Subscribe
Share
Share
Embed
Remember Salesforce and Seibel? Well. We're about to see a lot of similar things happen thanks to AI.
In today's episode, we discuss how AI is eating software - who's already being disrupted, and what a moat looks like now.
In today's episode, we discuss how AI is eating software - who's already being disrupted, and what a moat looks like now.
- (00:00) - Introduction
- (04:47) - AI's Impact on Software
- (08:34) - Understanding AI Agents
- (13:50) - AI's Role in Business Processes
- (18:16) - Challenges in AI Adoption
- (25:19) - The Rise of AI in Software Development
- (26:01) - The Importance of Smaller Models and Data
- (28:12) - The Role of Memory in AI
- (29:54) - Knowledge Graphs and Their Impact
- (31:08) - Customer Success and AI
- (33:12) - Navigating AI for Incumbents
- (38:27) - The Future of AI-Driven Companies
- (42:34) - Practical Tips for AI Integration
This episode is brought to you by by Fullcast, the only AI-powered platform that streamlines your entire sales lifecycle — from plan to pay. With modules like territory and quota management, routing, and capacity planning, Fullcast adapts to your unique needs — whether you need one solution or an all-in-one platform.
Ready to see the difference? Visit Fullcast.com and mention the Revenue Formula Podcast to unlock an exclusive premium gift, just for listeners!
Never miss a new episode, join our newsletter on revenueformula.substack.com
Creators and Guests
What is The Revenue Formula?
This podcast is about scaling tech startups.
Hosted by Toni Hohlbein & Raul Porojan, together they look at the full funnel.
With a combined 20 years of experience in B2B SaaS and 3 exits, they discuss growing pains, challenges and opportunities they’ve faced. Whether you're working in RevOps, sales, operations, finance or marketing - if you care about revenue, you'll care about this podcast.
If there’s one thing they hate, it’s talk. We know, it’s a bit of an oxymoron. But execution and focus is the key - that’s why each episode is designed to give 1-2 very concrete takeaways.
[00:00:00] Toni: Is software still eating the world or is software being eaten now?
[00:00:04] Jonathan Moss: pre Salesforce, right? Pre cloud, pre SaaS, Siebel and Act had 90 percent of the CRM market. Salesforce wasn't even a thing, right? Cloud SaaS happens, guess what?
[00:00:14] Now Salesforce has it. So I don't think there's any reason why this can't happen again. And I think AI will be the driver of it
[00:00:21] Toni: That's Jonathan Moss, EVP of growth and GTM at Xparity. He shares how AI is disrupting companies such as Chegg, how modes are disappearing, and much more.
[00:00:34] Jonathan Moss: there's this whole question of, could AI replace a CEO? And what it found was in the normal day to day stuff, AI could handle obviously those things at a higher volume, better quality, et cetera, because it knew all the data.
[00:00:45] But where it struggled was With collaboration, with coaching and development, or with unforeseen circumstances that businesses always go through. Market conditions changing, buyer behavior changing, because if you think about what AI, AI is trained on history. It's not trained on what's happening in the market today and what's changing.
[00:01:02] Toni: Before we jump into the show, today's episode is brought to you by Fullcast. The only AI powered platform that streamlines your entire sales cycle from plan to pay.
[00:01:14] Customers report up 80 percent cost savings, 20 percent growth in pipeline, and 30 percent boost in RevOps efficiency.
[00:01:22] With modules like Territory and Quota Management, Routing and Capacity Planning, Fullcast adapts to your unique needs whether you need one solution or an all in one platform
[00:01:34] Visit fullcast. com, book a demo, and mention the revenue formula podcast. To unlock an exclusive premium gift just for listeners
[00:01:44] And now enjoy the show.
[00:01:47] It's it's fun and now it's kind of, you know, I I keep referring to it now as our, it's, it's our little hobby. You know, this guy, this guy also goes hunting. Like that's a big thing here in Denmark, going hunting, killing things.
[00:02:00] I think this is what, what, what Danish and Americans have in common, I guess. But you know, and, and for those hobbies, you need to pay a lot, right? Kind of going hunting is pretty expensive actually. Yeah. And I'm starting to refer to this year now to my family. It's like, this is my hobby, but we'll get paid a little bit of pocket change for it, actually. Kind of, it's, it's like a, a profitable hobby.
[00:02:20] I wonder
[00:02:20] Mikkel: if that's like hunting, because there you go. Well, you know, I can sell the meat and then make some money. Do you ever sell the meat? That's the point. It's like, oh, we can sell the ad space, but we need to buy expensive cameras. So we ended up just doing, you know, spending the money on that. I think net, it's net
[00:02:36] zero.
[00:02:36] Jonathan Moss: So, so, net, net, net, it's the same. I heard, I heard I was listening to, was it the Jacco episode, and you mentioned you had upgraded your gear, so, so I heard about that, and, and to your point, yes, hunting is a thing And we, we love doing it. I will say I used to hunt when I was little or with my dad.
[00:02:52] I don't hunt anymore. I've turned on to fishing though. So fishing is my thing. And so we'll you know, we did that over the holiday break. We're going to go over Christmas and. We spend time out in the country of Tennessee, which is where I grew up. And there's just this big lake, a bunch of woods, but my my kids have decided that they wanted to try hunting.
[00:03:11] So now I have to get back into it and figure it out. So, so there you go, Mikkel.
[00:03:16] Toni: But, but is it like normal, normal kind of fishing rod? Because in the U. S. they do a lot of fly fishing as well, right? Kind of, is, is that what you're
[00:03:23] Jonathan Moss: They do. No, no, I'm not. I do have, I don't have the, I have the rod where you kind of hold the string and throw it, but it's not, it's not like a fly, so I'll
[00:03:30] throw it
[00:03:31] Mikkel: Spinner, spinner.
[00:03:32] Jonathan Moss: There you go. Yeah. So Spenner and I put you know, we put, we have a dock, so we'll put like five or six in there.
[00:03:37] In fact, I got a picture I'll show you all. My son caught a 30 pound catfish. He's only nine. He couldn't even reel it in. He was like, and he, and the crazy thing is only on a 12 pound line. So I don't even know how it didn't break, but he couldn't reel it in. He was like, I was, I was up. He was like, Dad, Dad, I got a big one.
[00:03:55] I was like, Oh no. So I ran down there and try to help him get it in. So.
[00:03:59] Mikkel: Well, at least it didn't get him.
[00:04:01] Jonathan Moss: Yeah, exactly,
[00:04:01] Mikkel: not by the sound of it. It could have, it could have happened. It could have happened.
[00:04:05] Jonathan Moss: for
[00:04:05] Mikkel: Oh man. I mean, we could talk
[00:04:07] about this for hours. I feel like there's going to be a bonus. There's going to be a bonus section. He said he didn't have a hard stop. No, exactly. Oh, you're going to, Toni's going to hate this episode. He's not going to be able to participate
[00:04:19] now. It's all, it's all going to be game. It's all going to be game.
[00:04:22] Jonathan Moss: We can talk about podcast gear. Look, I got a mic too. So, you know, I had to, I
[00:04:26] had to
[00:04:26] Mikkel: That's true. We did notice, we did notice. Yeah. And I mean, speaking about game, we're going to talk about a different game Artificial Intelligence. You've been posting quite a bit on this topic. I want to say, I don't know how long, actually. I, I stopped scrolling past some point, but it's definitely an interest I can tell of yours and super happy to have you on.
[00:04:47] One of the posts you recently did was about how AI is eating software and I thought that probably would be a good place to start. Can you tell us and the listeners a little bit more about what is, what is actually happening?
[00:05:00] Jonathan Moss: absolutely. So, so I think first we've got to understand a little bit of history. And so every, about every decade since the sixties, we went through some sort of technological shift. Right, so you think about mainframes to networking, networking to PCs, PC to internet, internet to mobile, and then the, the kind of transition to cloud and SaaS, which we're, we all talk about, we all lived in over the last you know, decade or so.
[00:05:25] Well, we're entering this new techno you know, technological platform shift of AI. And so, you think about kind of what AI at its core can do. So, right now if I think about the way software is built, There's a lot of non intuitive UX UIs, so many people have to learn how to even use software, and HRIS systems, ERPs, all these things are really the worst at this, but, and the way I like to frame it is, Software has been built where we actually have to work for it.
[00:05:55] It doesn't work for us. So we got to click, we have to select, we have to type, we have to do all these things and CRMs are notorious for it. And so I don't know of one person, and if you do, I would love, if there's anyone out there, cause I know a lot, everyone on that listens to you all has, you know, uses a CRM or has used a CRM.
[00:06:13] If you wake up every day and say, man, I cannot wait to log into the CRM and use it today, I want to hear from you because I want to understand exactly why you enjoy it. Cause I don't think there's one person out there that enjoys using a CRM, but the CR, the CRMs were built with this kind of purpose of, you know, a source of truth for customer data, streamlining processes, improving customer relationships and doing all these things.
[00:06:36] But I don't think that it ever fully accomplished a lot of the things that was meant to do just because the way it was built. And so with AI, every layer of software is evolving from the UX, UI being more personalized, being more dynamic to the back end operations being more automated, being more intelligent, actioning things.
[00:06:56] So, so if I take that same CRM as an example and I layer AI in of what it would look like in the future, so the CR, the CRM becomes just solely the system of record. Thank you. And AI becomes a system of intelligence, context, and action. So I have this way where actually AI is working, you know, is working the software for me.
[00:07:20] So the software is actually, actually starts working for me more so than I have to work for it. So I think that's number one. Number two is this is taking that same example for CRM. The CRM sits on a data of, or a goldmine of data, right? That we don't leverage. Mainly, probably 80 90 percent of it, because it takes time, it takes resources, it's unstructured.
[00:07:43] AI can now take all of that information and provide me insights, provide me recommendations, and actually make me smarter and more intelligent. And then I think one more step in the future when we have AI agents, AI agents will be able to do a lot of this autonomously. So I can set guardrails in place for what I want it to do without me, And what, what I want it to come to me with recommendations.
[00:08:07] And then I'll decide if I want it to do those things as well. And so I think we're kind of going through this, this shift and I'll stay on the CRM example. So pre Salesforce, right? Pre cloud, pre SaaS, Siebel and Act had 90 percent of the CRM market. Salesforce wasn't even a thing, right? Cloud SaaS happens, guess what?
[00:08:26] Now Salesforce has it. So I don't think there's any reason why this can't happen again. And I think AI will be the driver of it.
[00:08:34] Toni: Before we go deeper into this because there's a bunch of things to kind of hook into and drill into actually people are throwing this word agent around a lot.
[00:08:43] And I just wanted to, you know, I don't know if you have the official definition or if you're the, the, you know, commercial people understand this kind of definition version, but like, tell us a little bit,
[00:08:53] what is an agent for you?
[00:08:55] What does it mean? And how is it different from, you know, just a workflow or an automation that people might be used to already?
[00:09:00] Jonathan Moss: yeah for sure. So I think a great question and I don't know that there's an official definition and I probably won't be the one to say it but here's the way that I contextualize when the AI agent is. It is software but the soft but the software the agent has a specific objective. It's specialized in a specific, you know, task, knowledge, et cetera.
[00:09:22] And it's built to act autonomously and action things on my behalf. So the way I think about it is, and a good example that I use a lot, cause I think it's easy, cause everyone travels, right? If you think about today, we actually have a lot of agents and travel is one of them. So I can use Expedia as an agent to book travel.
[00:09:42] I can use you know, a travel agent. If you will, but I have these agents that can help me with travel. In the future, what an AI agent will do is it will know who I am and what my interests are in my family. It will go out, it will not only build an itinerary, but it will book my travel for me. It will set everything up for me.
[00:10:02] And that way I don't have to really do anything. So think about it as it's a software that has a specific objective and task that needs to complete, and it does it on behalf of me. That's the way that I look at what an AI agent would do.
[00:10:14] Toni: Kind of following up on the first part, right? So we were talking about, you know, how is it eating, you know, how's AI eating software basically?
[00:10:22] You mentioned the CRM as one really good example here. Would you say this is also who's going to get eaten first, kind of, or if you were to kind of be thinking about it, well, you know, out of this whole realm of possibilities what is, or whom is AI going to come after first?
[00:10:39] Jonathan Moss: Yeah. Very interesting question. And so I'll give an example of where it's already in software and then I'll, I'll kind of transition that into kind of the go to market perspective and what I think. So if you think about a company called Chegg, I don't know if people are familiar with them, but basically it's an online learning platform, homework, help, tutoring, things of that nature.
[00:11:00] It was significantly disrupted by CHAT GPT when CHAT GPT came out in November 2002. And so to put it in perspective in November, December, it was at a 29 stock price and a 3 billion market cap. Since then, if you look at it today, it's 2 and some change and 200 million. Market cap. So this kind of gives, you know, because obviously you had the ability to do almost everything that Chegg did through, you know, go through ChatGPT for 20 a month.
[00:11:27] So if you think about that, why was Chegg disrupted? It wasn't part of a workflow. It wasn't really embedded in my data, you know, and I found a better tool that I could, that was more efficient for me, cheaper for me, and could actually do more things than what CHECK did by itself. So I don't think CRMs will be the first ones to be disrupted.
[00:11:49] Though I do think that there's an opportunity in the future to disrupt them if, if they don't do a lot of change. Because I think they're too embedded in the workflows today. And they're the system of record, right? So if you think about having to rip and replace a system of record, it's, it's gonna be super difficult.
[00:12:04] So I don't think they're the first place to go. Where I think the first place from a go to market perspective would be, and I think the most vulnerable, is probably the sales engagement platforms and the enablement tools. The reason that I think that is because they're the layer that sits on top of the CRMs today, right?
[00:12:21] They're, you know, you can consider them like a system of engagement that's engaging. And I think that, you know, number one, there hasn't been really great adoption of them overall. I think mainly people have used them for like automated emails, which we all know we spam cannons, if you will, that are awful today.
[00:12:38] And AI is going to make them even worse if you put it in the hands of people that aren't trained. So, I also think they've become very expensive for what they're worth because what's happened is that you've had the market consolidation, right? Like, used to Gong did conversational intelligence, Outreach did sequencing you know, Clary did kind of the deal forecasting and stuff.
[00:12:56] And since all of them, have kind of converged on each other and started, you know, they were partners to begin with, and now they kind of do all the same things because they've become a horizontal platform, which has made them more expensive. But what's also from the buyer side, that means I'm not using the full capabilities, but I'm paying for them more than likely.
[00:13:16] And so I think they have a way to easily be ripped and replaced without a lot of impact, if you will. And I think there's a lot of tools you know, out there. Well, I wouldn't say tools. That AI, depending upon how you look at it I would say you could use AI to replace a lot of the things that they're doing from, you know, the, the, the sequencing you can, from the workflow perspectives, think about tools like Zapier or Make or Copy.
[00:13:40] ai. They can build these kinds of workflows for me using AI. And so I, I do think they're probably the first and most vulnerable at this point.
[00:13:48] Mikkel: yeah No, it's interesting.
[00:13:50] I think the other I mean obviously we took it from the vantage point of eating software But there's also the question that I'm wondering about now is well Is it just software because we also know that there's a bunch of roles. Is it gonna eat people? No, exactly. How
[00:14:02] hungry is this thing? You know what I mean? and I just wonder like And I think the trigger was you said hey it works for me To potentially replace me at some point. Like what functions, are there anything you're noticing there as well? I just think it's interesting to kind of bake into this conversation too.
[00:14:20] Jonathan Moss: It's super interesting. I, so I, you know, I hear a lot about this, right? I think actually Jensen Huang of NVIDIA had probably the best quote and I'm paraphrasing it because but ultimately he said, instead of looking at AI replacing 100 percent or 50 percent of roles, you should look at it at, as AI replacing 50 percent of the tasks or things that 100 percent of roles do today, right?
[00:14:44] So what, what is AI good at? And there's also another study I'll you know, I can, I can share, but ultimately AI, there's like a MIT and Harvard study where AI was put up against CEOs and said, could, you know, there's this whole question of, could AI replace a CEO? And what it found was in the normal day to day stuff, where if it's like super small decisions, things of that nature, AI could handle obviously those things at a higher volume, better quality, et cetera, because it knew all the data.
[00:15:12] But where it struggled was actually With collaboration, with coaching and development, or with unforeseen circumstances that businesses always go to market or go through. Market conditions changing, buyer behavior changing, because if you think about what AI, AI is trained on history. It's not trained on what's happening in the market today and what's changing.
[00:15:32] And so I don't, I do think that there's probably some roles. You know, I know AIs, SDRs are one you think about a lot. I hear a lot about them. I think there's a lot of challenges there. So I do think there's some roles that could, you know, be, that will be impacted, no doubt. But every technological platform shift has had that, right?
[00:15:52] They've, Certain roles have went away, but new roles were created. And I think AI will be doing the same thing. So I think it's, it's wrong to say it's going to replace a bunch of people. I do think that it's going to replace roles and roles will look different, but I also think it's going to allow us. To get out of the things that we're doing today that are not valuable.
[00:16:11] So administrative stuff back to back meetings you know, putting documents together. I think it's going to allow us to really get back to the things that we're great at, creativity, strategy, collaboration, learning, trying to predict what's going on, I think there's a lot of, a lot to say that, that AI will help us get back to that.
[00:16:30] So, so that's the way my vantage point, right, wrong, who knows, but I think that's, that's really kind of the way that I look at it today.
[00:16:37] Toni: You just mentioned kind of something really interesting, kind of AI is history, right? Because it basically is a very great summarization tool of, of everything that's passed, right?
[00:16:45] All the stuff that has been done already. Do you think that's a structural problem for, for the technology in general?
[00:16:51] Actually, and maybe this is more of a technology question, maybe kind of that's not your expertise area, but do you think that that limitation will kind of stay in place and create this cozy divide for, hey, AI, you do all the boring stuff that we just, you know, we can't remember everything, and we do the interesting stuff?
[00:17:07] Because that's kind of a very kind of happy version of, of
[00:17:10] trying to tell the story.
[00:17:11] Jonathan Moss: That's right. Yeah. No. Yeah. So from a technological standpoint, you're right. That's, that's outside of my domain expertise. But here's what I can tell you that I do know in the way that I would look at it is It can scenario plan for sure. So, but the thing is, is you've got to be able to give it what scenarios you think are out there to test, right?
[00:17:28] And it will be able to do that better, more, more efficient than us. So I do think there are scenarios to where you know, no, no pun, I guess using scenario word, but ultimately there are things that where it will be able to do that, but it, it's going to have to be programmed by someone or you're going to have to give it the, I think domain expertise is the best way to put it.
[00:17:47] Yeah. If you have domain expertise, you will, you will rise above anything else. If you're kind of a, a generalist across many things I think that you, you, there's a, there's a chance that your job gets disrupted. If you have domain expertise, you're the ones that can use AI to its full potential because you can kind of, you know, predict where the ball is going or where the puck is going, as Wayne Gretzky says.
[00:18:08] So I do think there's scenarios where that happens.
[00:18:11] Toni: And for everyone who's worried and scared of this thing.
[00:18:14] There's, there's one upside.
[00:18:16] Which is, at least what I'm seeing is Outside of the LinkedIn echo chamber where everyone is doing everything on AI already, if you really look outside in the real world, kind of no one is. Right?
[00:18:26] Kind of the real AI adoption is far lagging behind the hype. We saw some crazy, you know, terrible numbers from Microsoft's co pilot adoption. I don't know if there's something kind of new coming out, et cetera, et cetera, et cetera. Right. What is, what is your, what is your take actually on this lagging AI adoption?
[00:18:46] Is it just us hyping it up or is it, you know, what's, what's, what's the reason?
[00:18:50] Jonathan Moss: Well, you know, everything goes through the hype cycle, right? And I, I do think that AI is in that. I don't know what part of the Gardner hype curve it's on, but, or part of that it's on, but it's on, it's on it for sure.
[00:19:01] And, and you're exactly right. You know, I think there's a couple things here on the approach.
[00:19:05] So one is what I see and, and kind of read into, because I see the same reports that you're seeing, and I think it's, it's very interesting. But I think a lot of people think of AI as like another tool. And you know how good today we are at implementing another tool and having people adopt tools. Or they look at it as a magical wand where like, Hey, I can just implement this new thing and it will just solve all my problems.
[00:19:30] And so it's, it's less about like, what should AI handle? And it's really more about what outcomes should AI own. And so we're really thinking about it the wrong way. The other thing is, is just because, We have AI doesn't mean that we can skip on the fundamentals because it's still garbage in garbage out, if you will, just like a CRM.
[00:19:50] So, you have to have the foundational data, the processes, understanding what you want to accomplish, and we're very inefficient at that today. So while AI can make my, my inefficient processes better because it can do better volume, quicker, cheaper, all those things, I still have to have a process for it to understand and what to do.
[00:20:09] And so. So let's use like, for example, the personalized outreach example, since we were on that a little bit earlier, sales engagement tools. I can't just, and this is why I think AISDRs are struggling and will continue to struggle is. If I'm going to reach out to someone as a business, I still have to have key fundamentals.
[00:20:28] I have to know who my ICP is and who I'm reaching out to. I have to understand my segmentation, to understand the persona. I have to understand their pain points. I have to understand what my value proposition is, what channel they're in, who's the buying. I have to still understand all of these things in order for me to deliver a personalized message that actually hits and makes sense.
[00:20:48] What I see is like a lot of companies don't even have that. Don't even have 50 percent of that. And so you can't just unleash, you know, AI without that, that basic, those basic fundamentals and expect that you would actually get a good return from it. And the cool thing is though, is AI can help you figure a lot of those things out, but you still have to figure the, those basics out.
[00:21:07] And so, so let's, let's, let's use it for, let's think about a couple of examples. So, predictive scoring in the CRM. So lead scoring and things like that. Right. So it's been around for a while. It's valuable in a sense, but it doesn't really, you know, move the needle. So if I were to say I'm going to implement AI to do opportunity and lead scoring, great, but what does it actually do?
[00:21:29] And I have to rethink the customer life cycle and the journey and what AI can be the most impactful at. Or another version of this could be on the support side. So if I want to apply AI to speed up existing processes, so think of like, faster ticket handling customer service. That's great. But is it really moving my needle?
[00:21:48] No, I need to rethink. How I can eliminate tickets through intelligent self service and things of that nature. So we have, so the problem is, is that we're thinking about it wrong because we're thinking about it as just a tool that I implement or a magic wand, silver bullet that's going to save, you know, a lot of costs or stuff like that.
[00:22:03] And so I think about that is, that's kind of one bucket. The other bucket, and this I think was BCG, I believe, I hope, hopefully I gave the right person credit, but I saw this in one of the research reports I gave was that out of all the companies that they've implemented AI in. 70 percent of the problem is people.
[00:22:22] It's not the technology. And so change management, enablement, training, all the things. So I do think that there is, while we're in this hype curve and while there's a lot of you know, there's a lot of capabilities of it, it can do a lot of things. It's still going to be a long, long drawn out cycle of implementing it, the people changes, the process, the data.
[00:22:43] There's still a lot that has to go on in order for it to be effective.
[00:22:47] Mikkel: So do you think we've found those valuable or built rather those valuable uses yet? Because, so you mentioned, for example, AI closing tickets and I can see the scenario where a manager goes like, Oh, great. It can close, you know, a thousand tickets for us. Then we don't have to bother with those canned replies, but we're going to keep the staff. So we're just going to add, you know, cost on top. Do, do you think we've come to the point yet where we've built out all those valuable uses yet, or like, how do you see that side?
[00:23:15] Jonathan Moss: Yeah, no, I don't think we have. I do think to that point, that example I think is a good one. Cause it's then like, what else can I have that person do to drive impact? In the business. Is it you know, becoming more of a customer success side? Has it become more technical? What is it that I could potentially use?
[00:23:30] Because they have a lot of knowledge. And so, I think about it of, of how can I, you know, augment or replace the labor or the job to be done, if you will, with a different job to be done that would be more impactful. I don't think we figured out all the use cases. You know, the technology is changing so rapidly.
[00:23:47] I mean Just look at the 01 reasoning model that came out. Like, what can we use? I mean, I don't even know all the ways that you can use it, but there's probably a ton of new ways that we can use it that we'll figure out over time. So, I do think that there are some low hanging fruit today. I do think that you've got to be ready for what's to come.
[00:24:05] So, thinking about your data, your processes, and all those things. So, I do think there's a lot of things that we still have to do to get ready for it, but I don't know that we've, we've figured out all the, you know, all the different use cases. I mean, I can Give you a ton that I use, but ultimately that's probably one tenth of what we'll, you know, I'll be doing five years from now.
[00:24:23] Right. So,
[00:24:24] Toni: I think the, the, the point around kind of the model maturity kind of, you mentioned O1, which is in, in AI world now, pretty old news already,
[00:24:32] right? Kind of there's plenty of new stuff actually has come out since. I also kind of firmly believe the, the sophistication of the model is not what's holding use case
[00:24:41] and adoption back.
[00:24:42] Kind of, kind of the frontier models are like far, far out there. And actually kind of when you look under the hood of most of the AI tools out there, they're using those frontier models only for a very thin slice of the, the work that needs to be done. And then using much smaller, simpler you know, LLMs in order
[00:24:58] to get, you know, most of the heavy lifting done. Economically, but also, you know, it's faster and just, you know, not necessary kind of to use the, the, the, the, those massive models kind of in the background. And one thing that always comes up when we're talking about AI and, and, and software and building products specifically is like. Well, what's, what's the mode going to be, right?
[00:25:19] Because there's this, there's this specter out there like, hey, you know, we have cursor, we have a couple of, you know, I mean, Git, GitHub Copilot and pool sign and what have you, kind of all of those kind of teams that are basically helping developers to develop faster. And then there's even some teams that, you know, Write code even without this being an augmented reality, you know, approach for a developer, right? So there's this, this saying out there now, like, hey, you know, it's super easy to copy all of those software tools, right?
[00:25:44] And let's just be that as it may, maybe we can talk about this in a minute, but you know, if this is true, actually, what do you think the true mode actually becomes, you know, in this, in this world of it's really easy to build
[00:25:57] Jonathan Moss: Yeah, it's it's super interesting. So two things you said, and then I'll, I'll answer the question.
[00:26:01] So one is. These smaller models. I do think that you know, and, and, and kind of going into one of the moats is around the data. And so I do think these smaller models do prevail. I think you use the larger frontier models for certain things, but developing your own kind of smaller model, not only for cost, but also what you see is you hear a lot about hallucinations or wrong answers.
[00:26:23] I mean, you know, I'll, I'll even admit there's, you know, you got to double check everything. I do a lot of data analysis using using ChowGBT. And even some basic things like I asked it you know, to tell me who my best performing AE was in this scenario. And I get very specific with ACV. So who's the best in ACV?
[00:26:42] It actually gave me the number two, not number one. So like even things like that, it makes mistakes on now it'll get better all the time, but ultimately I think because it's so large, like it can make these small mistakes. If I make it smaller and more specialized to what, you know, it is that I need to accomplish it becomes, it becomes faster, it becomes more accurate and it becomes trained on the data, which I think is the number one mode is the data that I want it to know and want it to prioritize versus it doesn't.
[00:27:09] Prioritizing a vast array of different things. So I think, I think data is very important. And so I think this is where if you will, the Salesforce and others have an advantage, but ultimately that is going to be, that is going to be key. I think the second thing is workflow. So how is it embedded in my workflow and how is it helping, you know, again, like I said, kind of working for me versus me working for it.
[00:27:31] So thinking about the workflow, someone goes, so who is your user? What do they do on a day in and day out basis? And how can I build a workflow that makes everything frictionless and seamless for them? Or let me flip it. Who's my buyer? And how can I make their journey or their buying experience less friction by using, by thinking about the process?
[00:27:53] So I think the workflow, a very embedded workflow, and this also is where I think You know, a Salesforce could you know, if you will be, this is why you won't rip and replace them, but I don't know that they've got it figured out to where they can, you know, do things in my entire workflow. And that's where I think that you know, the AI emote could happen.
[00:28:12] I think another one is memory. So this is one that we don't talk a lot about you know, unless you're kind of in the AI world. So, Think about you know, when we talk about agents or, or AI and stuff, right? I talked a lot about the data, the workflow, et cetera. Memory is super critical and will be even more critical.
[00:28:28] And what memory really is, is as AI gets to know me personally or my business personally, what I'm doing, it learns, it adapts, it, you know, again, personalizes. Right. And so what happens is. It can start recalling my preferences, it has context, it has historical interactions. So, when you embed AI with memory, it will become harder to rip out unless the company themselves has built the memory.
[00:28:55] Because what happens is, is if I, if I, you know, if an incumbent comes in and tries to replace that moat, All that memory is gone, right? So now I have to, now AI has to re learn all of that over a period of time. So I think memory is probably one of the bigger moats and that's going to be, it's not that it's technical, you know, someone can't build it.
[00:29:14] It's that it's that first kind of speed to value, time to value, getting embedded, etc. I think that's, that's a, that's a big one. I think user experiences now become forefront of modes. This is why I also think that incumbents like, you know, Salesforce, I know I'm doing this CRM thing, but I'm kind of trying to build the story, if you will, on it, but ultimately why.
[00:29:34] Salesforce will will struggle to become the user experience that I interact with because the user experience is awful. You know, and so I think user experiences become a new moat. And so those that have really, really strong interfaces that are dynamic, seamless, intuitive, personalized, all those things.
[00:29:53] We'll win.
[00:29:54] And then I'll probably, and I'll say one more, which is also not talked a lot about, about, is you can build these things called knowledge graphs. And so, I'll give you two examples and then I'll give you, go, go to market examples. So, knowledge graph one is Google uses a knowledge graph for their search algorithm.
[00:30:09] So they're connecting all these different pieces, this vast amount of data, of unstructured and structured data, and they're building relationships with it. Facebook now meta. Used a social graph, which was a knowledge graph of all these social experiences, which, which you know, allowed it to be successful.
[00:30:25] I just saw, I've seen a couple of companies. So one in, you know, healthcare, which is what I'm in. And I've seen that, but ultimately from the GTM perspective, I saw one the other day, they had built a knowledge graph around network relationships. So their knowledge graph connect everyone that's in a company from board to customer, you know, customer, internal employees, et cetera.
[00:30:45] And LinkedIn or other places to try to find what are the relationships between everyone. They've already built a knowledge graph that can do that. And so think about that from a, you know, warm intro, referral, et cetera, perspective, it's, it's all served up to you. And I think that those types of things become moats in the new world of kind of this AI.
[00:31:06] So I would say those are probably the five.
[00:31:08] Toni: One, one follow up here, and this is kind of a secret hope of a bunch of people. Maybe it's not so secret, but you know, customer experience, customer success will be much more important and will maybe become a mode. Sometimes when I hear this, it's a little bit of the Hey, hopefully CS isn't going to stay the forgotten child of the go to market.
[00:31:28] Like finally, there's, there's a change for that. I just, sometimes I myself think it's a little bit thin. And, and I think kind of you didn't, you didn't add it to your, to your list. There was very tech stuff which I think
[00:31:38] is super interesting. But just double checking. What's your, what's your perspective on is CS suddenly going to soar to the front and, you know, be the moat for, for the go to market?
[00:31:49] Jonathan Moss: interesting. I don't think so. In fact, I think the world of CS, just like CELs and just like SDRs, I think looks, I think all of go to market looks completely different. I don't think CS becomes a moat because you know, and it's, it's similar to why, you know, an AE or a SDR is not a moat because there's just the knowledge has flipped, like buyers no more, the buying process is terrible.
[00:32:14] I mean, even on this customer success side, you know, we, I mean, who, who, who wants QBRs? I mean, who, especially with information that either A, I have myself or B is not relevant to me. You know, so I think that CS in itself is not a moat. I do think CS changes in the way that it's done today and potentially becomes more technical in nature or you know, more kind of, enabling and architecting, orchestrating, if you will, how AI is impacting customers.
[00:32:46] And so this is why I also think probably a whole nother topic for another podcast that I think RevOps for the first time becomes Elevated. And actually that team grows where other teams potentially you know, go down because ultimately they can orchestrate a lot of this through the technology, the processes, the data, things of that nature.
[00:33:05] And so, I do see in, in a world where, you know, customer success completely changes and I do not think it is a mode
[00:33:12] Mikkel: What do you think I mean, we've talked a lot about the the technology at this point, but there's a bunch of incumbents on the other side who might no longer have a mode, you know, you, you mentioned Czech, massive company getting disrupted, like really, really fast here. There, you know, I, and I recently, recently even spoke with a company that didn't have an AI strategy yet because they weren't sure how to go about
[00:33:34] it.
[00:33:34] How should, how should they navigate this environment right now? Because. They could choose to just sprinkle in some AI and then put the text on their website, potentially rate and extend raise an extension route. What do you think the successful companies do here to adapt?
[00:33:48] Jonathan Moss: Successful companies that have, or building AI products or successful companies that are used, that need to use AI. incumbents. Okay. Got it. Got it. Understood. Okay. Yeah. So I think the, the biggest thing here is, is a couple. So one is, If I'm a, if I'm an AI company and I'm building you know, software today the incumbent side of things is they're thinking about AI as an add on or an augmentation.
[00:34:09] It's not the foundation, right? And I mean, I mean, Salesforce is a great example, HubSpot, others, right? They're saying like, Hey, I have this, to your point, AI that that can do things, but ultimately it's just a bolt on. That bolt on because it's not built into the software itself, right, is going to have.
[00:34:25] process challenges and things, right? Because it can only do a specific function like AI customer assistant, right? It can only do those types of things. And so it's just bolted onto a. Inefficient legacy process. I think number one, so their infrastructure is not set up to fuel AI. I think that that's number one.
[00:34:42] Number two is these incumbents you know, the, the companies that exist today they can't risk cannibalizing their existing product or platform. And we see this a lot. So the innovators dilemma, right? You've built this software you know, you have kind of two options. You can disrupt yourself and really set yourself up, which is very expensive and your shareholders.
[00:35:02] Your you know, your investors, all of those people are not going to be excited about that because revenue is going to take a hit, et cetera. And so, I mean, think about it, like would Sears or Kmart of the world been great by just slapping them online with a website? No, Amazon rethought the entire process of buying right from the customer first of what e commerce would be.
[00:35:24] And so I do think that you know, that these AI first companies are thinking about it more so from a. First principle standpoint of how is, how do I have the AI infrastructure? I think that's number one. Number two is, I think AI, these kind of incumbents can't get you from, you know, a complete, if I think about a process and I think about what is the problem I'm trying to solve from end to end.
[00:35:47] It can only get me part of the way. So like, I can maybe think about a marketing campaign and it can give me that part, but then what about executing that campaign? Well, I've got to do something else. Well, then what about how does that transition to you know, lead qualification and my BDR, SDR, and my AE?
[00:36:02] Where an AI first company that's building this can think about the entire workflow or the entire process from end to end and how can I use AI to make this kind of these transitions, this data, this process be more intuitive, where if I'm an incumbent, My process and legacy system systems are already built.
[00:36:19] So I'm, I'm pretty much out of the game there. Unless I, again, disrupt myself. I think the other side of this is you know, there's things like multimodal, like, text, images, video, audio. I think audio becomes a big factor where I can just talk to software and it does something for me. You know, these kind of incumbents are gonna try to do this, but again, I think it's gonna be.
[00:36:41] Very dispersed about what they're going to do. So I think there's a lot of different things like that, that AI can do. We could get into iteration and remixing but like, well, let me give you an example of remixing. So remixing is kind of interesting. So I don't know if you've used gamma slide deck generator at all.
[00:36:56] But what it can do is, let's say I decided I wanted to build a deck. And I got to the end and I said, man, I really think this could be like a memo or an article or something. I can just ask it to turn all of that information into that and it will do that for me, right? A legacy incumbent can't do that.
[00:37:12] They can't, they can't remix the entire thing once I've built it. It's almost like I build it and then now I have to like redo it all. So I think there's some big advantages of those incumbents. You know, and they'll move quicker, they're more agile. They'll, they'll, they'll try different strategies. And so I think that incumbents have a, a long way to go to try to figure this out.
[00:37:31] You know, like I said, there's, there's plenty, you know, we've talked about a lot of history. There's history shows us all of these transformational shifts and what has happened. So,
[00:37:39] Mikkel: It's also just funny. I, I constantly get surprised. I saw a post from a uh, a guy he was basically setting up Google Ads and showcasing their new AI, which could could see where his mouse was hovering and tell him, okay, now you want to go here and here and here.
[00:37:55] And it was like speech. They were talking with each other, guiding him through setting up something in Google Ads. And I was just like, Why doesn't it
[00:38:03] Toni: just do it?
[00:38:03] Mikkel: Yeah, no, it's like, okay, so are we gonna, do we need intercom now anymore? Because I can just chat with an agent that tells me where to click in the software,
[00:38:11] and it's like, we're still, we're still so early, so it kind of makes me wonder, like, this good old thing that invented the the Ford GT, Skunkworks, if that's gonna be a thing we start seeing in this industry, because incumbents, they're just, they're so path dependent.
[00:38:27] Toni: What do you think?
[00:38:27] So kind of in connection to this year, what do you think about the, the, these ideas swirling around you know, wonderful clickbait on, on, on social hey, we're going to see a billion dollar one person. Like do you, do you think we're going to actually get there? I mean, anytime soon or what's your, what's your thinking
[00:38:44] Jonathan Moss: no, I don't think we get there anytime soon. I mean, look, maybe. We don't know what we don't know, right? In fact, you know, if you listen to a lot of the, the, even the experts, if you will, the Altman's and Dario's and, and, and those that are actually building these models, you know, there's not even really a definition, a full definition of what AGI, you know, artificial intelligence is like, what is it?
[00:39:07] When have we arrived? We don't even know. So for me, for any of us to, I think, sit and say that there'll be these one person billion dollar companies, I think is a little early to even know. Is it possible? Maybe. But, you know, I don't think we're going to know for a while. I do think, however, there are, you know, I think a few things that AI will help.
[00:39:26] One is you 100 percent don't need as many people as you did previously. That, that, that's for sure. So I do think you're like ARR per FTE will go up. You'll be able to do things quicker from a coding and shipping perspective. I mean, For a guy who doesn't code, you know, I can use Claude or other things to, to basically let me give you examples.
[00:39:44] So, I, I had to prove a point. So sorry if sorry if anyone that that works with me is listening to this, but so I had to prove a point in one of their meetings, we were going down a path of, Building like an app building legacy tech that, you know, our customers kind of wanted today. And, and I kind of had to flip it and say, wait, let's not have them drive what we're building, but let's understand what is the objective that they're trying to accomplish and how can I accomplish that?
[00:40:11] So we got to that point, and literally within three minutes, I built in Cloud a UX UI that actually solved what they needed and was able to showcase it to them like in meetings. And everyone was like, wait, what? And I was like, yeah, like if, if I, if someone who doesn't, who just understands the problem and kind of, and how to use AI, has no coding ability, can ship a UX UI that actually works in three minutes, I mean, you know, this is, this is the power that it has.
[00:40:41] So, but at the same time, I still need people that understand and have, you know, certain domain expertise. Like I can't, I don't think there's a world where a CEO can have an A, just AI running all of their go to market for them, unless they understand how to operate, going back to some of the points we made, you have to understand the fundamentals of go to market.
[00:41:01] You have to understand the customer and buyer journeys. You have to understand what goes into that. Could you have. A couple of people that understand that very deeply versus 20 people do that. Sure. If you're, if you're set up correctly, things, I absolutely agree. Could you have less product and engineering teams?
[00:41:19] Absolutely. Could you have less HR and IT folks or other? Absolutely. I do think there are less people that are needed. And I also think there's less capital that you need, which is actually very good for us because right now, as you all, you know, well, no, well, I guess unless you're in the AI hype you know, you're not getting a lot of the capital and there's not a lot of capital that that's going around.
[00:41:40] And hopefully that'll change. Cause there's a lot sitting on the sidelines, but ultimately. I don't think there's a world anytime soon where there's a one billion dollar company, one person shop. I do think there's a place to where instead of maybe needing, you know, a thousand people to get to a one billion, I'm making up a number, but a thousand people to get to one billion, you may only need a hundred or two hundred.
[00:41:59] And so I do think that world exists for sure.
[00:42:02] Toni: So we were actually just the other day recording,
[00:42:04] I think it aired today, maybe the episode around 1 million A per feet, 1 million a r per FTE, basically going in exactly that direction, right? Kind of talking about the exact kind of same thing also. You know, if, if you don't need so many people, well, you don't need so many support functions.
[00:42:19] You don't
[00:42:20] need all of those useless VPs and C levels like us, right? Kind of, you don't, you suddenly don't need those folks anymore. So there, there's a whole shift going on, but you know, I think, I think this happens this conversation usually happens on a very conceptual level. Do you have
[00:42:34] some tips for some folks to, you know, that are thinking about, How can we mold our organization to go in this direction, right?
[00:42:41] Not to jump there. It's not going to be kind of a jerk on the steering wheel kind of situation. But what do you think is, what, what, what do you need to, you know, put in place in order to move it in our action where you can basically utilize AI more in order to automate things the way that your staff is doing right now.
[00:42:59] And maybe yes, you know, letting some of that stuff go, at least not hiring back.
[00:43:03] Jonathan Moss: yeah. So I think one, one, one statistics, sorry, I just pulled it up and get to that. So going back to your 1 million FTE. So, HG insights just released an, a thing called the next generation of sales AI 2025. It's a pretty interesting article, but it looked at 75. Companies that are, you know, in sales, AI, I guess, companies, what's interesting is 60 percent of them were 2 to 10 employees.
[00:43:29] And 76 percent of them were right at or less than 10 million. So if you think about that 60 percent 2 to 10 and 76 percent less than 10, you're right at where you could, you could definitely have, you know, spot on if you all just release that definitely a 1 million FTE company for sure, I think is definitely possible.
[00:43:48] So, so just a little insight there. So back to your question around those that are looking to do something, how, where do they start and how could they accomplish it? What I would say is. And sorry for sorry for all those those great tools out there. I don't think you need to start with a tool.
[00:44:05] I can do, and what I have found is I can do 80, 90 percent of what I need by going directly to the LLM itself. And so I think where I would start is if I understand my business and what are the most important things that I need to accomplish that are highly administrative, that are repetitive.
[00:44:25] that are manual in nature and that don't need very strategic thinking and things of that nature is I would look and say, okay, let's identify the top five or 10 of those. If I think about my process, but first, as a reminder, you got to have your data, your process, and all those things figured out. So let's assume that you understand all of those.
[00:44:46] The next level is, Where are the biggest friction points that are manual that that are, are repetitive that I can use AI to help you know, be more intelligent, more efficient, et cetera. And then what I would do is I would utilize, I mean, AWS has Bedrock Google has tools, et cetera, that actually you can go in and they're really no code type tools that you can build kind of AI agents, connecting to LLMs and things of that nature, or just use the LLM itself.
[00:45:16] And then start finding ways to you know, rapidly iterate on how can you, you know, build a workflow or, or have this, you know, have the manual task be actually, you know, done by AI. And I think what you'll find is it's not as hard as you think. And so what I usually say is just get started with that.
[00:45:34] I don't think you start with a tool. I think you can layer in tools once you get to the point where it's like, hey, I, I don't know how to solve this or I don't have a good way to solve it, and it's a little bit more complex. Then I can start looking for the right tools to solve that problem versus thinking that I'd just go buy an AI tool and implement it and it's going to solve, you know, what I need.
[00:45:55] So, that's kind of where I would say start and I think, like I said, I think you'll find, you'll find that it's a lot easier than you really think that it, it is.
[00:46:03] Mikkel: No, I think that's true. We were talking the other day about mid journey. So basically image creation that I've, I started playing with for the first time really. And I'm, I, I used Photoshop way back, right? Way more complex compared, but it's still a skill. You, you
[00:46:16] still, you know, you still, it's still a skill you need to develop in order to get the right output. But man, it's just. So much easier, actually, compared to learning Photoshop from the ground up. So,
[00:46:28] So I think that that's definitely a good tip to go actually directly to the LLM first. And also just, I think it, it starts expanding the horizon of ideas of what you can do with this technology. Right. So,
[00:46:40] Jonathan Moss: that's, I, I'll
[00:46:41] Mikkel: I, think
[00:46:41] Jonathan Moss: the, the, most, sorry. The most powerful thing I can tell you is if you can build workflows or learn how to build workflows in like a Zapier or a make, or a, you know, a copy.ai, if you will, and you can embed and you can use I mean, or, or use directly to the lms. Like, I think to me, those are the two things that could make the most impact and the most success.
[00:47:04] I mean, even if it's just. You know, one example could be as a lead comes in, like I'm gonna give you this example. Lead comes in I can pull in that information from, you know, from my CRM right into Zapier. It can go do an analysis, or copy that AI, either one. It can do an analysis for me. It can understand who the person is.
[00:47:22] It can, it can scrape. It can learn. Then it can, you know, use OpenAI Chat, GPT Cloud or whatever to build out the conversation notes or the meeting prep. And ultimately tell me a little bit about the persona that I'm going to go into a discovery call or a call with, right? So even that in itself is like time saving and is not hard to build and is super impactful because it's low hanging.
[00:47:43] So don't overthink it. Think about the, I think that's a great one that's repetitive, right? Because you get leads in a lot or hopefully you do. It's, it's something that takes a lot of time and is manual for someone to actually do it correctly with researching. So think, think about that as an example and where all are the processes or places across, if you're in go to market buyer customer journey that I can use something like that to help.
[00:48:08] Toni: Jonathan, I think this was also a great way to maybe kind of sunset the show here. And Sunset
[00:48:13] Mikkel: the show? What are we,
[00:48:14] Toni: are
[00:48:14] Mikkel: we
[00:48:14] Toni: ending the show.
[00:48:15] The episode. English, English is the second language, you know? Yeah, we, we, we actually are going to be replaced by, wasn't it Gemini? I think kind of had this, had this podcast thing going on there.
[00:48:28] Jonathan Moss: They do.
[00:48:29] Toni: awesome. Yeah.
[00:48:29] Jonathan Moss: LLM. Yeah. What's interesting about that too. I don't know if you saw, but they actually just released, I think it was, is it this week or last week, man? Stuff's changed so much. I think it was last week and I don't know if you saw this. So, you know, it, you know, it can generate an audio podcast, right?
[00:48:44] Now they just released the feature where you can interact with the hosts. So now I can interact with the AI hosts and ask a random question or things of that nature. So it's kind of, kind of interesting. Right. But but yeah, it's, it's definitely changing for sure.
[00:48:59] Toni: The maybe I can start this over. Jonathan, great,
[00:49:03] great way to end the episode. Oh boy.