what an LLM is very good at doing is searching through text based content, reasoning through that content, and then generating some output in the case of sales, probably an email or maybe a call script based on what was found in that contact. So that's number one.

And they can just do this at scale. They don't sleep. They don't take vacation. They're not hungover. They're much better at doing this. They're not going to get annoyed after, you know, two weeks of doing it. I think the second one, which is, I would say maybe more fundamental to software as a service specifically is kind of just product building in general.

I think that the UI of Salesforce, if you think about that UI is and how salespeople are interfacing with it. You know, as a sales rep, like you are looking at a list, you're sorting that list, you're filtering that list, you're doing all these clicks and drags and drops to get to somebody to reach out to them. And I think in the generative AI world, what you're really going to be able to do is, you know, chat UX is a good example of this, although it's just one, but I think you're going to be able to just ask the

system, the question that you want to know, and it's going to come back with the answer versus again, today, the way that it works in Salesforce is I have to click and drag and drop and filter and sort, and then I get my answer. And I think in the future world, as it relates to building product, there's going to be products where you can just interface with a chatbot or maybe the chatbot just presents you with information every morning and says, these are the interesting things for you to care about versus having to actually go, you know, find that data. It's just going to proactively send you information that is relevant for your role. Yeah, that makes total sense. Uh, all right. That's all from my side, Eric.

Oh, you're lying. You have more questions, but guess we're getting close to time. Yeah, actually. So Brendan, my question is how do you build trust with the person who is receiving the signals? And I'll get very specific here. As a data team, we have sort of a tiger team that runs data at RutterSack. And I'm involved in that team. And we send a lot of signals into a tool called Sixth Sense that sort of collects our sort of comprehensive

like marketing website usage data, product usage data. We send it into Sixth Sense. They combine it with other intent data and sort of they can create composite scores and other interesting things. It's a very powerful tool, actually. It's amazing how accurate they are in terms of accounts that are interested in maybe a certain product or whatever. But not all of our sales reps trust it.

Yep. And so when I think about the level of detail that the data team at Ruddersack went to, to send data into Sixth Sense and sort of like my personal level of trust of their like heat score, say for like an account or a lead or whatever, I actually have a pretty high level of trust because I think the fidelity is pretty strong. How are we going to deal with a world where,

even if the RevOps person and the internal data team are saying, guys, this account is good. You need to pounce on this or whatever. And that's a struggle in this world, right? But then you have generative AI obfuscating that even more, whereas I feel like we make data-driven cases all the time to reps. And this isn't a dig on salespeople.

I'm just saying this is the natural course of a lot of things. Salespeople want to know that what they're dealing with is solid, right? The worst thing they can do is waste their time. That's the absolute worst thing they can do. And so when you have a generative AI tool making recommendations, you kind of have one, maybe two chances to sort of get it right. How do you think about that in terms of this tool becoming a useful thing? Because it's hard, even if you show everyone all the data and prove to them where this is 100% right, you know? Yeah, I mean, I do think that it's important. So two things.

One, I think as a go-to market person, it's my duty to anyone listening, like involve your go-to market team. Don't build that in a silo. Be sure that you're bringing in a couple of stakeholders from the go-to market side, including the end users who are going to consume this data, right?

So that might be a sales rep. Go get one or two of your best performing sales reps and pull them into this conversation and let them kind of help you at least towards the end of the build of this thing, or maybe at the very beginning, I don't know, help you think through like what would and would not be interesting data for them to have access to. I think that's super important. I do think the data has to be transparent. Like I think that these black box algorithms, I'm actually quite bearish on.

I think they're not a good thing, especially in this context. Again, I'll go back to lead scoring and it's similar to the example you just gave. We had this at Zoom. So we had a lead score that was quite sophisticated, actually. And it was built by very smart people. The data science team at Zoom built it. And when I joined Zoom, I was like, okay, cool. There's this lead score that I've seen in Salesforce.

It's just a score, but how is this made up? There was no context. So that's number one is like, you need the context of why something becomes an 88. Okay. That's hopefully most people are doing that to me anyway, that's like table stakes. But number two is like, how do I go back and look at this kind of cookie crumb trail and understand what is that score comprised of. And it actually took me, I got pinged around to multiple people until I finally got to some Google doc that was like, here's what the score is made up of. And what I realized is, and again,

it was very sophisticated. It was good. It had multiple different variables and weighted averages against them. And they were pretty good. The problem was by the time it finally got in front of the salespeople, there was a couple of misses early days. They kind of were like, I don't know if I fully trust this. Then it wasn't used. And then it just dropped off. And so very quickly, there's no feedback loop to improve it.

Because of course, like your V1 is not going to be great in the same way that a product isn't going to be perfect. You're going to iterate against it. And a lead score should be the same thing. But you're exactly right. If the trust is broken early on, like, that's it. As a salesperson, I don't trust that score. I'm done. And then it's never going to improve because I've already lost that trust. And so I think that it's actually super critical in the same way that when you first test the software, like that first mile product experience is so critical. So if you drop off, like I'm just done with that software, I'm probably not going to go back to that software. And so I do think that again, it needs to be open. It can't be a black box algorithm. There needs to be context sent along with whatever it is where the data is sent. And then you need to involve the end consumer of it. And then the fourth one is you need to be able to actually get buy-in from them and then improve it over time. So there needs to be that feedback loop, which I, this is not, you know, I say that about Zoom. I've seen this time and time again, where some marketing team or some

data team builds a lead score and then it just isn't used by the sales team. So I think that it's yeah, it's a very common problem. And I do think that generative AI has to become a little bit more transparent. I think you're going to have to, I think these open models are going to have to exist for people to be able to trust like a chat UX in the way that you need to, especially when you're interfacing with actual customers. I love it. Well, Brendan, we're over time. I told Brooks this is going to happen because I knew it would. But honestly, I feel like we scratched the surface.

I think we need to get much more into generative AI. And so we'd love to have you back on soon for a part two, just to dig into that piece of it in particular. But this has been really amazing. I think we've learned a ton about go-to-market. And I love the conversation about data models and just sort of how the warehouse is going to impact things. So thanks for joining us.

Yeah, I appreciate you having me on. It was a blast. What a fascinating show with Brendan Short from Groundswell. Costas, I believe that's the first guest that we've had that's building a company that is sort of a bet wholly on generative AI. Is that right? Am I remembering correctly? I think so, yeah.

Yeah. We've talked a lot about AI and ML workflows and ML ops, but it was really cool to have Brendan on there. He's a go-to-market guy, actually. So he's a founder and CEO and has a long history on the sales side, believe it or not. I mean, I guess maybe we're making, because we covered a lot with Brendan, but two big takeaways. One was certainly on the AI side, where his vision for what he believes to be possible for his company using generative AI was really cool to hear about. It's easy to get caught up in the technical side of it. You and I talked about the ethics of it recently. The infrastructure side is really interesting, but he really believes that he can make things drastically better for

salespeople using generative AI, which is super interesting. But perhaps the even more interesting conversation was around the supremacy of Salesforce's data model. And that is really one of the topics I think that on the show has been... I thought about it every single day the week after the show, because you can say a lot of things that you want about Salesforce, but the lead contact account opportunity model rules the entire world. And we had a fascinating conversation about whether it's possible to dethrone the Salesforce data model that has become king. So it was fascinating. Definitely one of the more intellectually simulating shows that we've had

in terms of sort of broad market stuff in a while. Yeah, 100%. I think it is like a super interesting conversation that we're having, not only because of generative AI, which obviously is a very fascinating topic. And it's very interesting to see people trying to build in this space today. But also because Brendan has a very interesting background. He is not a first-time entrepreneurial founder. He has done this before.

He has done it during the previous iteration of innovation that had to do with SaaS and cloud. And it's very interesting to hear all the commonalities between the two eras in terms of innovation back then and now. What is similar, what is different? And there's a lot to learn actually from all the stories that he shared about Salesforce, about SaaS, how Salesforce became such a dominant product out there. And we also talked a little bit about what can happen to, what can potentially dethrone them, right? How this can happen? What are the threats against Salesforce.

So it's a very interesting show. It's all about data, again, but in some very surprising ways. So I think everyone is going to find it a very interesting episode to listen to. I agree. So do you think that Snowflake can become the new de facto?

Can they create the data model for CRM? We discussed that on the show. Maybe that's too far. That sounds like a Shop Talk episode. Yeah, don't disclose that much. Let's keep it cool. I won't disclose it. No, I just wanted to drop a juicy, enticing nugget.

But yeah, definitely listen to this one. Really fun to talk to someone from the go-to-market side who has studied data and is building a startup on generative AI and using intended data. So definitely take a listen, subscribe if you haven't, tell a friend, and we will catch you on the next one. We hope you enjoyed this episode of the Data Stack Show. Be sure to subscribe on your favorite podcast app

to get notified about new episodes every week. We'd also love your feedback. You can email me, ericdodds, at eric at datastackshow.com. That's E-R-I-C at datastackshow.com. The show is brought to you by Rudderstack, the CDP for developers. Learn how to build a CDP on your data warehouse at rudderstack.com. you