CEO panel: From Systems of Record to Systems of Action — CEO Roundtable

Julia Nimchinski:
And now we’re back with our favorite GTM analyst, Chief Strategy Officer at Sendler, Seth Morris. Welcome back to the show. How have you been? What’s new?

Seth Marrs:
Good, yeah, long time to talk.

Julia Nimchinski:
Let’s get into it. We have a CEO panel, all-stars here. Take it away.

Seth Marrs:
Yeah, sounds good. Okay, so just, just kind of set the scene for this one.
So, like, for your CRM’s been this… what I would say is a system of record posing as a system of action, and I think AI has really magnified these gaps and created a new set of technology that can deliver… actually deliver on the process of a system of action, and I think when you look at the hard skill exchange Trends and predictions report, you saw that AI adoptions growing rapidly.
But very few companies are actually embedding this in the process, so now you’ve got these… this opportunity to build this system of action, a lot of people that are interested. And now it’s about how do you push people into understanding what this works… what this looks like, how it works, and today we’ve got a group of CEOs that are at the center of making it happen.
So, I’m looking forward to… to helping share their expertise, and just… just to get started with this, let’s start with… with introductions. So, if you… You could just say the company you work for, and how your business is positioned to help transition to this system of action. Tuba, would you mind going first?

Tooba Durraze:
Yeah, no problem, thanks for having me. So my name’s Chuba, I’m the founder and CEO of a company called Amoeba AI. By background, I am a data scientist who studied neurosymbolic systems. I did my PhD looking at how machines can reason in complex environments. The problem I became obsessed with wasn’t a technical problem, it was actually an organizational problem, funny enough.
Every company I worked with at that time had more than enough data, more than enough tools, but the time to decision-making was faltering because of different interpretations and no shared clarity.
So that gap between data and decisions is what led me to believe, to build Amoeba, and We’re building what we feel like is the next layer in that architecture that you’re speaking about between system of record and system of action. What is actually the system of intelligence that drives that? I’m so happy to be here.

Seth Marrs:
Fantastic. Great to have you. Michael, how about you?

Michael Litt – Vidyard!:
Yeah, thanks for having me. Michael Litt, I’m the co-founder and CEO of a company called Vidyard, and we spent a decade helping businesses make customer communication more human and more measurable and more effective through video, which is very much in the realm of system of action.
I think what makes that hyper-relevant right now is that companies are moving from systems that mostly store information, as you kicked off with, you know, CRM is kind of the canonical example of a system of record, to systems that actually help teams do something with that information.
And in go-to-market, that means helping sellers and marketers create better outreach, personalized communication at scale, understand buyer engagement, and I think this all kind of sums up in the conversations behind customers and various platforms in this… in this concept called Next Best Action. What’s the next best action that you want an AE or a go-to-market professional to take?
So we sit right at that layer of execution, help teams turn, communication itself into something actionable. So where AI can help generate better messages, video can help create more trust, more clarity, and the engagement signals that you can get from video views, can guide better follow-through.
And so for us, the system of action is really about turning that customer interaction into something scalable, measurable, and ultimately operational.

Seth Marrs:
Fantastic. Yeah, interesting. I mean, the spectrum here is really, is really cool. It should lead to a great discussion.

Michael Litt – Vidyard!:
Cool.

Seth Marrs:
Alright, Sam, you want to go next?

Sam Liang, Otter.ai:
Yeah, sure. My name is Sam, I’m founder and CEO of Otter.ai. some of you might have used Otter. We created the AI Meeting Assistant category. I’ve been working on this for 10 years now. we actually think about, you know, to enable a system of action, we need to have enough context. So we created a system of record first. by focusing on meeting data.
Traditionally, most enterprises, lose basically almost all their meeting data, so we help capture the data, organize it, synthesize it with, additional data sources, like emails, Slack message, a Google document. So, we are creating a conversational knowledge engine.
So, moving beyond just the transcription, we’re talking about a conversational knowledge engine, which actually, included both a system of action, I mean, both a system of record and a system of action. So, so far, we’ve got more than 35 million users, with over $100 million in ARR.

Seth Marrs:
So, lots of fuel to power the AI engines and these LLMs to be able to give you insights. It sounds…

Sam Liang, Otter.ai:
Yeah, absolutely, because we see that most businesses are run on meetings, right? So you talk about your annual plan, your projects, you do status update every week, so meetings is where most knowledge workers spend most of their time in. So we see that it’s super critical to capture the data, share the data, and then enable AI to take actions.

Seth Marrs:
Got it, got it. Matt, how about you?

Matt Darrow:
Well, hey Seth, good to see you again. I have a doppelganger here. Julie, I have to call out, when you guys flash the intro slide, I was apparently the CRO of Gainsight, which I’m not, even though I know Nick well. I’m actually the co-founder and CEO of Vivin. We’ve raised $130 million from Excel, Menlo, and Salesforce to create the most powerful AI teammate for sales.
a teammate who actually works alongside reps in their most critical moments in a variety of different mediums, including fully featured with voice, personality, and facial expressions on live sales calls. I got a lot to say about Systems of Action. I think we’re in good company here, Seth, but I’ll keep my intro short and sweet, and keep it rolling.

Seth Marrs:
Cool, cool, awesome. Yeah, really looking forward to your perspective. Sahil, hey, great to see you again. Why don’t you go next?

Sahil Aggarwal (Von):
You as well. Thanks for moderating this panel, Seth, and good to meet everyone else here. My name is Sahil Agraval, I’m one of the co-founders of Vaughan. We actually named Vaughan after John Vaughan Newman. The famous mathematician from the first half of 20th century, who, came up with game theory and modern computing architecture.
We call ourselves Claude Code for Revenue Teams, so we want to be the platform where work happens for revenue teams. I think as we’re talking about system of action, system of intelligence, there is definitely something that will come above the layer of system of record. And there are companies that are building agents that are doing work end-to-end autonomously.
We are building a platform that humans will use, and will be the platform that humans will log in at the start of the day, and will log out at the end of the day. And that is where, whatever work humans do in 1 year, 5 years, 10 years. Hopefully, they start and end it in our platform.

Seth Marrs:
Got it, got it. So, for everyone that’s listening, you could… you could see from this group, there’s a very diverse set of perspectives. I think… I think one thing this also brings out, like, listening to all of you guys do your introductions, is system of action isn’t one thing. Like, there isn’t a single definition of system of action.
There’s a chain of things that need to happen across… So, why don’t we jump into this just by, like, talking to that? So, I want to have you guys just say what your contribution to the system of action is across this space as you’re helping go-to-market leaders. Michael, you talked a little bit about it in your intro. Can you expand a little bit on, kind of, how does Vidyard fit in?

Michael Litt – Vidyard!:
Yeah, for sure. So, I mean… I think… I feel like I have to start with, like, a summary of the difference between system of record and system of action, just to… just to help with this, but…

Michael Litt – Vidyard!:
You know, system of record, I mean, you’ve called it out, it’s a CRM, it’s where information is ultimately stored, it’s what should be true in the business, it’s mostly passive. And then system of action is what people use to move the work forward. So, you know, it takes the context, it takes the signals, it takes the intent, and turns that into recommendations.
And this is where, like, all the innovation is ultimately happening in AI. It doesn’t just, happen, you know, because a human is going to do that work, necessarily. The system of action can ultimately make those recommendations. So our play here as Vidyard, is that, you know, we are a video messaging platform, people, sellers, marketers, use us to communicate with video across their website.
So, that’s video embedded on their homepage, that’s video that’s sent out by account executives, or that’s video that’s automatically generated using avatars from inputs that we get from the system of record. And so, the whole concept of video, and I think why it’s important in this next generation, is it’s very easy to generate text.
It’s very easy to use an LLM to send an email, and we’ve all been victim from too much email for too long, and I think that’s just gotten a lot worse, thanks to, you know, how easy it is to generate AI slop. So a video in your inbox is a true pattern disrupt. It’s that true human-to-human interaction. In a sales motion, like, you want somebody to be liable for your success.
You want to know that there’s somebody there that’s truly thinking about you, especially in high-value, considered purchases. So video is that, component that can ultimately create trust.
And when we think about the next generation of buyer and seller, Gen Z, you know, this is the generation that’s grown up on Instagram and Facebook and TikTok and WhatsApp, and a lot of these platforms are all about asynchronous video. And so, they expect video to be a part of both the buying and selling process, and that’s where we exist.

Seth Marrs:
So, I mean, your system of action there is you’re enabling sellers to use video in this moment and action it in the easiest and in the flow of work in the easiest way possible, so they can get that channel or get that medium to their customers for engagement.

Michael Litt – Vidyard!:

Yeah, and our opportunity is to, you know, have a script generated or automatically generate that video from the system of record and from that data store, which is constantly informed by all these other systems of action, so that the information, the script, what’s being communicated, the type of video being shared, maybe it’s, hey, we recommend you send this micro-demo of this feature set, hey, why don’t you do a walkthrough of the pricing proposal versus sending it via email.
We can make those recommendations based on all those insights.

Seth Marrs:
Oh, fantastic, fantastic. Sam, how about you? How do you see the system of action and your role in it?

Sam Liang, Otter.ai:
I think, you know, as I mentioned earlier, you know, our belief is that for the system of action to have enough information to take action, you need to have accurate system record first. Traditionally, when people think about a system of record, people think about CRMs, Google Documents. Their tickets in JIRA system. Those are all important.
However, most people overlook the amount of knowledge and amount of insights in meetings. You know, again, although some people start to use meeting note-takers now, majority of the meeting knowledge is still lost. And even if you capture some meetings, in your, a meeting note taker, they’re usually, fragmented and siloed. You know, everyone keeps track of maybe 10, 20 meeting notes themselves.
Usually not shared, right? The sales team may, may use you know, a note-taker to take some meeting notes, but it’s usually capped to themselves. The product managers don’t know, the marketing teams don’t know. So, we see that it’s really important to actually create a organizational-wide knowledge base that capture the meeting notes, the meeting knowledge, share the meeting knowledge.
That’s… the second part is really important. We see that most Meeting notes either not captured, or after they’re captured, they’re not shared with the right people. So we see that there’s tremendous opportunity to create a system that, enables this radical transparency, radical information sharing.
So that almost all meetings, between sales, marketing, product, engineering, design, they should be shared, they should be searchable, they should be, accessible by AI. I should be able to ask AI, hey, to say, what are the top 10 sales meetings that happened today? What are the common customer product request today was last week, right?
Traditionally, you know, if you’re a C-level executive, and you’re a CEO or CRO, you may need to wait a week, two weeks for someone to give you a report. But now, if you think about it, most of the, insights, most of the information. happen in meetings. And also, compared to documents, we see that documents They’re important, but most documents becomes obsolete, really quickly.
You know, the document that you wrote 2 months ago, maybe half of that information is already old. However, when people… whenever people have meetings, they have the most up-to-date information. It’s the most fresh, most up-to-date, so that information needs to be made available to everyone in the company as soon as possible. So we need that system record first, yeah.

Seth Marrs:
So you’re enabling a system of action by making it very easy to capture the information needed to draw the insights that push action forward.

Sam Liang, Otter.ai:
Absolutely, and then after every meeting, if you think about it, there’s a lot of action items generated. in those meetings. Right? So, then, you know, we invoke, agents, or create a suite, you know, invoke a suite of agents, and say, oh, you need to, create a PRD out of the product brainstorming session. Right? That’s an action item out of that brainstorming meeting.
Then AI can go ahead and create that drop for you immediately. You don’t even need to wait. AI can do that. You say, I need to schedule a follow-up meeting with Thess. Okay, then the scheduler agent can go ahead and do it. So, you need to create that system record, then the action can be executed by agents.

Seth Marrs:
Awesome. Okay, so let’s… let’s… let’s go deeper on agents. Matt, why don’t you talk a little bit about your place on the… in the system of action and the stuff you’re doing with your agents?

Matt Darrow:
I’ll, Maybe let me expand upon what Mike was saying, because I think that’s where my head is, too, which is, like, my bar for a system of action is it’s gotta engage with a third party, and that party is either a system or a person, but it has to do so on its own volition.
And I think this is where, okay, a system of record is a data store, and it could be, to Sam’s point, a data store of meetings, it could be, to Mike’s point, a data store of stuff in your CRM, but if you actually want AI to act with its own volition.
Now the bar is raised significantly, and I would say, you can sort of take the veil off of things that are masquerading as systems of action that really aren’t, that don’t meet that bar. Most famously. not to cause a stir here, it’s like Agent Force. Agent Force operates much more like an ETL tool. If this, do that over here.
It’s a very, very rigid declarative structure where you need to spec out hundreds of custom tasks and workflows to quote-unquote take action on things that might be in your record set, so to Sam’s point, like a conversation history. The bar needs to be pushed for the AI to make that decision on its own, and to act on its own volition.
So… Our breakthrough, and I think, too, about not knowing you, just given your introduction, your research, this is probably near and dear to your heart, we firmly believe that if you don’t give AI a world model that is a top-down declarative understanding and structure on how it needs to operate within its domain, it cannot act on its own volition, and you’re going to be stuck in this world of.
Humans creating one-off agents to ask these pedantic tasks in ETL workflows. That is not a system of action. If you think it is, your bar’s too low.

Seth Marrs:
Got it. So, I mean, well, yeah, I mean, to a certain extent, there is a system of action for the sellers to work on, but what you’re talking about is you should have agents alongside the seller that are taking action without the need of the seller to tell them what to do. They both have the same goal, and they’re working towards the same goal. Got it, got it. Makes perfect sense.
Sihil, what’s your take?

Sahil Aggarwal (Von):
I think first, I would definitely agree with Sam that the system of record is actually not CRM, but order is the system of record. Gong or other players who are in this category are the true system of record. For the last 25 years, a seller or a CSM will go on a call, and then will use their own individual judgment, and then fill the system of record, which is CRM.
But now with AI, what we can do is we can just mine the calls and emails directly, and that becomes a system of record. In fact, I have become far more conscious that my note-taker joins my calls. There are 20% cases when it does not join my call, and sometimes I won’t start the call till my note-taker joins it, because the value is insanely high.
So, 100% agreed on how a system of record itself is changing, and going to raw data. In terms of how I think about system of action, I think… We as humans, we know exactly what we want to do. It’s just that doing stuff is hard. Doing stuff takes time, needs analysis, doing stuff needs expertise, skills, tools.
As an example, we organized a dinner last week, and after organizing the dinner, I know exactly what needs to happen. I need to send a personalized note to people, thanking them for coming in. I need to look at our history before, so I reference that history, so it’s not just a cold… it doesn’t feel like a cold message.
I need to reference the conversation that happened at the dinner, and then I need to hopefully ask them to take a look at one. Now, doing all of that will take me 6 hours, if I was to do it right. What AI, or what AI enables the system of action to do is, now I can just tell AI to do it, and AI, if it has all the context, if it has intelligence about me and my business, AI can go do it end-to-end.
So what used to take 6 hours will take 5 minutes. And I think that’s the shift from system of record to system of action, which is… humans know what they want to do at a particular time. If they don’t want to do something, you can put all the recommendations in front of them, but they won’t listen to you.
They know what they want to do, so we… our belief is, is we’ll just make it 100 times faster to do what you want to do. You want to reach out to your close-lost people, to build more pipeline. You want to do sales coaching, you want to update CRM workflows, like. Our thing is that we are a system of action, where you know what you want to do, and we’ll help you do that 100 times faster.

Seth Marrs:
So, I mean, what you’re saying and what Matt’s saying has some nuance, right? Like, what you’re saying is, hey, tell it… tell the system to do it, and it will autonomously go figure out how to do it, and what Matt’s saying is, hey, we both have the same goal, we’re gonna work alongside each other to do it. Both system of action type approaches, but… on different planes.

Sahil Aggarwal (Von):
on different planes, and that’s why I would hate to be a buyer in this market.

Seth Marrs:
Yeah, it’s not… it’s not easy.

Sahil Aggarwal (Von):
Yeah, it’s… being a seller is… being in sales is hard, but being a buyer is 10 times harder in this economy.

Seth Marrs:
For… for sure. Tuba, what’s your perspective?

Tooba Durraze:
I think, I’m gonna talk about the middle, the connective tissue. I think system of record… it’s like IoT sensors. You’re gonna collect data out of anywhere and everywhere, and you’re gonna store it in all kinds of ways for it to be accessed by other systems. Where was it gonna be a thing? Like, data is just gonna be there. It always has existed.
I think in terms of action, like, I lean more, I would say, from my research, Matt’s direction, which is what are the autonomous actions you can create using systems, but Sahil’s point is valid, where, you know, if the human knows what to do. like, doing it faster in a more efficient way is a great kind of thing in terms of system of action.
The one thing I’ll say is maybe, like, 5% of the people in the world know what to do, and exactly how to articulate that. We know that. I think the piece that doesn’t get a lot of attention is, again, the connective tissue between the system of record and the system of action. How we think about it is kind of giving Your data, your business, essentially, like a… like a prefrontal cortex.
The analogy I’ll use here is a kid, you know, when you’re younger, you have the data for, like, there are roads I have to cross. You don’t have enough, maybe, life experiences, but the action of crossing a road, you understand that concept. You understand that, you know, there are enough roads around you.
Only when you get smart enough to realize what is the right time to cross a road is when, like, you are actually allowed to cross roads without holding your parents’ hands. So then enabling that piece, which is knowing what action to take.
When and how, whether it’s human-oriented action, whether it’s a machine-oriented action, is really, really important, and I feel like that piece sometimes doesn’t get enough attention.
So I think increasingly in the AI world, where you’re going to live in the space of, like, kind of offloading a ton of your work to AI, having agents autonomously come up with what is the thing to do, and when to execute it, that middle piece of, like.
the system of intelligence, again, I keep coming back to that, I think it’s going to increasingly become even more important than, you know, how to solve the problem, or what to solve the problem on.

Seth Marrs:
Well, it seems like you’re saying it’s the decisioning around all this, like, all these things that we just talked about, and how do I make the right decision for, like, when I want to use Vaughn versus when I want to use Vivin versus when I want to use Vidyard, like, all of those together? Is that kind of what you mean?

Tooba Durraze:
Yeah, the brain in the middle, which is, like, telling you what the right decision to make at what time, using what data as context, executing how, is kind of the piece that we’re anchoring on.

Seth Marrs:
And do you envision that as, like, a collaborative piece, where the human that’s working in it is working alongside all this stuff and engaging it with agents together to make that work?

Tooba Durraze:
I think, okay, I’m contrarian, I’m, like, I think it’s a compute problem, so I will say at some point. You know, like, it is more a machine-oriented thing versus… Like, humans aren’t, like, you know… take the plane as an analogy, I use this a lot. Yeah, there’s a human in there in case things go wrong, like, but planes are, like, on autopilot most of the time these days, right? Essentially.
So I think I say human in the loop, that’s my version of human in the loop. Someone keeps governance, but increasingly. like, why would you want to monitor, like, a data cleanup effort, or, like, a email send effort? Like, why wouldn’t you want… like, let the machines do our laundry, do our house chores, let them just do those things that we don’t want to do.

Seth Marrs:
Gotcha, gotcha. Fade back, and do the things you want, rather than the things you need to, and let the machines do that type of stuff.

Tooba Durraze:
Yeah.

Seth Marrs:
Cool. Okay, so I’m going to move on to the next question, and this is one that comes up, I hear it a lot, I’m sure you guys hear it too, is, oh, I can’t do any of this AI stuff because my data’s bad, so every time I try to run it, I get these really weird responses, so I really… I can’t progress with these things.
So, it… how do you think these tools and these systems of action that you guys have… I mean, you… you… there’s a spectrum that we’ve just talked about here. how do you handle that, and how… do these tools fix that problem, or do you need to fix that problem before you can use these tools? So, Sam, why don’t you start us off?

Sam Liang, Otter.ai:
It’s actually related to what I said earlier, is, like, in terms of accuracy of data, the… a lot of time, you know, the old data, become, obsolete really fast. It’s just, the businesses move so fast. you know, the customer requirement may change, the, schedule may change, there’s something, you depend on might have changed without your knowledge.
A lot of time, I see that, that’s wasted in enterprises. You know, we… we have a small team, just over 200 people, but there are multiple departments of sales, marketing, design, engineering, recruiting, right? There’s a lot of, cross-functional dependency. In the past, we see that a lot of time, people operate based on old assumption.
Things, you thought you knew, like, you know, you determined months ago, okay, this is scheduled, this is what do we do. But oftentimes, things change really fast. The, the, something that, discussed in one team is not quickly shared with the other team.
So, they, they forget that, oh, something have changed, they, they operate it based on old assumption, they’re based on, they operate based on old schedule, they assume that feature will be done. you know, a week ago, but it’s delayed, and… and then your schedule will slip. So you didn’t know in advance.
So we see that We really need to, make sure the latest information is shared, instantly, not, like, a week later or two weeks later. So that’s really important, right? For CRM, Sahil, so mentioned that, right, the… a lot of a sales rep, you know, they usually manually enter data into CRM, Number one, most people don’t do it very diligently.
They either don’t do it, or when they do it, they enter incomplete data, or inaccurate data. So you… with AI, it’s much more objective, it’s much more comprehensive. You know, right after the meeting, the CRM record is instantly created by AI. You don’t need to manually do it. So that’s… that’s really critical.
And also, on that one technical detail, it’s actually, In meeting… when, when you take meeting notes. You really need to know, who said what. What’s the point of view of Michael? What’s the point of view of Simon? And when you create action items, you really need to know Whose action item is that? You know, you don’t want the AI to assign it to the wrong person, which will create confusion.
So that’s actually… technically, you really need to make sure, you know the speaker name really, really, accurately. So that… that’s one thing we actually focus on to make… really make sure we recognize exactly who said what.

Seth Marrs:
Yeah, so if this works as planned, I mean, data gets better when you take the human out of it. So, remove the humid.

Sam Liang, Otter.ai:
And, and also.

Seth Marrs:
kind of the Hubith point.

Sam Liang, Otter.ai:
Yeah, important to, integrate all data sources, emails, Slack messages, and meetings, and of course, documents, but we see the, there are three major communication platforms most people use today. Email. Slack or Team Chat, or meetings. So these three platforms, the data needs to be integrated. They usually talk about the same thing, just on different platforms, right?
You email someone to set up a meeting, and then you have a meeting, then After the meeting, you may continue the conversation on Slack. So, all three platforms, the data should be all integrated. That… so that you get the most up-to-date, most accurate data, and you have the most comprehensive view of that same problem, no matter where the discussion happens.
So that, that’s, you know, really critical.

Seth Marrs:
Cool, okay. So, I’m gonna ask, like, Sahel, I’m gonna have you do it, and then I’m gonna go to the next question, because I want to get a few more questions in. So, and if anyone else has something they want to say on the data side, I… but I know for sure you do, so, what’s your take on this, when it comes to data?

Sahil Aggarwal (Von):
Yeah, I think there are 3 nodes in this whole thing. Node 1 is just your raw data. To Sam’s point, your call recordings, your emails, your Slack messages, your text messages. Now when humans convert that into a CRM-readable information, which is, like, I will go in and fill close date amount, next steps, MedPEC, That is where the messy data is. That is where the bad data is.
And then… before LLMs, we have all been trying to draw insights on this node, too. And so when I go in and I try to ask, like, hey, which which deals do not have Next Step. I’m relying on a seller updating Next Step on a daily or a weekly basis, so I can use that information to drive my forecast.
the concept of messy data, or incomplete data, actually does not apply anymore, because you can just point these LLMs to your node 1, which is your raw data. And these LLMs can just create opportunities for you, see which opportunities don’t have a next step by going to raw data. So you bypass that whole thing. We yesterday did a webinar with one of our customers from Oyster yesterday.
They have been using the platform for 2 months, and they have stopped building fields in Salesforce now. Like, why would I create more fields in Salesforce when all it is going to be is messy data? And if I’m going to ask AI to update those fields for me, then I don’t need those fields to be human-readable anymore. I just need insights on tell me which deals don’t have next steps.
and give me the 6 deals out of 200 that I can go and then coach people on. So I think when people think that I’m not ready for AI because my data is messy, you just need to be concerned, am I recording my data? Am I using an order to record my calls? And I think as long as you are, you’re good, and you can just point these LLMs to… to do your inside work for you.

Seth Marrs:
So all the structured stuff that everyone’s worked so hard on is becoming less valuable, because it’s not going to be accessed the same way with the same tools, so it can be accurate as long as you capture the pieces Pure, without somebody manipulating it.

Sahil Aggarwal (Von):
Yeah, I mean, there is a system of intelligence angle, for sure, that needs to apply to it. Like, we are working on that. Obviously, Amoeba is solely focused on this, but, you do need that intermediate layer of system intelligence. It’s not just LLMs have a limited token window, that they can go through. But I think as long as you do that, yeah, like.
Yeah, I actually don’t know what future of Salesforce looks like in 5 years, and maybe that’s why Mark Benio wants to call it Agent Force and be all in on agents.

Seth Marrs:
Yeah, probably… probably true. Okay, I’m going to shift and go to you. I want to talk to… I really wanted to ask this question, and Matt, I wanted to ask you this one in particular. So, like. there’s a huge amount of talk around AI replacing sellers, what that’s going to look like, and that there’s going to be fewer sellers going… in the future, there will be less sellers.
How do you feel about that? Like, what is that going to look like? Will there be less sellers, or will there be the same number of sellers that will just be way more efficient?

Matt Darrow:
I think I’m gonna get a lot of head nods, but absolutely, there will be less. I’m not a doomer, where the sales profession is being deleted, but sales is a big term. So I’ll give you, like, 3 buckets, and I’m sure others will want to chime in here, too. The first is. let’s just move left to right in the sales org.
On the SDR side, like, if folks are watching this and haven’t replace their inbound SDR function with AI, they’re very far behind. Like, that role’s just gone. On the outbound SDR side, even folks that are heavily automated, right, this role still exists, it’s still very valuable.
It’s just that the capacity and the ratios are really different, because you can go and up-level the individuals doing that role in the tool. So I still see that as really, really important. And then when you move to the account executives themselves, like.
On the low-end SMB side of the market, I mean, people either already have, or they’re actively trying to just completely, fully replace the AE in that segment. These are 5, 10K deals, and people have successfully done that. like, 2 years ago, they were experimenting with, like, let’s offshore it, and now they’re like, just throw AI at that problem.
So, if you’re in that world, you’re already gone, or you will be very, very quickly. And then if I look up to, like, mid-market and enterprise, this is where I’m not the doomer for the rep. I think the salesperson there is absolutely critical. They’re on-site, there’s customer relationships, there’s complex sales, there’s competitors, and But what’s gonna happen there is the supporting cast.
And the further you move upmarket, the more and more, historically, the deal teams increased in size. You have the AE, quarterback and the sub-product line AEs. You have the SE quarterback and the technical specialist, the industry expert, the value consultant. All of those roles that are the planets that orbit around the AE, those are the ones that are going away, and going away quickly as well.
So, when you look at the account executive, okay, their capacity numbers are going to be expected to increase. The supporting cast is gonna go down. So, last comment, are AEs going away because of AI? I firmly do not believe so, but I think it materially changes the solar system that’s orbiting around them, as well as some of the inbound folks, too. And you see this in, like, the BLS data, too.
If you compare and contrast, like, labor statistics versus, like, sellers versus engineers. Right? Engineers are suffering a much more severe consequence right now and projected over the next 18 months than sellers are, but this is coming. This wave is about to come. Go over, Mike, would you… you’ve got… you’ve got a smile and a head shake, though, too, but…

Michael Litt – Vidyard!:
I was gonna say the number of engineers that have been hired in the last 12 months has actually increased dramatically across the technology industry. So I think there’s something at play here, which I’m sure everybody’s aware of, called Jevins Paradox.
Which is, like, in 1861, the steam engine became hyper-efficient, and everybody was like, oh my god, the demand for coal is gonna drop off a cliff. And the paradox there was that the demand for coal skyrocketed because it became so much cheaper to ship stuff all over the world, and that was the beginning of global commerce.
And so I think what’s happening in engineering right now is people are super productive, thanks to these tools. And so the demand to produce software that performs actions and completes tasks and, you know, where software has never even been dreamed to exist before has increased dramatically. So, I… I like to think about… so that’s, like, my only nuance, Matt. I otherwise totally agree with you.
I like to think about, like, how this applies to go-to-market roles. And I agree, like, I think… something that’s happened for a long time is there’s this asymmetry of knowledge that exists between a buyer and a seller, right? And that’s… that’s the value exchange process.
The buyer understands the problem that they have, they don’t know how to solve it, the seller’s like, okay, we have this technology, we have this tool, we have this service, we have this product. That can help bridge the gap, and that information asymmetry is the sales process that gets solved. The internet solved a whole big portion of this, like, 20 years ago.
People started Googling stuff, figuring out, you know, that answer, and serious decisions back in, like. You know, 2010 was talking about how 65% of the purchasing decision is made before someone ever talks to sales. But they’re still talking to sales, because I think there’s this human-to-human bond, which really matters, especially for high-value purchases.
If I buy your $100,000 piece of software, or million-dollar piece of software, or million-dollar service, or whatever, million-dollar piece of infrastructure. I want to know that you, on the other end of that deal, are responsible for my success, because if this is a bad purchase. and I get fired, I want to know that somebody’s gonna feel bad about that.
And that’s the human connection, that’s the trust thing, that I think is always going to exist in really high-value interactions. Because a machine, an AI, no offense. to anybody here, I don’t think anybody would take it, is really… it doesn’t have, like, human-oriented incentives, right? It doesn’t care if somebody lost their job because they made a bad decision.
Its job is just to sell that piece of software. So, a little bit…

Seth Marrs:
So, I’m going to put something out with… and I want to go to you guys, and I want Tuba to chime in here, because she kind of talked a little bit about where she was going and the machine side of things, but the part that keeps coming to mind for me with this one is… and it goes back to what you said, Michael, around Jevon’s Paradox.
Is it that the jobs… I think you’re gonna see a significant shift, but… at the end of the day, it was more… Forrester wrote a report, The Death of a Salesman, like, 10 years ago, and it got all this news around sales going away, this, that, and the other, and then fast forward 10 years later, and there’s 20% more sellers than there were before.
It’s like the world… has this… like, this way of changing with it, I think sometimes we forget that we’re not gonna be flat… I listen to you guys talk, you guys aren’t flat-footed. The minute something changes, you’re gonna change, and I think that’s gonna happen in… it feels like that’s gonna happen in go-to-market.
Like, Matt, I want to give you a chance to respond on that, like, what’s, like, how do you see that?

Matt Darrow:
Well, that’s why… and I go off of maybe not, like, maybe speculation and conjecture… conjecture maybe is the wrong word. It’s just what people are doing. Like… people are removing and replacing a lot of these roles and ancillary functions already, and that’s where the systems and the technology is today versus where they’re gonna be.
But I do, like, I do, and I want to double down where Mike is going, too, and like, the… the… the intrinsic value of sales is really important for human-to-human interaction in valuable purchases, so I’m… again, I’m not necessarily the believer that the sales agent will talk to the procurement agent, and then everybody will be selling $5 million deals that way, but I do believe it’s different than the 10-year-ago Forrester report, just because of what’s possible now, and how the work can actually get done.

Seth Marrs:
Yeah, it makes sense.

Tooba Durraze:
I think I’m gonna take the floor for you.

Seth Marrs:
You go.

Tooba Durraze:
respond to that. I think… We’re talking about maybe two separate dimensions. One dimension is, yeah, I think solving for the problems of the now, the way we see the world right now, yeah, I think human connection is going to exist.
like, I feel like, to Michael, to your earlier point, one of the reasons inbound sales agents are so successful is because there’s now so much information that exists around products, so that gap is, like, becoming lower and lower. So it’s good for them to kind of do that bit, so then deals can be passed off to AEs, potentially, further down the line.
But in a space where you know, you’re still selling software in the same way that we’re selling right now. The human connection needs to exist, because, for lack of a better term, you do need someone to hold responsible in that way. In the future, where software is there to solve for certain problems, I see very easily a world where it’s like an agent-to-agent marketplace.
Like, why would I need to be involved in you know, someone purchasing, like, HR software, which is, like, listening all of my employees. Like, do I really need to be involved in that? Can I rely on machines? Again, it’s a compute problem to make the best decision for, like, what is the best software for us right now?
And I think most of these things will end up being not long-term contracts that we’re seeing right now, but more, like, a consumption-based or pay-as-you-go type deal. I do see the world sort of moving in that direction, but I think a lot of the controversy for us exists because we’re trying to apply that lens to solving for this world right now.
In this world, I feel like humans do need to exist in that interaction.

Michael Litt – Vidyard!:
Yeah, I totally agree with you, and I also think it’s, like, software sales is one category. I think agent-to-agent, marketplaces, PLG, but then, like. you know, if you’re selling… I got a friend who owns Canada’s largest excavating and earth-moving and infrastructure project company, where, like, they’re bringing in bulldozers and, you know, hundreds of people on a job site.
Like, you’re probably not buying that contract through a bidding process, through a portal to a government. you know, using an agent-to-agent interaction. You want to see the person, the CEO of that company, who’s responsible for that $40 million infrastructure project. And, like, know that if it doesn’t go well, you’re gonna see them in court.
And, like, I just don’t know where agents and these digital services live.

Tooba Durraze:
AI, that does change if those are not people doing the job.

Michael Litt – Vidyard!:
Not robots, but I think they’re still gonna… yeah, I… yeah.

Tooba Durraze:
At some point in the future, I’m not talking about… I don’t want people to come for me on LinkedIn, I’m not advocating for, like, human.

Michael Litt – Vidyard!:
I’ll be living on Mars by that point. I’m gonna be digging holes. robots.

Seth Marrs:
Well, let me throw something out to you guys. will any of you allow that to happen? So here’s the thing that, like, you hear this all the time with AI. Oh, you know, if I have perfect information, and I know exactly what to do, then I don’t need a person to do it. I could just let an agent do it and talk to the buying agent.
I don’t… if I… if I cornered each one of you, are you gonna let that happen? I’m not gonna let perfect information go out about my company. Are you? Like, do you really want to work in a world where you’re forced to basically compete only on features? Like, do any of you really want that? Like, that kind of gets a little bit, like, I don’t… I think a person’s gonna stop this from happening.
Like, where do you guys sit?

Tooba Durraze:
But why, Seth, why? Okay, you live in a world where you’re talking about, like, I need to make money by selling software. In the ideal world, you’re probably on a beach, in a cabana somewhere, right? And not having to sell software.
I’m saying, I’m saying there will exist a time where, again, selling, not equating selling software to doing laundry, there will come a time where, like, this is kind of, like, the means to make money are going to be, you know. Not what they are right now.

Sahil Aggarwal (Von):
It does assume that, software itself is commoditized. If you look at Otter versus Gong versus… there are a dozen note-takers, Sam has his own personality in the way that he built Otter, and I think… that needs to shine through. Whether that shines through via an agent, whether that shines through a human, it’s very hard to say.
But I am going to pick the note-taker, not just because of features, but how I feel when I log in. Like, is it speaking to me? And now, in a world where software becomes commoditized, maybe yes, and every piece of software looks the same, and then we’re just all competing on features. But Claude code is very different from cursor.
not because of the model that it uses, but, like, it’s a CLI, it’s a command line interface, and Cursor is a chatbot that worked, like, 2 years ago. My simplistic take on this is what will happen is that we all need to 5x to 10x our seller quotas. That needs to happen with AI. We ran a survey where we looked at 1,200 sellers’ calendars and Gong calls and recordings.
And we saw… 1.2 hours, that is what a seller spends on external calls in a day, on average. And Sam, you probably have so much more data on this, and you should run it across your 34 million users.
I have been on conferences where I’ve spoken in front of 200 people, and I’ve said, like, if your sales team is spending more than 3 hours per day on average on external calls, I will buy you courtside tickets. And you know how many people have taken me up on that? Nobody.
So… and we are all CEOs, we are leading sales teams, and if your team is spending 1.2 hours per day on average on external calls, then we need to fire half of our sales team, and give that pipeline to our top 50%, and increase their quotas by at least 5x. And AI should allow them to do that now. Like, they should be taking 4 to 5 hours of external calls a day. That is what needs to happen.
Like, maybe AI replaces sellers, maybe it does not, maybe the number of sellers increase, but we need to 5X their quotas.

Sam Liang, Otter.ai:
Yeah, I totally agree with that. The, The… if it’s a computed transactional sales. Right? It’s very simple. AI should do it. You don’t need a human to do it. I mean, they can just buy it online, just use their credit card, right? Then maybe a chatbot can help them. Or actually, AI voice agent can help them as well.
However, you know, for complicated product, you know, larger deals, million-dollar deals, in the next, I don’t know, 5-10 years, probably you still need human to help. Beyond 10 years, we don’t know. AI can probably do everything. But a big part of doing sales is actually, building trust. building human relationships. And also, a big part is not just sales.
The best salesmen actually truly understand the customer’s problems. Truly understand the customer’s problems. Then they say, hey, you know, whether this product really fix your problem, or I need to customize our product to fix your problems.
So that requires… still requires a lot of human intelligence, and again, you know, 10 years later, I think AI can do those too, but we’re talking about maybe the next 3-5 years. You still need a human to do those, and to… to Hugh’s point, it’s the… the lot of the mundane work, the grunt work, can be re… can be taken care of by AI now. You know, the sales rep should be talking to customers.
a lot more, you know, the 5 hours, 6 hours, if not 8 hours every day. You know, that’s what the human is good at, is to build the human relationship and build trust. And all the, you know, mundane, tedious work can be taken care of by AI completely.

Seth Marrs:
So, a question for you, for all of you. isn’t… If we talk about this world, isn’t a relationship a defect of humans in the buying process? I mean, in reality, right? I mean, if we’re talking about this, is the… what does the relationship do to help me get the best product? I mean, and I’m a person that believes in the relationship side, but, like, if you’re really talking about this.
It kind of feels like a relationship’s a defect in the buying process that leads to a suboptimal purchase.

Tooba Durraze:
It can’t.

Seth Marrs:
scary to me.

Tooba Durraze:
Yen.

Seth Marrs:
Wait.

Michael Litt – Vidyard!:
Yeah, it can.

Tooba Durraze:
I think it can.

Sahil Aggarwal (Von):
You know, my favorite restaurant is not your favorite restaurant. like, we are all unique people. Every human is different from each other. We might have some shared beliefs, but we might differ on other things. You might be a Republican, I might be a Democrat, you might be an atheist, I might be a God-loving human being. I think that’s… Okay, Opus 4.6 is smarter than I am.
These LLMs have crossed that threshold 2 months ago. For me, probably, like.

Michael Litt – Vidyard!:
years ago.

Sahil Aggarwal (Von):
Yeah.

Seth Marrs:
I’m with you, I’m with you, Michael.

Sahil Aggarwal (Von):
No, the… I… I don’t think I’m even scratching the surface of Opus 4.6 or GPT-54. And these models are definitely smarter than I am, and they will only get better in 3 months, 6 months, 12 months. I think our edge as a race is… obviously, like, I don’t think AI… I’m not a doomer that AI will take over humanity, I think it’s just trained differently, but my take is that, like.
our edge is our creativity and our differences, how I think is different from how you think. My favorite restaurant is different from your favorite restaurant, and I think that’s a beautiful thing.

Tooba Durraze:
You like a plant, I like a plant, I see plants in your background. We might go plant shopping together, even though we’re not experts on plants. And then the LLM would be the expert on the plant, right? I might enjoy going with you to the plant shop versus the expert on plants. I think human relationships, even if the outcome is subpar.
sometimes that… The relationship supersedes that subpar outcome.

Michael Litt – Vidyard!:
I think you’re talking about…

Seth Marrs:
Yeah, go ahead, Michael.

Michael Litt – Vidyard!:
I was gonna say, I think we’re talking about brand now. Like, we don’t always… you know, the t-shirt has been commoditized, but we don’t always have a t-shirt.

Seth Marrs:
So is this a… so, okay, I’m gonna put one more thing, and I want to get your guys’ feedback. It’ll be the last thing we can talk about. I get that completely in B2C. Because B2C is about my personal preferences. Every one of us on this call has one goal. Grow our companies. grow our company.
So the… the crazy thing about B2B is that we all, even though we’re all individual people, the end result for all of us is we’ve got to grow our businesses. So I think B2C side makes a ton of sense. Can we be individualistic in a B2B world? We’re all trying to help our customers do that one thing, which is grow. What’s your guys’ perspective on that.

Michael Litt – Vidyard!:
And I think it goes back to brand, like, the cult of CEO, you know, CEOs being more vocal on social, think about… you know, now Elon is obviously selling consumer products, but also has, you know, large B2B plays in the context of other companies he runs.
You know, people will… buy and sell often on the brand of the CEO, what they stand for, what they believe in, so I do think that, you know, the same humans that buy running shoes and go to restaurants are the ones that are buying enterprise solutions, and… and brand, connectivity, what it means to them. Like, think about, you know, Dreamforce versus a HubSpot Inbound.
These are two very different experiences, two different brands. based on the founders of those companies and what they ultimately built there, people have preferences. Do you like orange? Do you like blue? I think all that… all that fundamentally matters.

Seth Marrs:
Makes sense.

Matt Darrow:
I think that’s the magic of B2B, too, Mike, that you mentioned. It’s highly personal, and that’s a little bit… and that’s sort of why it’s an exciting space to be in, in B2B, because it’s… you’re dealing with people’s person… like, their own personal motivations and their psychologies.

Seth Marrs:
that… I’m gonna stop right there, because we just went full circle to, like, talking about up front around how we’re gonna automate the world at the beginning of this conversation, and now we kind of figured out that, yep, we… there’s tons of great things that we could do.
But the human side of this is something that we all… that we all crave, so I think that’s a good… that’s a good ending spot, unless somebody else has something else. We’ve got 2 minutes left, I think that’s a good place to stop. Thank you guys very, very much for joining, for such a great conversation. I learned a ton, and I hope everybody who’s listening learned a ton as well.

Michael Litt – Vidyard!:
Awesome.

Tooba Durraze:
Thank you.

Julia Nimchinski:
What an incredible session. Thank you so much, Seth, and everyone. What’s the best way to support you, Seth?

Seth Marrs:
Yeah, I mean, follow on LinkedIn and join in for more of the, like, conversations on the podcast that I’m the Innovative Revenue Leader.

Julia Nimchinski:
Awesome.

Seth Marrs:
Right.

Julia Nimchinski:
Thanks again!

Michael Litt – Vidyard!:
Humans make things interesting.

Seth Marrs:
Yeah, for sure.

Sam Liang, Otter.ai:
us.

Seth Marrs:
Thank you, guys.

Sam Liang, Otter.ai:
Aye.