Turning AI Tokens Into ROI (That You Can Measure) | with CEO @ Paid, Manny Medina

0:00The world will be fully agentic. So there is no software that will not be agentic in five years. And you just have to figure out how to make money from it. That's Manny Medina, CEO of Paid, a platform which helps companies with AI agents get paid for the work those agents do. He explains why every software company is racing to build their own agents before someone else's agents plug into their platform, do the work, and take the money. And ultimately, why it's the beginning of the end for seat-based pricing models. So like, what the do you get for the hundred seats you bought? Nobody knew. Like you want value.

0:30So all of a sudden we're all smart about value and want to ask those questions on agents. We also get into why your cost of compute is going up even as the AI labs make tokens cheaper. We debate whether an agent that gets 3 to 5 % of tasks wrong can be trusted with your team's work. And we uncover the gross margin Manny thinks software will run on once money stops being cheap. There is a difference between positive gross margins and 80%. Like I feel like a normal software business will be running on. If you want to know where software is headed once agents can do serious work, this episode is for you.

1:02Welcome to Topline. Hey everybody, it's Sam Jacobs. Welcome to another episode of In-Person Topline. I'm joined by my co-hosts, Afid Zaman, the CEO of Self Talent Agency, and A.J. Bruno, the CEO of Quotapath, leaders in incentive compensation design, management, and performance. deployment we're joined today by a good friend the person that originally told me to focus on pavilion full-time the co-founder and ceo formerly of outreach and now the co-founder and ceo of paid manny medina thank you thank you for having to top line we're excited to have you awesome i'm a long-time listener first-hand caller well we're glad to have you here so the first thing you want to know is tell us about paid what are you building what are you working on what is the problem that you're solving.

1:44Yeah. So thank you. So paid is a monetization platform for AI agents. And what we did is that we took a problem that I saw at Outreach, which is how do you monetize agents where agents are performing the work? So when something does work and doesn't need a seat to the work, you need to get paid for it in a different way. So it's not just seats because the work is not seats. The work is the work. The work is the effort. The work is the output. The work as the outcome and there was nothing in the um in the software stack that will help you sort of like a meter the work being done b make it explainable to your customer what kind of work do they get for it and then c finally price it and um so we built a platform thinking from agents back as opposed to like looking at sas forward and that would allow us to future prove your business and grow faster because as your shipping agents you need to get paid for what they do tell us how it's going it's going amazing so we took the last year and a half building um it's been a bit of a moving target because the agent architecture has changed you know when we came out to the market it was a sort of very deterministic workflows you had n8n and land graph and all this different like frameworks to like build your agents and that went away in december last year with the releases of opus 4 and then saspocalypse happened that was a big boom for us.

3:03And then fable happened. And then, you know, agents became higher tenacity, more autonomy to perform work. So the leaps of agent performance have been every three months. And we have to like pause and start chasing, start chasing the tech. We should work back from where we think the tech is going. So we now assume that agents will have eventually, you know, full autonomy and high attribution. And now we need to make sure that we are able to see what the agent is doing and be able to price it. So instead of, you know, chasing every new framework as it comes out, we should just assume that the models will get super smart.

3:35The agents will have, you know, almost complete autonomy to bring the work and then we should understand what they're doing forward and do it today. And the benefit of paid is from an enterprise's perspective, it's let me manage my cost and my budget. Is that the... So we are first and foremost, we're now built for agent builders. So that could be AI native companies such as, you know, Serval or Squint or Alvarinvent. So we see a lot of vertical agents in industrials and manufacturing and consumer package goods, et cetera, or horizontal agents like IDFM. And you become a feature of their agent that they can sell to the end user?

4:10So we become their platform for making money. So we meter what the agent does, and then we rate that meter, which means that we assign a credit load to what the agent did. And then we show that bill to the customer in a way the customer saw what they got for the agent work. So right now, Agent 8 is our main customers. We're also working with SaaS companies that are shipping agents. We fundamentally believe that all software will be agentic in the next 10 years. So we're just trying to capture the world you don't want at a time. You will be fully agentic or the world? The world will be fully agentic.

4:43So there is no software that will not be agentic in five years, not even 10, in five years. Is that a byproduct of like two different worlds? one, companies that are SaaS and aren't getting with the agentic are just going to go away completely? Or do you feel like everyone's going to get on the bus? I mean, there'll be some level of just like losers and falling apart. I just don't, you know, I was the one who flew a flay in the runs on Francisco and said SaaS is dead. So like I used to think that SaaS is just going to take it lying down. It turns out they won't. They have a lot of knowledge. They have a lot of workflows embedded in their software already.

5:22So they are the primary beneficiaries of building agents on top of their stack. And they just need to figure out how to capture value. And if they don't, somebody else will. Yeah, because your TAM gets a lot larger if you have all these sacks. A hundred percent. It's kind of like a little bit, you know, I was thinking about this. It's a little like what happened to Salesforce, right? Salesforce created this API and like an ecosystem of like go-to-market tools were born, including outreach because you can access the APIs for Salesforce. The same thing is happening with MCP. So like every single SaaS company has now an MCP layer on top of it, which makes it available to any agent to run on top of that.

5:57Now, the question is, how do you capture value for that? You know, are you going to capture perhaps MCP call or are you going to build your own agents that, you know, are a beneficiary of the knowledge that you have? So this racing condition will force everyone to build some kind of agentic workflow on top of what they're offering. And you just have to figure out how to make money from it because it's not the same as proceed yeah so this happened in proceed so without getting on a soapbox this used to happen proceed and it happened at our suite well we're here for your soapbox so don't worry about it self-conscious about it all right we just talk about talk time so i'm a little i'm a little you know i'm a little you're the guest it's awesome hasn't said anything by the way he's very self-conscious so but one of the things i love about the show is that is that you've been talking to a lot of like you know former sass leaders who are now jumping onto the onto new gigs and they all sort had the same problem, which is in SaaS, you would sell 100 seats, right?

6:47And then at the point of renewal, procurement would come to you and say like, well, only 20 are being heavily used. And so I'm going to repurchase for 50 instead of 100. So now you have a contraction of it. And you have in your back pocket this add-on that you wanted to sell to them. And they'd be like, no, I'm just going to buy 50, be flat contract or down. And if you don't like it, I'll put an RFP out. Brace your hand if you haven't had that problem. Right? Like everyone has that problem. So in the agent world, you sort of do away with it by doing away with seeds. So you say, look, if you've sold 100 seeds and only two were hyper users, you don't care because you're charging for all the users and all the values those two seeds we're getting.

7:27And you don't care about 98 seeds we're having to use because those two are driving value for the entire organization. And you will have those dynamics more and more and more often in which you will have super users or you have agents deployed on behalf of a couple of users. So you don't have to sell per seed. You see what I mean? And the other seeds can do whatever they want. I think still, though, that you have to, the obvious counter argument, which is CFOs and people saying, how do I manage my budget? I mean, there's, I think the problems of consumption or usage-based pricing are, there are many.

7:55Yeah, yeah, yeah. Because what is an outcome? You know, like, how do we define the outcome? There's no consensus around it. How do we manage our budget? How do we forecast spend? Turns out that the two people that overused didn't do anything interesting. They were token maxing and they actually are completely inefficient. so how to what's what's the value conversation that's that's that's that's that's that's that's incredible so that is important now so like in sass funnily enough we never ask those questions right like you saw the hundred seats what the hell what did i get for the because seats make so much so much intuitive sense yes okay not excessively okay i'll work do whatever the fuck yeah so like in sass we had the same problem it's just nobody asked the hard questions so like what the fuck do you get for the hundred seats that you bought nobody knew and the the first chart that everyone to to show up is usage and who gives a fuck about usage like you want value so the value question was never asked so all of a sudden we're all smart about value and want to ask those questions on agents so like i think it's proper to ask those questions and you should answer those questions and this is why the onus is on the agent builder yeah to make sure that they answer the questions about what am i getting for the agent so that's that's one of the things that we do we have this this engine that allows us to see what the agent did and turn that into like normal English that a CFO can read and understand.

9:12Like if somebody token maxed your agent, what did you get for it? Or did they just build a betting app, you know, in your time? So a question around value, it has to be shown. So your buyer needs to understand what they get for the agent. That's number one. Number two is how do you manage predictability? And that's a total fair question. So there's ways around it. There's caps, there is limits, there is wallets, There's assignments. How do you do the other one though? Like how do you help? I'm an agent builder. I'm using this platform. I need to somehow show this customer that your people use a lot of this, spend a bunch of money.

9:48Here's the actual value they got a bit. How do you actually show that? How do you measure it? How do you show it? What is your relationship and working formula with the company that's buying your technology to figure out what that thing is and then show it, measure it and show it? So we have a framework and it's just a logical step. So for like, it's a ladder, meaning the first step in value is transparency. So your buyer and your buyers, you know, accountants and financial people need to see what the agent did and be described in normal terms. So once you get the bill at the end of the day, it's like kind of like the phone bill.

10:24I don't know if you guys were old enough to remember before it was all you can eat packages. You used to get this little booklet at the end of the month with every single dial that you make. My father used to make us sit down and be like, who is this? Yeah, exactly. Precisely. Sam and I remember this because of our age. Austin grew up in Pakistan, so he remembers it because that's what they did in Pakistan. Yes, right, right. So, you know, you will get that phone bill. And then, you know, of course, you're going to go to the highest numbers and see, okay, where did I spend money? Did I really call Pakistan from New York?

10:56And was that call really 10 minutes? You want to inspect. So, first of all, is the ability to inspect what the agent did and what do you get for it. So that's the problem with token maxing is that you don't know what a token is. A token should not be a unit of value for charging because a token is really kind of like the inside secret. It changes by lab and it doesn't mean anything to you. You as an agent builder, you don't have that luxury. You're not an anthropic boy. You know two models can do the same task for the same amount of spend. So you've used two models and you've spent the same dollar fee on them, but they use a different amount of tokens to do it.

11:30Exactly. That's crazy. 100%. So like token is not a measure of value. Token is an internal measure of like how do they build the LLM. You as an agent build, you have to show what the measure of value that you deliver to the customers. The first thing is transparency. Like show the customer what you did for them. Then explain to them what value are they getting for it. So like if you did, you know, a piece of research that would have taken a human being two hours, explain that to them. If you took, if you did a piece of research that consumed, you know, that took the agent, say, a day in that, but that would have saved a whole, you know, two months of work, explain that to them.

12:01Or if you produce an artifact, like a PowerPoint presentation or an analysis or something else, you have to explain that that's the value they got. And then, you know, from there to ROI, that depends on the customer, right? Because each customer has a different ROI equation. But that's why it gets really tricky, though, right? Because if somebody comes to me and says, listen, we helped you make, you know, all these presentations a lot faster, a lot better even. And you're like, yeah. And then I look at my P &L and it's just kind of the same thing as before, which is kind of on me, right? I haven't found a way for this to materialize into a true ROI in my business.

12:32But then I point back at the AI company, and this is a common conversation happening amongst people, especially in the large enterprises right now, which is like, we're using all this AI. What are we really getting out of it? So in all the functions other than engineering, this is a very real conversation. No, 100%. And even engineering is a real question. So what you're not seeing from the agents, because we work with them, is you're not seeing transparency. Like you say, you know, your build is X or you consume Y and you're not explaining to the user what is it they get for it. So explainability is like the first rung in the ladder.

13:04And as long as everybody's explaining what they're getting for it, then you can have a more productive conversation that, you know, I sat here, you know, produced 10 artifacts this last month. And you can ask him what those artifacts were and what did they use them for? But he did produce the 10 artifacts. And that is real work. Maybe you're using it the wrong way. Maybe you're using it the wrong way. Maybe you're using it for the wrong customers. But the work did get done. You see what I mean? It's when you hire a human being and you ask him to tile this place and the person finishes tile in this place, he'd be like, well, you know, you use too many tiles.

13:32It's like, well, you didn't ask me how many tiles to use. In the start of this shift, there was this huge conversation happening around how AIs want to enable real outcome pricing, right? Like not just like, oh, usage, but I help you get a deal done. And I make a percent of the deal the way a salesperson makes commissions. or I help you hire somebody and I make a percent of that person's compensation, this will be enabled. And we haven't necessarily seen that, like anything very close to that. What's your point of view on how far, how close to this we're going to get across? So agentic pricing has two boundary conditions.

14:08So one is autonomy, meaning that the agent do the thing by itself or did it need a human to do the work? And the other one is attribution. Did the agent do it by itself and you can attribute all the work to? the agent. Attribution is really hard. So this is why, you know, true outcome-based pricing is going to take some time. So you can see it in places like support because the agent is wrong autonomously. You can see it in places like inventory management or procurement where the agent did procure the part or they find the thing or they complete the transaction and like in accounts payables and those kinds of things, right?

14:45But in like the great majority of other uses, there's no real attribution. So the space in which you're pricing your agent is between effort, meaning it's a little bit north of talking consumption where you actually put in work to get that done and output that you produce something. That's really where you live in agent pricing. So most agents now that are very autonomous, they can go do a bunch of work, right or they can produce a powerpoint presentation or they can produce you know a piece of a speech or an email or whatever um and that's sort of usually where you you get to charge it's either an effort or an output or a combination of them of them both back to the point around tokens and the definition of tokens which is just really interesting are you putting pressure on the agenda companies basically your customers to show this transparency or is there an opportunity to get all the llms on the same page to understand what a definition you can't get the lms do anything Right.

15:43So you can't do shit do there. Trump Trump like get these guys on the same page. Let me call Sam real quick. I think you're doing it again. We're in the same wine club. Did you guys see the SNL skit with Dario? Oh, it was so good. So, so good. It was so good. But, okay, so obviously you can't do anything there, but like your customers are major users of these. do they have any pool to help like with pricing themselves or like no no one like no so like our customers consume llms and produce work so they're the intermediary the factory that takes in you know tokens and turns the tokens into actually right so you can now observe kind of what's happening from like a gpu outputs an hour you pay a token and then shit's getting done and these guys give you like this observability on top of that 100 so are you did you have to build the observability part of it.

16:38We actually had to. So that was our first, so it's funny, our road was a little meandering because when we started as a company, under the hood, would it look more like Datadog than we did like, say, Metronome or any metering platform? Because we wanted to answer two questions. A, what was the work that was done? But B, how much did it cost you to do that work so you can have good margins? But then, as you guys covered last week, margins, nobody started carry on margins. So I'm saying, oh, margins are important, margins are important. Get out of here. They're like, boring. The margin guy. Here comes Manning.

17:21So like, you know, and the funny thing is that our customers, when we were talking margin, we're like private equity. And I'm like, I love you guys, but it's not the right type of customer. I want companies that are growing fast. So we sort of like, you know, pivoted. We still have the cost management piece, but we pivoted into more of a metering solution that has a really flexible understanding of what is it the agent did. And that became our secret sauce. It's like we can meter the agent, but we can immediately understand what it did. And we can tune the value delivery based on the customer.

17:54So, for instance, a resolution could be, you know, I answer your question and you went away. But a resolution for you could be an answer to the question, you went away and didn't come back in a week. So now you have to keep a window open looking for that signal to come back to see if it got resolved or not. Where a resolution could be like the person got escalated to a human, but the human was able to respond within two minutes. So every single definition of an outcome or work changes per customer. You need a flexible platform to do that. So we ended up building that because that's where agents are going anyway.

18:25and then you partner with the harnesses so that you have this information and then you have like these harnesses that people can use as well or are you mainly partnering with the application layer companies that on the back end are using some sort of Our old customers are sort of like application companies and they produce two things they either produce the agent themselves and then they run it and lately and this could be a different topic of conversation lately what we're seeing is that some of our customers are on the path of the LLN company So they're retreating to start just building applications to just build harnesses.

18:57Because a customer will be like, fine, I want to use you, but I want to use my LLM. So they will be like, you know, sell me the hardness. Why do you think this draws? Because this is a really interesting thing. I saw Replit. That's a whole different conversation. So Replit just announced that they had shifted a bunch of their spend to the open source. And that's usually sign one of, I'm now going to train my own models as well, right? Like Harvey has done. and it's usually downstream of I'm trying to reduce the spend because Anthropic, OpenAI are too expensive. Then in the last few months, a lot of that spend has moved back to OpenAI because they've got such good and such cheap models.

19:34So Replit is now almost majority through OpenAI. And so you see this like oscillating a bit. It's not that clean cut that yes, if you want to reduce across, you'll only go open source. Where do you fall? Do you think that OpenAI and Anthropic will be able to capture the low cost market and the premium market, or the low cost market 100 % will sit in the open source, which means a lot of Harvey's training their own models. I think, you know, I heard that the best analogy that I heard about this is that the foundational models are like hardware companies. They're not, they look, you know, to us, they sound like software companies.

20:11They're really hardware companies. Their job is to put out the most expensive model out there and then use that knowledge to bring the cost down over time and then do it again and again and again. So like semiconductors work like this. You know, your Samsung TV works like that and that, you know, they spend gobs of money in coming up with like the latest and greatest and then they get to own everything that comes after in terms of cheaper costs. So how does it work? You know, you have Astra come out, right? But what happens is that Astra is super expensive but the first thing they do after Astra comes out and it's super expensive is they become more token efficient.

20:43so for the same task you use less tokens and then the token cost goes down so now you have token efficient and your token token cost goes down and in the meantime the distilled models are you know taking six to nine months to catch up and they actually never do because by the time the distilled models get there astro's already cheap and efficient now the distilled models and the new one is out so every time you see a model come out there's already one in the bag that hasn't come out that is about to come out a year later and on and on and on now this works well and hopefully we don't get into this because I don't know about this enough, but I'm sure we will because I just brought it up.

21:16This works well in a low interest rate environment where money is very cheap. That's about to change. And I don't know how that works. We're not low interest. Yeah. No, we're about to, you're about to see higher interest rates. And like the cost of everything is going up. Like the cost of GPU hours going up, the cost of compute is going out. You know, we're not making any more data centers. So I don't know how this plays out in the future, but that's how it's played out so far. And that would just continue to play out where the financial models. I mean, I do. Yeah, this is an area where we've stepped into this area because there's a geopolitical climate around it.

21:49There's a lot of things that are happening. But to me, it seems like the market's the market. And we're like, yes, there's a lot of noise about not building data centers. It feels like that's somewhere there's going to be a breakthrough in somewhere. It doesn't matter if Texas says, no, we're not building anymore. There'll be another state in the middle of Wyoming. They'll be like, no, we'll figure it out and we'll build it. So So those feel like short-term problems that we're going to be able to eclipse. The problem is it's all financed by debt and short-term problems. Right. There's some problems.

22:18One is that. The second one is a grid. Decades. The debt is a real thing. And then - It is what it isn't. But we've been hearing about that forever. This feels like GDP, everything has to be around our GDP growth. And AI is hinged to that. So I can't imagine. to god's ears we hope so well i mean that is i sure hope so it it does i it's it's really hard for me to sit back and say like like if ai doesn't exist we're all kind of screwed in some way it's really interesting 100 we're screwed yeah yeah especially why why why are we screwed why am i gonna say the birth rate exactly the birth rate's gone and shit everybody's gonna drink every time I'm he's doing his part I'm keeping one of the solutions I think the companies making their own models is something that it's hard to imagine that this stays and becomes a big part of what they do because it's a it's a huge effort right to train your own model so you're gonna grow your Harvey you go to some inference company you go to Fireworks and they're like we'll help you take some open source model fine tune it put your own data into it get it ready These open source models, as you said, are distilled, right?

23:34So they're six months to a year behind the frontier. And so then you have to go to somebody and they have to learn how to like customize it for you, give it to you. So you will be a year behind on your own model to the frontier model, maybe a year and a half. Yeah. That does somebody else, some new company will pop in and be like, we use Fable where these guys use some rinky-dink model and our shit's just better. Well, I'm such a big fan of the show that I'm going to quote you guys of what you said last time in this very show. Which one of us, though? It's really important. That if you have two, let's say you have Harvey 1 and Harvey 2, and Harvey 1 decides to go and optimize margins and go use the steel models and become more efficient, and Harvey 1 just goes around to Harvey 2 and says, you know what?

24:13I am using the best model, and I will never put your case law into the cheap model because Harvey 1 crushes. Perfect counterposition. On and on and on. Yeah. It's the perfect counter position to take in a moment like that. 100%. In that example, it is perfect. But I guess where I, but we hire dumb people all the time. Not because we want to. Some people want to. Who wants to hire a dumb person? McDonald's. We're commodities. We're talking about like knowledge businesses. We're talking about businesses that solve business problems for other businesses. But there's got to be some, I just feel like sometimes we - But if you go to a lawyer, you want to know that your lawyer is using this optimal model, you'd be like, no, I'm just going to go to the other law.

24:52I would be freaking out. I'm going to ask my service providers now, what are you using to serve me? If you go to a lawyer, imagine somebody is getting divorced, my favorite divorce lawyer. Okay, but Asit, we don't use Gunderson for every single legal activity. We do not. I do not use White & Case for every single legal activity. I use cheap lawyers all the time. All the time. I can't afford, I can't afford to hire fricking Wilson Sincere. But now in this new world. I'm just trying to show how many different big law firms I know.

25:32I think now that we're in this new world where we have intelligence, where all companies are going to use AI, there's going to be a new question of what are you using? Like what product are you using? And that's going to become a differentiator. It's like if you're using the shitty product, I'm a little bit concerned. Like in Paisan, you sometimes go up to the northern areas and the roads are really bad. And there are these people that take you from the airport to the resort you want to go to. And a common question is which car you want to use. Because those roads are really windy. If you're using some rinty ding Suzuki car, my chance of dying is 50 % higher.

26:07So do you ask the question? I think that will happen here too. But do you guys know people that use Harvey right now? I don't know anybody. You do. And what do they say about it? They love it. Really? Because I thought I felt like you've, I've heard you or others say that like they liked it when they first used it and now they don't. I've said like the people I know really like these things. I'm sure once you start using it, you can't stop using it. I'm sure it'll seem like how the hell did we do law work before? Remember law is jobs normally? Like law is a shitty job. Like hard. Like you get this thing that gives you an unlock.

26:39Like why not? Yeah. Manny, are you at paid looking or do you work with your customers on building their own LLMs or no? That's like you're out of that. No, I'm out of that. So what I do do is that my, one of the things that, you know, after I left Outreach, one of the things that I made myself, I made myself a set of promises. And one of the promises that I made to myself is that from now on, if I'm going to get into business again, I have to solve the whole problem. Like instead of selling tools, I'm going to sell you whatever is that your problem is. And like for everyone coming to me, they have generally two problems that they all dress up like, how do I price my agent?

27:15When in reality, how do I price my agent really means how do I do net expansion faster? You see what I mean? Because everybody went into the new logo acquisition and everybody's crushing. Yeah. You know, the enterprises are buying one of each. So new logo is not hard. Expansion is hard. You know what I mean? Sticking the landing, improving the value, getting expansion revenue is hard. And because new logo acquisition is so easy, people just took the money off the table saying, yes, fixed price, right? The problem with fixed price is when you go to get expansion revenue, it's a fixed price. It's a fixed price.

27:44So it's a hard conversation to have. So like at some point, you know, it comes to Roos and they come to us and be like, how do I price my agent? And in reality, that means it's how do I do NRR faster? And I want to answer the question. So I'm not going to say, yeah, you do that by buying my stuff. Now I'm going to say I'm going to bring a monetization consultant. Somebody has worked at Simon Kutcher, McKinsey, VC, or whatever. And he's going to sit down with you and do a quick back of the envelope design of how your pricing should be before I sell you anything. Because at that point, I'm actually solving the problem.

28:12So I do a little bit of what Harvey does in that my go-to-market motion has professional monetization consultants in it to be able to actually solve the problem. So you're a forward deployed. So we, yeah, it's a little bit like a forward deployed and we do a co-design with you and then we implement. How do you charge for that? We take a percentage of revenue. So we always take a take rate. Oh, so your outcomes and that's outcomes. Outcomes based, baby. And everyone's been pretty open to it. That is a clean, that's the cleanest outcome based. It's clean. You grow, I grow. Yeah. but you're also starting at the beginning of this.

28:45It feels like you're moving into like Chargebee or I don't know Stripe, Zora like are you competing in some way? So I do, we do run into you know the other metering systems but again our approach is solving the problem like I'm not going to sell you anything matter of fact like one of the things that drives me crazy I have customers from other platforms come and ask me the question of engagement and I'm like dude I'm a small company unless you buy my product I can't sell you just consulting. Yeah. But when we engage you get a consultant with it and we take our percentage of revenues throughout the life of the contract.

29:15This episode of Top Line is brought to you by Compound Growth Marketing. The way your customers buy has shifted before they ever reach a rep. They're asking at LLM who to trust, who to shortlist and why. Decision-making is impacted by investment at the top of the funnel and most B2B marketing teams under-invest because it's difficult to measure. That's the gap and it's expensive. Compound Growth Marketing bridges this gap for many of the fastest growing companies in B2B SaaS and cybersecurity. Their PDS, or Predictable Demand System, unites paid media, answer engine optimization, and GTM engineering into one discipline engineered to produce pipeline you can actually forecast.

29:51Backed by over 100 million in media managed. No vanity metrics, no guesswork. If you'd like a free audit on funnel media, including AEO, reach out at compoundgrowthmarketing.com and mention Pavilion. AJ, have you reached out and did you mention Pavilion? Did you say anything? I did. Last time we had an ad read, I actually did reach out and I have a call scheduled next week. Oh, that's great. So I truly do. And I mentioned Pavilion. I put the name in as well. Thank you very much. We will chat back in next week. Asit, reach out to them. I don't spend anything on marketing. Okay, fair enough. Well, anyway, folks, please hire compoundgrowthmarketing.com so we don't become unhoused.

30:31And please mention Pavilion, but only one L. This episode of Top Line is brought to you by DevRev. Guys, tell me something. Tell me something. I'm excited about. I'm really, really jazzed. Jumping out of my seat. I am energized by computer. It's not the computer that you're thinking, but it's called Computer by DevRev. If you work in GTM or sales, you know what it's like. You're running a dozen deals at once. It's on you to remember every detail. Who said what, what you promised, what needs to happen next. Things slip. A follow-up gets missed. A deal goes quiet. Computer by DevRev is an AI teammate that fixes that.

31:09Computer uses something called shared memory to hold the full context of every deal across every communication channel, Granola, Salesforce, the whole bit, even across the rest of your team. So nothing falls through the cracks and you always know where you stand. Your deals keep moving and you stay one step ahead. Take a look at computer.io forward slash top line. Asad, react. I think computer is a classical thing. You need it. these people are cool you know what i want to meet the founders of devrev because they're interesting they're here i hope so i want to go demo computer yeah and also understand for the last couple of months yeah we're i'm definitely gonna see i think the three of us need to go up there we're gonna go right now this episode of top line is brought to you by pavilion We're coming off the best Pavilion GTM we've ever done.

32:00Ten years after Pavilion started as a few operators getting together for dinner, we now have a community of over 10 ,000 across 50 chapters and 15 countries. And at GTM 2026, we laid out what we think the next decade looks like. AI is making capability abundant, but judgment, relationships, and access are becoming more valuable, not less. The GTM leaders who win from here will be AI fluent, P &L fluent, cross-functional, performance-oriented, and deeply connected. So we're building Pavilion around exactly that. We launched GTMSkills.com along with new programs like Leading with AI and GTM Engineering School.

32:43We introduced DreamFuel, bringing elite performance coaching to operators. And we're expanding access through Pavilion Scale, Pavilion Gold, and Pavilion Ventures. connecting members not just to ideas but to the people opportunities capital and ownership that can change your career the next decade of gtm is going to look very different from the last one if you're an operator and you want to be ready for it come join us at pavilion go to joinpavilion.com to learn more you mentioned you mentioned lessons from outreach i'd love to because that's something else public that you've been talking about which is what you've learned and what you would do differently?

33:21What are the big lessons? What did you learn? So first of all, how do you look back on that period fondly? Yeah. With anger? I, it's hard because it, you know, outreach is my baby. You know, I took it from nothing to what it is right now. And, um, you know, there's some things I would have done differently for sure. And there's some things that I will do again. Um, and there's some lessons that, that, you know, I'm, I think I'm overthinking at some point, that so for instance you know i decided that i wasn't going to hire a product marketer like i can do my own positioning like i can do my own enablement and i realized that was dumb like in my first one of my first hires that i reached was a product marketer and there's a reason one of your first marketer one of your first hires hires it paid is one no no i haven't hired product marketer yeah but now i'm upside down i feel like i'm working all the time because i don't have a marketer and i should have done that because i did that at outreach and my life was glorious you know what I mean and I didn't do that here and I'm like ah crap I should have done that but one of the things so what did I learn I will I will never build workflows again I will always going to go for a system of record so even if it takes me forever to build you know this company I will build a system of record so you won't sit on top of anything meaning like you're we will own the record if the record is important we will own the record you were like beholding to sales exactly nope exactly we own the record of what did the agent do the credits that you're charging and the auditability of the whole thing because that's a financial record So how much you charge and whether you get for what you charge, you know, it's a notable thing.

34:50Does that make it harder? Like, I don't, I mean, we're going through SOC 1 type 1 stuff right now, but does that make it harder for policy stuff or financial compliance? Yeah, of course. Yeah, of course. There's a much higher barrier to enter because you can't screw up. Yeah. And you can screw up in a world in which, you know, the code is writing itself. So the likelihood of screwing up is very, very high. so you know for us you know a billing error or a or a credit miss assignment is a common thing and we have to take it as such you see what i mean but so back to the thing that i want to do is i don't want to build worthless anymore because worthless is like being a shark you know i mean you you're only constantly moving you you can never rest because your customer would turn any minute and you always have to like continuously deliver value and which is fine but i'd rather you zone the record and do a good job on the record.

35:39What were some like lessons or like some principles you took from that, that initially were like, you're like, I'm not going to do this. I'm going to do this. Where now, like a few years pass, you're like, that was maybe the wrong principle to take or the wrong lesson to take. You had to like, give it up. Maybe anything of that sort. So tell me more. Like I find that sometimes when I've gone through failure and I like immediately sit down and I process it, I have all these interesting insights. I'm not going to do this. I'm going to do this. And six months in, the thing that I told myself I'm not going to do was like an emotional takeaway from my reflection of that situation and was actually like a limiting belief or like just the wrong insight.

36:20Yeah. Yeah. So this program marketing thing was a wrong insight. Like I thought that I didn't need it in a reality. I needed like more than anything else. How did you feel about your fundraising strategy at Outreach? Are you, did you i i felt like that i shouldn't have raised the last round and the reason was we didn't need the money and i sort of like that i felt like as a first time ceo i got pressured by like the market and like the momentum and i got taken away by everything else um and now have a much more deliberate you know fundraising strategy um in fundraising your previous rounds determine in your next round.

36:58So you should never take an inbound term sheet because if somebody sends you an inbound term sheet, that means that there's other people out there that will give you a term sheet. And of course that inbound term sheet will have, you know, an exploding date and there will be pressure and all this nonsense. There's never pressure on an inbound term sheet. Like actually, I heard an interview from the guys at Bending Spoons that said that, you know, when I make you an offer, that offer is valid today and is valid a month from now. The price is not going to change. And I'm still buying your company for that price.

37:31And VC should be the same way. If a term she comes in and they value at something, the value is not going to change by next week by when you want this thing signed. So you should take that and be thinking for your offer, I'm going to go shop it. And that should be totally fine. Travis spoke about this a lot, right? Around his process of fundraising now and how he runs this competitive process. He is. He's next level. I am, you know, I don't even hold a candle to him. But like, you know, you should always go look for the right partner for whatever it is that you're doing. You know, whether it's, I was talking to another founder the other day, and he's like, I only take operators.

38:03And that's a hard rule. So like, if you're a professional VC, can't be on my cap table. That's fine. So for me, like, I just take big brands. And, you know, call it whatever, but that's just my. I only date supermodels. I only drive Ferraris. 100%, yeah. Well, outreach was so story narrative driven. I mean, I, back in, we were an outreach customer replaced, I think ToutApp, maybe Yesware, like throwback. One of the two. Yeah. I mean, it might have even replaced SalesLoft because that's like, we were using the first version of SalesLoft where Kyle got like locked out of LinkedIn. Yeah. but um i remember looking like when my cro uh this company was trend kite my previous company i founded cero brought it and he was like aj i think we're gonna switch and i was like i don't know and then i looked into the story and i was like actually this is a cool company and the sales team it was we had just become a real true like mid-market company at that point so we had moved beyond and looking at it i was like oh i understand the story which i also think from a VC standpoint was also like the big reason why like VCs just love a great narrative they love a great story and you told such a great story there so it was easy to get caught up in that hype cycle as well and in hindsight it's easy to say like well I shouldn't have raised the last round but you didn't know it at the time right right right right so that's one of the things but it does seem like you're thinking about that for paid yeah I'm just more intentional while we're building like we know that a system of record just takes a little bit longer to build Workflow is far easier to build.

39:37You know, selling, you know, small mid-market deals is far easier to sell than selling bigger deals. But, you know, I see that - How much of that is resonating with the investors and other stakeholders in the market? Because you're in a market where, you know, this massive revenue growth, you hear of companies that have run from no revenue to a hundred million in nine months all the time. You hear of all these things and you are taking like a counter position. You are selling into the enterprise and you're building a system of bracket. Those are like two slower things by default, right? So that's a lot of the explanation and conversation I imagine you have to have.

40:11How receptive to that is the other side of it, investors, et cetera? Do they get it? You know, that's the other thing that comes to the experience that you just don't give a fuck anymore. Like if somebody will, somebody, look, when you are a take rate company or some of the fastest growing companies in the planet, you have your okay. and and as long as you internalize that and you tell that story you're gonna be okay and um and so like questions like oh how you're different where you you know how do you grow fast etc you know the funny thing about growing fast is like nobody talks about like when the starting point is of growing fast you know i mean you build build build build build build and then you lock in and you go real fast so you can make that starting point anytime you want you see what I mean, and then you can be a fast growing company.

41:00It's like NVIDIA, you know, to fumble around for a long, long time. And then all of a sudden it went up and did it right. So like, it's one of those like cute, fast growing stories that took out 30 years to make. Do you reflect on like, you know how you said, I wouldn't have taken the last round. Maybe if you didn't take the last round, you'd still be at Outreach doing your thing, but then you wouldn't be doing this thing. Yeah, like that's how, so I love that question because one of the highlights of my career at Outreach was that I had a chance. I became like the highest performing company for M12, which was Microsoft VC at the point.

41:33And because of that, I had a chance to do a keynote with Satya. And for the keynote with Satya, I got to sit in the green room, you know, with the man himself, right? And, you know, we shoot him the shit and he's like, and he's like, you know, what's going well, what's not going well. And I was like, and I just like gave him like, you know, a good 10 minutes of what's not going well. Because as a CEO, you're just like, you know, you just take the branches, fucking hit yourself. and he looks at me and then there's a lot of like i would have i should have i would have and he's like he looks at me and he said manny life is path dependent you're only here because of the things that you did if you didn't do those things you wouldn't be here that's for sure so everything that you said is absolutely true so like i could regret x y or z but then i wouldn't be here and that's and this might be one of the most important this may be one of the most important I think I'm loving what I'm building.

42:21I'm building exactly what I wanted to build. I get to see agents in every single walk of life. Like, my job is fascinating, dude. Like, I can't, every time I wake up, I'm like, I can't imagine I'm doing this. And being a second time founder in this platform shift has like huge advantages. A hundred, yes, a hundred percent. Huge, huge. A hundred percent. So like, you know, I love my team. My team is small, it's high performing. You know, everything that I wanted to do, you know, at outreach, I'm doing it now. I wanted to keep the team very small, very high-performing. I wanted to keep revenue growing higher than growing headcount.

42:56I wanted to hit operating leverage early. I wanted to keep ACV high because it's easier life to live. I wanted to keep the board small. You see what I mean? Everything that I wanted to do happened. I wanted to spend more time with customers doing fascinating things. The kind of agents that we get to see in the wild are just crazy stuff, dude. What are some of the craziest things you've seen? What gets you the most excited? Um, we have this customer called Oliver and Vendt and they make agent chemists. So they're deploying big chemical manufacturing companies and these agents are making chemistry, are creating new formulations.

43:34So I don't know if you know this. That's pretty cool. I was, when I was, you know, we were onboarding Oliver and Ben, they told me this following story that it turns out, I don't know if, yeah, I haven't looked this up, but it turns out that a few years ago, the red dye used in the tea of the Tylenol was considered a carcinogen. So all of a sudden, you know, Johnson & Johnson has, I don't know, two months to change the formulation of that tea that goes in the Tylenol. You know how expensive that is? Because it has to be the right color and it has to be the right taste and it has to, you know, use the stuff that you already have in the many warehouses around the world for making Talena, which is one of the highest selling drugs on the planet.

44:13So that formulation is a, you know, $20 million exercise. Albert and Ben can run five agents, go find all the possibilities, come back with a formulation in an hour, I don't know, four hours, give or take. That's amazing. So this is the kind of stuff I do on the daily. That's true. So I love my job. You see what I mean? That's true. Examples like this, I hear every day of something else crazy like that. I was just talking to an investor yesterday, and he works at a firm in New York, and they do biotech investment. That's their primary niche. And I was like, hey, how much are you using AI? Like, how much has it changed your job?

44:46He's like, my job is now 25 % of what it used to be. And so I was like, crazy. Like, what are you using? And there was no like, oh, I'm using RollGo. I'm using this. He's like, I have plot, chat, GPT. I oscillate between the two. So it was very like basic workflows. but he is now spending 25 of the time he used to spend doing things and so now he has 75 more time to do and it's just really interesting to see like where it works it works that well right yeah and and like you know problems such as there's this other company uh that we work with called squint and they make videos for for for machinery and and i mean the main problem in manufacturing is that people are not manufacturing anymore so like you know kids don't wake up and be like you know I want to be a machinist.

45:32So, so they have to figure out how to manufacture more people or like how to train more people that come from other walks of life to like, you know, operate the thing that is going to make, you know, Cheetos or whatever. So, you know what I mean? They don't make themselves. So somebody's going to have to run that. As it relates to people and loving your job, outreach seemed like such a, I hesitate to use the word family culture because everyone likes to use a family, but it did seem like a very close-knit group. And you worked with some amazing people. Probably a lot of people you want to work with again.

46:00Yeah, of course. Yeah. But at the same time, building a company in 2026 is like, is not like, it's a kind of feels taboo to say, I want to build a people team or people culture. Does that change how you think about that for paid? um so first of all you know i'm building the company uh a little bit different than the company's a lot smaller than we than we did in the past um i you know there's some aspects so we you know we talked a lot about like you know being close-knit more family oriented uh but at the end of the day like it's a sports analogy more than a family analogy like we're all here to perform Right.

46:39And the thing that was clear to me from outreach is that the only thing that matters is winning. Full stop. The culture is all great. Yeah. As long as you're winning. Say more about that. Is that, is that, you know, culture. Did you find that you were celebrating things at outreach that didn't correlate to revenue growth or success or winning? yeah i i feel like i feel like you know sometimes culture took a second seat to winning and winning should have been the main thing um now you have to win with other people right it is that oh i'll adage that if you if you if you want to go fast go alone if you want to go far you know go together uh so you still have to bring everybody along and that was that was the you know that's a that is the biggest job of a ceo is to bring everybody along everybody aligned but let's not confuse culture with winning.

47:28The culture is of the service of winning and winning together. When you're winning, the culture feels a lot better than when you're not winning. So much of culture is just morale, right? I feel culture is the exhaust of winning. 100%. If you solve for winning, you'll get great culture. Exactly. Nobody is in a losing company but I love the culture. This is a great place to win. They might not pay me this month. It's fantastic. When you think about the companies that we have right now, there's a lot of companies that are winning, but it's a little bit more nuanced, right? Because there's this idea of experimental revenue, experimental spend.

48:06And so what are you seeing over there? Because when I speak to companies that are really large enterprises buying technology, you see that in a lot of the departments that are not software development, people are saying, listen, we've spent a lot of money and we have like things it does for us that are cool and interesting, but it's not showing up in the P &L from an ROI productivity perspective just yet. They continue to spend because of fear and excitement, but at some point that's going to stop till the thing gets in. How much of the revenue you see at these AI companies do you think is experimental revenue?

48:40It's a great question. So I think that that is shifting. I think that that was the case maybe last year. I think this year people are sort of like locking down into longer commits meaning they know they're going to spend on ai they know that ai needs to do this body of work and they're going to pick a vendor and then they're going to go for it um that said ai is very it's like almost like sub one percent penetration in the enterprise because like we're still doing the whole like you know i'm going to do the old workflow but faster with ai kind of thing as opposed to completely rethinking the role They rolled completely agentic.

49:18And that's what I'm excited about. There's this whole new breed of companies who are selling software with transformation. Because in the past, we all sold software and we looked at professional services as taboo. And that is the biggest unlock that Balanthia brought to us. Do you know some people are saying this, that the unlock is the forward deployed model with it? But isn't the biggest issue the fact that the technology is probabilistic with a pretty high error rate even at the frontier? Like even the best models still will make mistakes three, four, five percent of times on tasks that have multiple steps in them.

49:50And so you can dump a bunch of FTEs on top of it. But the FTE is not like, you know, they're not doing the work. The AI is going to do it and there's going to be an error rate. And so that restricts what you can trust the AI to do if you're an enterprise. Yeah, but you had that problem before with human beings too. So, like, you know, we keep forgetting that AI is not a magic bullet. Like AI is a path that you jump on and things get faster, right? So like when you're seeing a lot of errors or a lot of misfires, then you may have a hardness problem. Maybe you have a memory problem. Maybe you have a context problem.

50:23Maybe you have something else problem. You see what I mean? And so like you start troubleshooting down the hole until you fix it, right? Because the end goal of doing all that fixing is going to be a much higher performing organization where you need a lot fewer people to do the same work. You know what I mean? Or you have a much more throughput. or you have a higher quality. When we talk, I mean, we hear this all the time, right? That people, instead of just enhancing the existing workflow, you need to reimagine. Is the act of reimagining fundamentally to just sell the outcome? Is that what needs to be reimagined?

50:59Instead of saying, you know, you still have to enter information into this form and then that leads to another form and that leads to another form and that is work and that is what I'm selling. Instead, it's just, is the clearest articulation of the way to reimagine, you know, the, everybody uses the electricity example. When electricity first came out, it didn't make any supply chains faster, but then they rebuilt the factory floor based on the existence of electricity. Yeah. Would you say like the single thing that everybody needs to start thinking about is just start with the outcome in mind?

51:29Yeah, of course. Like that, at the end of the day, like the process exists for a reason. And that reason is the outcome. and like all the middle steps are all negotiable. At the end of the day, you're just looking for the outcome. Because I think when I think about it, now we've gone to this world where most interfaces are the chatbot, right? It's like a blank screen and you're supposed to tell it what to do. And I think to the point, the next step has to be, we know what you're supposed to do, right? So just do what the thing you're supposed to do is. Exactly, exactly, exactly. Instead of asking me to imagine something.

52:02100%, 100%. It's nice to talk to the computer, but the whole point is like, I don't have to talk to the computer. 100%. And it's a funny thing. It's like what we lack, like the impediment between like where the technology is at and the benefit that you get from it is this lack of like context and awareness and like all this other, you know, things that are happening in the company you don't have access to. And that's why like, you know, I feel like there's a whole new breed of new startups that are coming out that are going to fix that problem on the enterprise and they're going to make a ton of money.

52:30I think the fear I have is like till the enterprises learn how to sandbox their own design. So like when you create technology, you test it in the sandbox. It's a controlled environment. If you're an organization, you almost have to have like this thing you create. You're like, let's have the sandbox for operations on how we will do this end-to-end agentically in the best way possible. Let's test it, let's stress it, and then let's deploy it as the main thing. Till that happens, the space at which enterprise will move is the same pace we've always seen them to move which is like seven eight years later we'd be looking at these companies and they're like slightly different but not that different that seems like the pacing of the frontier i think that's the thing that is i think that which that should give us comfort you know that things do move slower like uh than you know the the effective altruists and a polycule in san francisco are like everything's going to be over next year right No, it's not.

53:26People make slow decisions. It's funny because every technology has the same rate of diffusion. I mean, the internet went through the same thing. And we're like, oh, my God, everything is changing so much quickly. Like, no, at the end of the day, we're human beings. The majority of knowledge is intrinsic, not extrinsic. And it's going to move as long as the uncovering of the intrinsic knowledge. Crot-Crola has applied to power. The pace Crot-Crola moves at. Thank God. That's going to give us all an opportunity to make money. When you're bringing on customers or just kind of the top of the funnel and the go-to-market, are you doing much more diligence into the organization to figure out their durability?

54:00Because to the point of this year, companies are locking in. Well, they're locking in and picking the winners. Yeah. And so for you, it seems like it's really important to pick that winner, find that winner that would be a potential customer. Because otherwise, like, you know, from an NRR expansion, you potentially might have an issue if a company's like, just not thriving. It's funny that it's a good question. That's the other thing. I kind of don't care because I don't know who's going to win and not win. And I can't tell. Otherwise you'd become a VC and make like a trillion dollars. Right, right, right.

54:31But what I do do, the qualification is fairly, right now it's mostly inbound because most people have a problem with their pricing and agents and they come to us. So our question is more like, you know, is this a CEO problem or not? If it's not a CEO problem, we punt. And if it's a CEO problem, and then we solve the problem. Does that mean you're getting on the call with CEOs? Yeah, all the time. This is why I love my job. I get the stories about it. Are there even boards that are joining these conversations? No, they're usually precipitated by board conversations. Yeah. Of like, this came up at the board level, somebody on the board has to allow us.

55:05Are you doing any partnerships with VCs or anyone? Not officially, but I mean, every VC knows me. That's funny. I took so much money for outreach. I mean, like, I don't know, who doesn't know me? It's funny. Yeah. And it sounds like you're going, you mentioned you're going to CEO off sites and like conversations. So I'm sure at these tables, there's a lot of these companies, whether it's harness companies or application layer companies that are all talking about this problem as well. So those events are probably go. Yeah, I mean, the question is more scale, right? Like if there's only one man, it kind of, you know, travel takes up time.

55:40But no, but that's the thing is that at this point it's evangelizing that there is a solution to this problem. And this is one of the things I love on my outreach experience is that outreach was an evangelizing problem, right? It was convincing the world that sales could be a workflow, that it wasn't just like, you know, the long, you know, gunslinger who's just going to bring it in. And like, no, you have a pipeline. The pipeline is a fairly understandable workflow and there's a way to generate it. And as long as you have coverage, you're going to make the number. And like, that wasn't clear back then.

56:10And, you know, over time it became more clear. So it's the same thing now. It's like, you're going to sell work. The work gets priced in these foundry conditions. The work is all what the agent did and the work has to be observable and auditable. You know what you're saying? Like there's going to be all these AI companies you ingest and some will have accelerated revenue and then they don't provide the value and they'll churn in your own. I think like this world will look a lot like the VC portfolios. You're going to have like companies that you ingest into your organization. We at STA will have them, right?

56:41We're going to have a bunch of these AI companies that have raised a bunch of money and are like, let's go out and hire 30 salespeople. And not every AI company that's going to go out and hire them is going to make sense. But the one that takes off, that NRR will make up for this drop-off. I don't know if that's true because that's problematic in the sense you're going to have a 90 % failure rate. You'll have one out of 90 % failure. Because that's what the VC, that's the, you said, if it's kind of like the VC model, that's literally the VC model is like, I'm going to have one company that will make 100x, the rest are going to go whatever and not do anything.

57:12I mean, that's probably not. I don't think that's the model though. Like if you look at like a top VC fund, it is two to three return. With growth equity, sure. But I'm just talking like lower level at that point. Like see the seed in series A, 90 % of those companies are going to fail. Yeah. But like growth stage is like three, four would return the fund. I think in general, his thesis is agents need observability so we can figure out what to charge for it. A hundred percent. Right now we only engage with people that are a little bit more scale. Because those are the ones who need them. So the early stage ones are just taking away logos quickly and then doing fixed price.

57:45And they usually box themselves into a problem by the time they race their race A or B. And that's when they, you know, when the first renewal conversation happens, that's when usually we're like, oh, shit, like we screwed this up. We need to get ourselves fixed. Oh, that's interesting because you were working with a lot of companies that haven't really gone through renewals. Exactly. So we don't work with, a lot of them we don't work with because they haven't felt a problem. So pricing is not a problem if you're just taking down new logos at a fixed price. Because they told you, well, fixed price.

58:09This is not a priority. It's something that they're thinking about, but it's not a priority. The moment that they come to a renewal conversation, they're like, oh shit, I boxed myself into a fixed price. They're like, all right, so I have a bunch more that I just did last year that are about to come to renewal and I don't have a way. Fundamentally, your perspective is cost of inference, cost of compute is not going down. It's not going to go down. It's going to stay high. And so that if you're a CEO, that's like to the point of the Harvey comparison is like, all right, I need to figure out a way to get better margins because I'm burning too much cash and I need to get to something that looks at least like traditional software margins.

58:48You're screwed. Is that your point of view? No, I think so. Yes, until the last part. Like you have to get to a sustainable business. Okay. And a sustainable business has to be able to flex around the value that you delivered. You know, your cost is your cost, but like if you deliver some value, you need to be able to capture that. And if you're capped because you give a bunch of fixed price contracts and you're pricing a way that is not aligned to how your agent performs at work, then you're not going to capture value and somebody else will. Well, you were saying is the days of software margins are over.

59:1980 % are over. And I guess my response to that is, isn't that sad? Why? Why? Because SaaS was the greatest business model ever created. And these are all inferior businesses. My first job out of business school. So I went to business school. And then second month into my business school. Did you go to HBS? Yeah. And so second month. But I went to a different HBS. Because my second month of business school, 9-11 happened. So all the jobs that were available to me are gone. So by the time I come out of HBS, there's not many employers. hiring and the ones that were are not hiring as many. So one of the ones that are hiring is Amazon.

59:59So, you know, a guy called Andy Jassy, you know, thanks for an offer. And, you know, he wasn't a big deal back then. And when you work in Amazon, you are used to tiny margins. And you actually have to memorize the operating margins of everything at Amazon because you want to make sure that you're not, you know, Jeff Bezos, the famous expression of Jeff Bezos is your margin is my opportunity. A hundred percent. Like, and he, so you know that books are 20 % and you know that electronics is 7 % and you know that jewelry is 3 % and you just know those numbers call. I still know them because you just memorize them.

1:00:33And then, so like, you know, and then I went to Microsoft and I was like, what the hell is this? This big margin business situation. And then when, you know, when I started a company and I looked around the SaaS world, like 70, 80 % margins, what you do with this luxury? Like, you see what I mean? Spend it on growth. Yeah, you do. but even that is inefficient you know so the the um the ability so i i feel like you know we were in an anomaly for our longest time because money was cheap so you can you can do the things like that but i don't think that like at the normal business you know we'll have those those kinds of margins like that you know but you do need positive gross margins right of course yeah at some point but i mean there's a difference between positive gross margins and 80 percent But I feel like a normal software business will be running on 40 % margins, 60 tops.

1:01:23Especially when they - Gross margins. Yeah, gross margins, yeah. So from there, then you have to figure out how do you take out your sales and marketing and your DNA and your engineering costs. That's why you got to keep OPEX extremely - I think the number that everybody I hear from is that at the end of the year, if you had 30, 25 to 30, that's like a good level to be at. Gross margins. Yeah, gross margins. Gross margins. You'll probably - Like that's a negative margin business right there. I know, I know, it's so hard. Because you probably don't have like, the technology is breaking through quite fundamentally every year right now.

1:01:53So you don't have that period of time where you get negative margins, you have to increase prices, rationalize. So then your blended yearly falls at 25, 30. And that's like - But there's so many variables to this. These inference down things are below that. So what I'm seeing from my perspective in these agentic businesses is that they are hyper-verticalized. So they're solving very narrow problems in very narrow industries. And there may not be huge, but they will be very profitable in those niches because they're, you know, if they own all of it, it's kind of like Viva. You know, Viva was - What a great business.

1:02:25What a great business that was. They do like three things, but they only do it for pharmaceuticals and that's it. And it turns out it's a huge business. Yeah. And I feel agents are going to be the same thing. You're just going to have a bunch of Vivas. The Sierras and the Decadrons, et cetera, which have like kind of started very verticalized. But like you look at Sierra right now and it seems like they've used support as the entry point and now become a broader platform play inside these enterprises. How do you think about things like margin for them then? I would imagine they would have low margins too.

1:02:56Well, so Sierra is a bit of a different animal because they went for outcome-based pricing right away. Yeah. And unlike everybody else who gets bent around the axle, oh, what's an outcome, blah, blah, blah. Like they have a catalog of outcomes that they go through. You see what I mean? So they don't have a outcome. They have a bunch of different outcomes that they already tallied up for some other customer. That's gangster. What's an outcome? It's this, this, this. What do you want? So it's kind of funny because sometimes I explain that you should have a catalog of outcomes and send different depending on the customer.

1:03:26And people get freaked out. I was like, whoa, whoa, my sellers can't do that. I'm like, you have a bunch of dunces in your company. You should hire smarter people. That's what Sierra does. And so their outcomes are great. Their margins are great because they sell outcomes. So when you - But do you think that's uniquely possible and support and becomes less and less possible as the broader platform. Of course fucking not. Like, they're just smart. Interesting. And they are, you talk about limiting beliefs. Like, I feel like I'm, you know, remember that movie with Bruce Willis where like he sees dead people?

1:03:53Six cents. Right, six cents. It's like, I feel like that. I'm walking around and you're seeing a bunch of people with limiting beliefs. And it was like, no, if pricing is not this thing in the box, then everyone's going to die. I'm like, no, you should make your pricing way more customizable depending on what your customer value is. For the record, Bruce Willis was actually dead in the movie. He didn't see dead people. The kids saw dead people. Oh, wow. I don't know if that was necessary. You're that guy at the party. Everybody's free to show you the joke, and you're coming like, I just don't want anyone to get confused about that.

1:04:22Manny, I was totally joking. You were like, what the hell is this guy doing? Is he doing it again, Mark? Oh, my God. We'll call 19 in the room. Manny, thanks for joining us on Topline. We loved having you. Thank you. We're going to have you back. Thank you. Great to see you, and congrats on paid, and we're super excited to support your journey. Thank you. I appreciate that. Thank you for having me. This is awesome. It's fun to be in the room when the magic happens.

1:04:54If you are looking for more Topline, check out the Topline newsletter at toplinemedia.substack.com. That's toplinemedia.substack.com. Or if you're a video person, check us out on YouTube. Topline-media is what you want there. Have a delightful day, everybody.