AI-Native GTM and the New Science of Scaling

Julia Nimchinski:
Welcome to the show, Mark.
Mark Roberge:
Hey, Julia. That’s one thing I uniquely love about your shows, is we get these, like, serendipitous transitions to see old friends that we hadn’t seen in a while. So, wow, it’s cool to follow Brian. It’s awesome.
Julia Nimchinski:
Very cool. Excited to have you here, and just to set the stage for the audience, Mark Roberish is one of the defining icons in modern GTM. As HubSpot founding CRO, he built the revenue engine from zero to IPO, the learnings captured in the sales acceleration formula. And he now teaches at Harvard Business School and co-founded Stage 2 Capital, giving him a rare view into how hundreds of leading startups scale, and why they don’t. And today, serendipitously, we are releasing Stage 2, releasing the Signs of Scaling first chapter. And we’re gonna send it all, I believe, like, exactly now, to all of our community. Nice. So, yeah, perfect timing, excited to have you here. What’s Newark?
Mark Roberge:
Yeah, well, I can… I’ll end on the… the book launch. But yeah, I mean, I think we spoke 2 months ago, and it’s crazy, because, like, in AI world, like, that means it’s 2 years ago. So it’s exciting to kind of break down some of these elements, but I’m happy to share kind of what we’ve seen. I think we probably… look at, 200 to 300 AI startups a month, and, you know, usually invest in, like, one every two months. So happy to kind of share some of the trends we’re seeing. And of course, I have the great pleasure of heading over to HBS every week, to teach and get you know, test out some of these hypotheses and let students from every continent across the world and every industry bane on it from their unique perspectives. So, happy to share some of those trends.
Julia Nimchinski:
Very excited. Well, are things first, you know, what I’m gonna ask? Will AI disrupt the horse and buggy?
Mark Roberge:
Yeah, that was, an interesting one. I guess we can get, like. way societal for a second, which I do think we need to talk more about, in tech, because I think most technicians, aren’t in it for personal gain, first and foremost, but, like, just want to bring great product to market and change the world in a positive way. And I think, like, we all need to reflect hard to make sure that we’re doing that. And this one’s a real tricky one. You know, I don’t think I saw as much… fear during, like, the internet age that a lot of us live through. So I think we gotta just, like… I’ll share a couple narratives that I hear, just as we think about, like, the impact of AI. The first one is, like, well, is it ultimately gonna kill us, or, like, should we, like, slow it down? And I guess, like. I don’t know if I have… I don’t know if I’m deep enough in to know, like, the life-threatening aspect, and I just… I don’t know. I don’t know about that one, but I think the one I do think quite a bit about is the economic impact of, like, whether it is compatible with capitalism, as we’ve created today, and whether it could create massive unemployment, which none of us would want to create that. So I think the first question is, should we just, like, slow it down and, like, let the regulation and, like, the government policy catch up so that… and then retrain people accordingly? And unfortunately, it’s more complicated than that, because not only are we dealing with an economic shift, but I’ll just kind of quote what others are saying, which is. There’s also a military risk here as well, that, like, many powers are, whether political, government, or other, you know, representation, are aggressively innovating on AI, and should someone, like, gain a massive advantage. They could take over everyone else. And so, like, that’s what people say, and I think that there’s a reality to that. So I think it’s extremely dangerous to, really kind of over-regulate and slow things down from just a… like, a military protective standpoint, because that power could fall into the hands of an entity that doesn’t believe in human rights and freedom as much as we do. So I think, you know, we have to, as technicians, like. probably educate the broader society on that reality, so that they understand what’s at stake here, and take the adoption of AI more seriously. So I think the second question to that, Julia, is whether it actually does is kind of compatible with our version of capitalism. Does it create massive unemployment? Most of the economists I know argue that it won’t. They say that if you go back in the archives of the advent of the Industrial Revolution, of electricity, of the invention of the computer, the digital age, and even the internet, there was massive fear around job loss and unemployment. And in each of those phases, more jobs were created, and society was better off. So, I think a lot of the economists are putting their hat on that. That’d be great. I think we got a shot at it, but I also think we need to plan if it doesn’t occur. And that’s something that I’ve just been thinking about around, like, how would that unfold if you know, 30-50% of the jobs as we know it today are done better by AI, and it’s difficult to retrain some of those folks to be, like, applicable today. And I guess the only things I’ll add, Julia, is, like. a few observations I have about that is… The U.S. was a really cool journey for freedom. and human rights, and I think it’s really made the world a better… not perfect, but, like, really… really did a lot around those missions. However… We have dropped quite a bit in the happiness index, for being… the highest GDP growth and… or the highest GDP… We’re… we’re number 22. And I think there’s an opportunity for us, if we were to go down… if… If AI did not become compatible with our current version of capitalism in the same way that we’re rethinking sales funnels and rethinking organizations and processes, there might be an opportunity for us to rethink our economic structure And the society here to increase happiness. Maybe to work less, but to earn more. Maybe to have a better life, even though we’re working fewer days and fewer hours, having more time for our children and our elders, and our community, and our neighbors and our hobbies, and our health. And having a smaller government, and lower taxes. Like, that’s what something like AI does, is it helps you to optimize everything for the better, so… Maybe we’ll end up there.
Julia Nimchinski:
In current administration, do you think that that’s… How things will unfold.
Mark Roberge:
Oh, no, no, I don’t, and I don’t think either. Like, whoever had won, I don’t think… I think right now, the other thing we have, unfortunately, is we have the divided states, and we need a leader. I think we need a philosopher, an economist to rewrite The Wealth of Nations and be our modern-day Adam Smith to lead us into what’s next. And I think we need our… a new leader to unite us. We’re very divided right now. The United States is more powerful than divided states, and the divided states compromises and threatens Human rights and, and, and freedom, for the whole world. We’ve been in worse situations before. During the Civil War, we were shooting at each other, and leaders came and united us, and I’m confident a leader will arise and unite us again. But I haven’t seen that person yet.
Julia Nimchinski:
Fingers crossed. Well, let’s transition to GTM. It took a lot.
Mark Roberge:
I think we need, like, an in-between in there, Julie. So how do you do a demo? Let’s do it. Let’s do it.

Julia Nimchinski:
Okay, AI nativeness. You talk a lot about this these days. Yeah. How do you define it? How do you think about it? What are you seeing?
Mark Roberge:
Hmm, yeah. Yeah, I think everyone’s kind of, like, asking how… how’s sales gonna work? How’s go-to-market gonna work? Like, in my world, of course, how’s finance gonna work? How’s, like, co… you know, R&D gonna work? How’s all this? But, like, Been thinking and writing about the go-to-market side, And I think the way I like to frame it is phases that represent Incremental abstractions of human work away from today’s point of work. And there’s 4 phases that I would… propose. The first phase has become quite a reality in 2025, and I think will become quite mainstream in 2026. And that phase is, like, we still have humans selling to human buyers. But, far more efficiently. I think the KPI within that phase will be selling time. So selling time in the world of go-to-market represents the percentage of a week that a seller is in front of a customer or prospect. And in the industry for a decade plus, that has been around 25%. And I think we’re seeing pretty strong evidence that that could go up to 75%. In fact, I think, like, people like Kyle Norton at Owner.com are claiming they’re able to achieve that now. By eliminating just, like, administrative work, preparation work, you know, the overhang that I think AI will do a fantastic job of doing. So I think 2026 will be a massive breakthrough of that, and of course, that will take many years to, you know, proliferate into, like. More conservative industries for tech adoption. Phase 2, I think, could be, the seller is replaced by an agent. We’re seeing a couple vendors making some progress with that. The buyer remains human. I think their most promise occurs in, like, PLG funnels, as an example, because, like, the bar is very low. Like, today’s PLG funnel is, like, usually a sequence of. landing pages and product pages, and to have, like, a, you know, a freeform agent there helping you through and answering questions, I feel like that’s an opportunity today. And I think it could easily start to help in more transactional, SMB-type funnels as well. I think, you know, Phase 3 would be the buyer becomes an AI agent. I think Carolu… formerly at Atlassian, I think you’ve worked with her a little bit, has been, doing a lot of thinking around that, and that makes sense. Like, if you’ve ever been part of a… a buying process at a large organization. These are, like, massive committees. There’s, like, 100-page-plus RFPs that go out. There’s no way that humans can orchestrate that better than an AI agent to, like, truly evaluate the needs of an organization and evaluate the vendors without the human bias we have around likability of a salesperson, or, like, taking out to a fancy dinner, or whatever. So, we could see that world. That does become pretty Star Trek-y. And then, I think the fourth phase Is, the functional boundaries of the organization blur, and it’s no longer around, like, how does go-to-market work, but how does an organization work? And whether something is in the product, or engineering, or marketing, or sales, or CS, or finance department becomes less clear. I think we’ve seen a couple vendors that, represent that. Ironically, I was just jumped off a webinar with Momentum, one of our investments that is given an example of that, where they, you know, the use case they’ve been making great progress on is, like, every product and engineering team would love a very concise, accurate summary of what the market is saying to their support people and salespeople, to make better decisions on roadmapping, and that is one of many cross-functional use cases that a vendor like momentum enables, and I think that’s an example of what we’re talking about here in Phase 4, is today’s organizational silos of. a sales team, a marketing team, a support team, an engineering team, a product team, a finance team, these are designed because of limitations of humans. We don’t see a lot of people who get a master’s in finance that go code and get a master’s in computer science and go run finance teams. these big organizational silos exist because of those limitations of our capabilities. But once AI starts to do some of these things, you can really start rethinking them, because organizational silos do harm and create inefficiencies. And so I think that’s… that’s where we end up. Again, it’s hard to, like, conceptualize when or what, but Phase 1 is becoming far more evident.
Julia Nimchinski:
Momentum is one of our Year-long partners, I believe.
Mark Roberge:
Nice.
Julia Nimchinski:
Yeah, great, great you mentioned them. Jonathan is joining later today. Oh, good, yeah. CR is Head of Agents. And, yeah, to your point, Mark, Momentum is this really bright example of an AI-native company rethinking the whole thing from ground up. I’m curious, why do you think we can’t skip from Phase 1, or can we, to Phase 4? And just start building agentic organizations.
Mark Roberge:
Yeah, I think with any adoption cycle like this, there… There are massive limitations in the, human adoption. and technical capabilities. And so it’s really the classic product-market fit, you know what I mean? Like, and that’s what kind of causes the Gartner hype cycle, is… You just, like, take the internet again as an analogy we can relate to. When it first came out, like, we were dialing up through AOL, and it took 2 minutes to even connect to the internet. We really thought the internet was just an online brochure of your company. We couldn’t conceptualize yet user-generated content, we couldn’t conceptualize the iPhone and FaceTime, we couldn’t conceptualize the sharing economy like Uber, we couldn’t conceptualize, like, infrastructure technology like Twilio or Snowflake. Right? So, like, those things are coming, and we can’t even predict what they are, and they’re going to, like. Dictate the future successful companies and products that we’re gonna grow to love. But that, that really is, the limitation. I mean, the other side of that is, like. as humans, we’re not really ready to adopt these things either. You know, when… the story that I think most famous from the prior era was when Jeff Bezos quit his quant hedge fund job in 1997, he didn’t quit it to become a book salesman. He quit it because he envisioned A future where people bought everything from food to furniture to clothing online. But we weren’t ready to do that as humans, we still had to touch things, and the technology and infrastructure wasn’t there either. So, he did what, most… a lot of entrepreneurs need to do right now, is design big, start small. He listed 50 products that could be purchased today, and from a variety of… of… evaluation lenses chose books, because they didn’t get damaged in delivery, there was no question around quality, there’s a million new SKUs plus every year, and it’s a pretty decent profit margin. And so, while he exploited the revenue, short-term revenue opportunity there, he was able to build a decentralized warehouse infrastructure in cheap places of the world, which didn’t exist before, as well as a delivery mechanism that got you a book from click to your door in 2 days. And so when the world was ready, and the technology was ready, to buy clothing and food and furniture, he could capture that opportunity faster than Walmart. And that’s really what we’re… another example of what we’re seeing here is just, there’s human resistance. That’s a reality as well.

Julia Nimchinski:
Definitely. You talk a lot about AI-native orgs. I’m curious what are you seeing in your portfolio, in the transitions from, I don’t know, like, the flattened organizations. How would an AI native company be structured?
Mark Roberge:
Yeah, In a way, A lot of our… the people at the top in today’s organizations, whether it’s the CFO, or the head of engineering, or the head of sales, they’re there because of a mastery of human management. A mastery of hiring great people, of mobilizing great people, of coaching great people, of incenting people. I think that gets de-emphasized quite a bit, and it’s replaced by a mastery of Strategy implemented into process and technology. And so I think we’ve talked a little bit about that, like, perhaps a way for us to conceptualize that today is in the RevOps world, or DevOps world, where their entire job is to You know, do that, is to, like, accelerate the capability of a function through the adoption of processed data and technology. And so I think, like, our future leaders will, look more like that. And they also will probably be far more cross-functional. I think in the last two decades, we saw a movement towards specialization, especially in go-to-market, with the invention of the SDR to cold call, and the CSM to deploy. And AI… the reason for that was just, like, again, limitations of humans, and as AI takes over more of that, I think we’ll move in the other direction. and we’ll start compressing these roles, we’ll start seeing more full-stack salespeople, and we’ll start seeing new roles that transcend different functions. So I think that gives us a glimpse of what they might look like.
Julia Nimchinski:
this note, I’m curious, you’re really a genius in hiring. I’m curious, how does that impact Your hiring strategy today. If you were… CRO and building an AI native sales org? Would it be a generalist? What would be your first hire, second hire, async? Stay in fact.
Mark Roberge:
Yeah, I think soon it will be more of a generalist. I think, depending on the role, like, in the SDR side right now, someone who’s, like, really comfortable with the tech and prompting and iterating is advantageous. We’re seeing a couple examples out there, I think it was Vercel. Where the CRO, who’s formerly from Stripe, was… he had some really great breakthroughs, there. So, I don’t think we’re quite there yet, but, like, certainly the openness to, to some of that, like, process automation and tinkering, and I think, like, the stage we’re in right now is very experimental. And I don’t remember, I think it was, like, Google. 10 years ago, that, like, had, like, 10% like, experiment time on purpose. They, like, really, like, carved out time in the week for people just to have their own ideas and tinker. And I think that’s needed today. And it doesn’t have to be everyone as you’re scaling up, but you need people like that. We’ve also had the rise of the GTM engineer. Which you’ve talked a lot about on the show. I think that’s a really cool role and meme. It’s becoming a little funky, like, when SalesOps and RevOps came out, where it’s, like, means a lot of things now, and it’s also just a, in a way, like. I mean, I think the purest people say what a GTM engineer does is it really, like, automates the process, you know, stuff that’s human tasks. I mean, that’s what RevOps has done for 20 years. You know, I remember the first the first team… when we first built the HubSpot sales team, we were running on Salesforce, it was the only cloud-based CRM, and it was something like 17 clicks just to log a call, send an email, and set a follow-up task. And we had to hire a consultant to make it 3. I mean, that was, like, streamlining tasks, and, like, it’s on a different level now, involving LLMs and Python, libraries and, like, really cool, like, modeling tools, but it’s still kind of the same thing. So that level of tinkering and experimentation is, like. needs to be part of the org. I don’t know if it has to be a part of every single sales hire, but it needs to be part of the org.
Julia Nimchinski:
Love it. Mark, if you map this moment to the hype cycle, worthy of Blaze 2026.
Mark Roberge:
I think we’re gonna have… A bust… It’s hard to know. But I think we’re gonna, like… the bubble will burst.
Julia Nimchinski:
But at the same time.
Mark Roberge:
the true AI-native companies that take over the tech sector, will be running. they might be seed or Series A funded. So I think the companies that, like. raised it over $100 million before having product-market fit will go the way of, like, Groupon and WeWork. And the companies that, like. haven’t quite raised at that level yet, and won’t be able to because of the bubble bursts, will build the real businesses. So, that would be my prediction. And I think we’ll start to see, like, real production efficiencies unfold in almost every department.
Julia Nimchinski:
You co-wrote a chapter in one of the books, in the book of Momentum that was also released recently. And, you talk a lot about the new revenue architecture. And could you please unpack the role of 40 becoming 100, and its impact on performance, valuation? And yeah, downstream.
Mark Roberge:
Yeah, I mean, that… I’ve been thinking a lot about what are gonna be the… what’s the biggest innovator’s dilemmas as we move from cloud to AI? That… and what that means, I think that the late Clay Christensen invented this concept and studied it over 100 years, from moving to mass production, moving to electricity, moving to… Computing, moving to the internet. And essentially, there are situations where Once you understand the new way, whether it’s a tech architecture, or a business model, or a pricing model, whatever, it’s actually easier to build a business from the start, from the ground up, to exploit it, than it is to pivot an existing business. in the internet… I mean, in the on-prem to internet era, there were 3 major innovators’ dilemmas. One was the actual rebuilding on the cloud as opposed to a floppy disk. Two was distributing through an inside sales team and marketing and generating product-generated leads. And then three was the pricing model, going from on-prem one time to subscription. That was easier for a ground-up startup to exploit. What are going to be the biggest innovators’ dilemmas as we move to this AI era? My hypothesis on the biggest one is not on the product offering, but on the way the organization runs, and the way the organization writes code, builds product, sells product. And counts the beans. And I just think it’s going to be easier once these new best practices unfold in 26 and 27 for a white… white space, startup to… to exploit those, and to go from Rule of 40… Rule of 40 to Rule of 100, to go from a rep that’s producing, you know, 300K a quarter to 2 million a quarter. To, you know, having engineers write a quarter’s worth of code in a week, and to have finance teams close the books in an hour, not a month. I think those things are coming, and I think it’s gonna be easier to start from things from scratch than to pivot a big company.
Julia Nimchinski:
Love it. Mark, speaking of innovator’s Dilemma, do you think we are currently experiencing a sustaining use of this disruptive technology, or… Does it have to be this way?
Mark Roberge:
What do you mean? Like, does it…
Julia Nimchinski:
Yeah, so, I don’t want to be nerdy, but…
Mark Roberge:
Yeah.
Julia Nimchinski:
In the transition, in the architecture of the disk drives, we went through from 14, I believe, to something like 2. And then there was a point of time when the disk drives were literally shipped with frames. Teaching. make them fit in the older architectures. Right. I’m curious if you see this as a parallel to what’s happening, you know, in the tech sector with AI SDRs and all of the Just enhanced use cases, sustaining use cases?
Mark Roberge:
Right, yeah.
Julia Nimchinski:
incrementally, like, disruptive technology.
Mark Roberge:
Yeah, I think we’re still in the incremental stage. I think a lot of people are taking the existing workflow and trying to automate it with AI, and that’s not the opportunity. I think the opportunity is to reinvent the workflow. There’s a lot of examples. I think probably the easiest one to conceptualize is the concept of a support ticket. You know, from a first principles, like, someone comes to your product and struggles, so they find the chat. and they submit a ticket, and there’s a whole bunch of software and teams that help you with that. But AI has the potential to identify the behavior that leads up to a ticket, and to help them on the spot within the product. So those are the types of things that we need to rethink is… just rethink the workflow, not iterate it. So… and I know, Julia, you gotta run, to the… to the next one, and I… you… you started off with the… yeah, today we’re… we’re actually announcing the Science of Scaling book, just so folks know, you can pre-order it now. All the proceeds go to mental health. It’s a work that I’ve been working on the last 10 years in this journey. The number one reason why I’ve seen companies fail unnecessarily is choosing the wrong time to scale and the wrong pacing. It’s largely driven by a fundraise. And it’s not about where you’re at. And that’s what the work does, is it allows you to calculate when to scale and how fast using your own data. whether you’re using that, doing that through humans or through agents, that’s what it’s about. So, I’m… if people do bulk orders, I’m happy to do, webinars or in-person events for them, that’s something that I’m offering. So, thanks for the platform as usual, Julia.
Julia Nimchinski:
Love it. Thank you so much, Mark. And last question. Your top AIGTM prediction for 2026.
Mark Roberge:
Yeah, I think the rethinking of enablement… I think enablement has fallen to the bottom of the categories for go-to-market, because it’s seen as right content at the right time. And I’ve seen early signs that enablement is about, you book a meeting. it sets up the whole process for you, stands up a buyer agent for you that the rep can practice with, attends the meeting, and coaches the rep along the way, and updates everything after the fact, including the account plan, as well as the rep skill scorecard. That’s a very cool reinvention of the workflow that’s coming to life right now. I… and there’s other folks that agree, I think enablement will shoot up to the top.
Julia Nimchinski:
Thank you so much.
Mark Roberge:
Thanks, Julia.