The Intercom AI Transformation

Live from Schenectady, New York, it's the Metrics Brothers, Growth and CAC.

And I'm Growth, better known as Ray Reich, founder and CEO of Benchmarket.

And I'm CAC, better known as Dave Kellogg, EIR at Balderton Capital, independent consultant. And the author of Kellogg.

And together, we are the Metrix Brothers.

And we go together like Cortés and one of his ships.

Ah, so no retreat in us, huh?

No retreat. Now, Ray, I gotta say, no retreat. And that, that's what we're here to talk about, which is burning the ships in SaaS. But I have to say that Cortés did not in fact burn his ships, Ray. Do you know what he did with them?

I do not. This was on his way to take over Mexico?

Uh, he scuttled the ships. He did not actually burn them. The mythology became burned, but he actually scuttled or sank, uh, the ship. So yeah, so to get full commitment from the crew, right? There's no going home. But it was not— ironically, burning the ships was not actually burning the ships. Uh, but we'll continue using the metaphor to And who are we going to talk about who burned the ships, Ray?

Well, that one is very appropriate for Intercom, now a $400 million AR company growing at 26%. And they did scuttle— is that the right word?— a lot of their SaaS resources.

Indeed, indeed. And in so doing, they killed off literally tens of millions in ARR, rebranded the company, launched a new product, did a pivot slash— I'll call it a pivot, which is the current language, to AI. So, this story, Ray, and the reason I wanted to cover this story in this episode is, first, it's extremely well told. He did a press tour on this story, and he's on Lenny's podcast. CJ did a post on outcome-based pricing. He did a couple of tweets himself. This is an extremely well-told story. It is an inspirational story about, hey, can you take a stuck multi-hundred million dollar SaaS company and remake it for AI? which he in fact did. Now, I'm going to call him he because I can't pronounce his name, which we'll come back to in a minute. But before we deal with the Irish spelling of Eoghan, I think we should hear a word from our sponsor.

The Metrics Brothers is proudly sponsored by Benchmarkit, the B2B AI and software industry's largest, broadest, and most contextualized benchmarking index. Dave Kellogg and Ray Groth Wright have put the Metrics Brothers podcast on hiatus for remainder of 2026. All episodes we have ever recorded are still available on your favorite podcasting app. You can reach Dave at Kellogg and Ray on LinkedIn at Ray Reich. Thank you so much for listening to the Metrics Brothers podcast. Okay, Dave, let's get into this story. So, you know, you see a lot of SaaS companies being challenged, and I use the word become AI-first. And this one seems like a really good, at least edge case example. So why don't you give us some details?

Yeah, I'm not even sure I'd call it an edge case. I mean, it's certainly not your typical company, and this is not a typical story, but it can be done. So I do find this inspirational, which is why I wanted to cover it. So first, you've got Intercom, you know, somehow today, $400 million in ARR, as you mentioned, somehow reaccelerating. Their new AI agent called Fin is headed towards $100 million in ARR on its own, growing at 3.5x. So superb growth. All of this is happening while the SaaSacre is happening out there. The SaaS market is getting hammered. But if you look back in time, just 2 years, Intercom then looked a lot— well, actually, it looked a lot worse than SaaS does today, right? Because they had had 5 straight quarters of declining net new ARR. Growth had basically stalled out. So, this is really a company-level reset.

Yeah. And in fact, can I put some more numbers behind that, Dave?

Please, Ray. Yeah.

Because typically, you know, we talk about growth endurance. And growth endurance is what percentage of the previous year growth do you keep in the following year? And it used to be around 80%. The last couple of years, we've seen it down into 65%. So at $100 million ARR, $200 million ARR, you're going to see that growth decay. But Intercom, they were growing 4%, 4% in Q3 of '23. And just to kind of put some headline numbers out there, they increased that growth rate to 11% in Q4 '24. And in Q4 '25, Dave, They went up to 26%. So unheard of, well, unheard of, very rare growth reacceleration starting from a low point of 4%, just a little bit over 2 years ago.

Now, some of you might think this is kind of a founder mode story because there is a founder mode element to this. The founder, Ian, had gotten sick. And as I understand it, he was bitten by a tick. So, I don't know if this was Lyme disease. They never mentioned it in the stories. But you had a founder get very sick, step aside for 2 years, have a hired CEO come in. And then, as the founder started to get better, he discovers that the business is effectively stuck and doesn't like it. So, he walks back in against this backdrop of low growth endurance, flattening ARR, actually declining ARR. And says, we need to do something. And it wasn't like he did it all by himself. But it definitely was founder mode to the extent that a founder came back and said, we need to change for 2 reasons. One, we're kind of stuck. We had worked very hard. This is like a 10, 13-year-old company. We had worked very hard for a very long time to create something of value. And it's stuck. It's not worth very much anymore. And it doesn't seem to be going anywhere. So, we need to do something. And in some ways, he credits correctly so that the fact they were so stuck made it easier to drive change, right? It's actually hard to do this if you're growing 10%, right? And if things look okay. But they were really like, they were in trouble.

Yeah, I will tell you, he also had one external factor that was a little bit different in this environment. Now, he still needed to act upon that because he came back in October '22. Dave, and ChatGPT launched in November of '22. So he brought the key executive team together. They had a 72-hour kind of strategic kind of session that they got together offsite. And after that 72 hours, they decided to go all in on AI as the core part of their product going forward.

Correct. A very important part of the element. AI is very much part of this makeover story. In some ways, if AI is the problem for most SaaS companies, these guys turned AI into the opportunity. It's like, hey, we're stuck. We're not going anywhere. We're not even sure we are who we wanted to be. And I think as he's looking at this backdrop, ChatGPT, you could just imagine looking at ChatGPT and going like, eureka, we need to go all in. And there was really no reason not to go all in because they were fundamentally stuck.

Yeah.

So for them going all in meant writing off like $50 to $60 million in ARR. I think they rotated like 80% of the R&D team into it. So this wasn't, hey, throw a couple of developers on AI, which I think is the classic conservative SaaS CEO response. This was, I would call it a more Benioff-like hard rotate. Like, okay, we'll leave a few guys on the core product to fix a couple of bugs. But everybody else, we're going to go all in and build this new AI product.

Hey, hey, Dave, you just mentioned something. You said like writing off some traditional current ARR. You've sat on a lot of boards. You've been a CEO multiple times with investors on your board. How does that conversation with the board go? Because to me, they often are the, we can't do that to our existing customer base and ARR.

Yeah, I think it depends on the kind of board you have first. Ironically, PE boards might be more open to it than VC boards because PE boards, all you need to do is convince really one person, the deal partner, that if we suck up some pain now, we're going to have a happier ending later, right? So, it's a difficult sale, but there's only one person you have to sell on it. What makes things tricky in VC is while each of them individually might be more aggressive, and more kind of forward-leaning, hey, yeah, we need to go build for the AI future, in order to say, hey, we're going to let this part of our business shrink and we're going to expect to have very high churn, i.e., very low GRR in certain segments of the business. And we're effectively not going to care about that because we're betting everything on the new future. To do that, it's difficult because you might have 5 or 6 Type A personality VCs who might go, yeah, that's an interesting idea, but can't we do it without writing off that ARR? Can't we find this balanced middle? And one of the reasons I think founders like this story so much, because a lot of founders I talk to love this story. So, if you are a founder, go listen to Ewen on Lenny's podcast because this guy's very real. He's not slick. He's not fancy. He's very real and down to earth. And he's not dramatizing when he says we needed to basically turn this thing. And it wasn't going to be through minor moves. And I think he felt like a lot of the wisdom out there he got from his advisors and his VCs, he felt was bad. And it would have been, I'm guessing, this much more, can't we take a measured approach? Can't we bet a few people on it? That's what committees do, you know?

Well, one of the things I've seen when the market's moving away from us, and I saw this a couple of times at like $30 million to $50 million, I also saw this CEO kind of hedge their message. That because a lot of the employees were skeptical of making wholesale changes of something that, hey, it got us to $30 million. It got us to market leadership. So, they hedged their message a little bit. And they don't make some of the larger, I will say, wholesale organizational changes needed to really go after the next phase of the market. Do you see that sometimes where CEOs are worried about how employees will accept Large wholesale changes?

Yeah, look, in general, change is difficult for anyone. And whatever it's worth, Ray, I do believe the hardest thing to change is a good performing business that you think is doomed. I had that challenge at MarkLogic. We were performing very, very well. And it was like, we need to change this because no SQL is coming. That's the hardest thing. I think the second hardest thing is kind of a stuck business because here, It wasn't like they were bleeding out. I can't remember how much they were shrinking. It was single-digit percent, I think, right? They certainly weren't growing. Value is getting destroyed. But look, sometimes it's the CEO who wants to hedge. Sometimes it's the board. Sometimes it's the committee. Usually it's the committee nature of the board that leads to a more conservative outcome. But yeah, it happens all the time.

Yeah, by the way, I think, and this is from Owen's kind of post on X, They were growing about 24% in the second half of '22, and that went all the way down to 4% in the second half of '23. So, you're right. There was a factor that forced them to change. But a couple other things they did, Dave, and you mentioned, they introduced these new work streams. They were small startup-like teams of 10 to 15 people, and it was drawn from engineering, machine learning, sales, and marketing. And there was a head, and they called that head the directly responsible individual, DRI. And he or she really led that team like a true startup. They were 100% segregated from the people keeping the core Intercom business running. And in fact, that's when they created their Apex team. And their Apex team started with about 10 to 12 AI machine learning people, specifically Tasked with getting the third-party models actually brought into the company and optimized. And they just announced this last month, Dave, and not to kind of get ahead of ourselves, but they have now built their own, they call a vertical customer agent model. And it has about 3 to 4% higher resolution rates than what the third-party LLMs from Claude and OpenAI were delivering. So they're all in, even on model development now.

Awesome. Awesome. So let's talk a little bit about the economics of all this, Ray. Look, the core argument is simple. You're going to trade— I can't remember who had this quote, darn— you're trading a structurally great business for a harder business, right? SaaS works at 75 to 90% gross margins. AI is more like 50 to 70%. So this is, you know, if you think about it from a finance point of view, it's like, Hey, let's jump out of this profitable boat into this less profitable boat, right? Let's actively pursue a lower margin business. And it does mean it's going to change your economic model. It means for the customer that software looks a lot more like labor, which increases your TAM. For you, it obviously is going to increase your COGS dramatically, which will affect the rest of your P&L. And for a given dollar of input, you don't get as much output. So, so, so what are you seeing here, Ray, when it comes to the numbers?

Yeah, well, if you look at pure AI native that are heavily leveraging third-party LLM API calls, I'm seeing it in the 40 to 55% range is gross margin. Now, if you're a blended SaaS company, you're still doing a lot on your legacy cloud service provider and you only have some AI, I've seen that blended range in the 55 up to 70% range. But a couple of things have happened. Number one, we all expected token costs to come down, and that has. But the intensity of the prompts has gone up dramatically. So we're actually seeing companies right now spending more on inference than they were a year ago, even though per-token cost is down. And Intercom doesn't publish their gross margin, or I couldn't find it, Dave. So But I would assume that one of the reasons they built their own model wasn't just for performance, but it was for cost of goods sold optimization also.

Yeah, I mean, Ray, a couple thoughts on that at a high level. One, you know, these people want to get paid back for their hundreds of billions in CapEx, right? So as soon as the market share— as soon as kind of the greenfield market share wars are over, they're going to stop subsidizing tokens, And tokens are going to get more expensive, not less. And I think a forward-thinking company is like, hey, where can I reduce my dependency on foundation model inferencing? And in most cases, actually get a better result. So I hear you there. I think this also— well, it wasn't just their cost structure that changed, it was their pricing model as well. So unless you want to say more on cost, I was going to flip to price.

No, let's go to price because it was a key inflection point, or appears to be.

Yeah. So look, as part of this, and I do think that some spaces lend themselves better than others to outcome-based pricing. Like, you know, back in the day, like Twilio, per SMS message sent, right? That lent itself well to consumption-based pricing, right? Some other things don't lend them, like post and financial planning. It didn't lend it, you know, per plan produced, per model run. There was no natural per, as I call it, use for the outcome-based pricing. I don't know if they moved to it or not later, but it was not obvious. But this customer support is one of the, in my opinion, most obvious places and easiest, like per ticket resolved, right? There's just a natural per. So, as part of this transformation, they basically said customer support is changing. We need to change our software with it. We need to build Fin. And Not only is Fin going to have a different cost structure that we need to figure out in our P&L, Fin's going to have a different price for the customer, pricing model for the customer. And we're going to do that per resolved ticket.

Yeah.

And, you know, just like we're going to change one thing, just like, Ray, when we moved to SaaS, people are like, oh, it's just charging per month instead of charging per year, right? You could look at it with one eye and go, yeah, you're right. Instead of one big check, you know, one big check at the start of a 5-year, 10-year project, you'd have one check smaller check every year. It's just a pricing model change. But to me, if you got it, it was a lot more than that because that first check wasn't actually usually big enough to pay back your customer acquisition costs, let alone generate profit. So, it was like, if you want to make money on this customer, you need to be vested in their success. And that was the massive difference. It wasn't just like a, oh, a little changing pricing model. And the same thing is true with outcomes, right? As you move from subscription to outcome, it's not a little change. It's not like, oh, your bill's roughly the same. And we're just going to get paid slightly differently. And we're arrears at the end of the month or in advance for a certain number of tickets. It's that we only get paid when you get a ticket resolved. And that's a huge philosophical difference.

It is. But they also had, I won't call it an unfair advantage, but they did have a couple advantages. Number one, they already had a large install base of customers that they were providing customer support to. So, they deployed Fin internally. And they found out that they were getting about an 81% internal resolution rate with their customers. And it was engaging over time over 97% of the inbound customer service calls. So, that's one kind of data point they had. They also are today, they know they have about a 72% customer resolution rate. So, at that $0.99 per resolution, that they're pricing at with a 72% resolution rate across now 8,000 customers, it's a pretty easy model to say, hey, this was my cost using human resources. This is my cost using AI agents. So they kind of knew what the profitability was going to be at $0.99. Does that make sense?

It does, Ray. I mean, certainly that was kind of a data advantage that they had built into the story. So, if we come back, I think we want to start to wrap up here, Ray. Well, do we want to talk about this, the comfort trap? Do we want to talk about that now, though, or Ray?

Well, you know me, I love numbers. The one other thing I have to say is NRR, net revenue retention. After they deployed this $0.99 per resolution, they were at 112% NRR. Back then, last quarter, and this was— Owen actually shared this— 146% NRR. So you got growth rate going up dramatically. You've got NRR growing from 112% to 146%. This use case is really worth looking at because all the standard financial performance metrics are off the charts positive.

Yeah, no, it's an awesome story. I want to come back to— we touched on this in the beginning. We called it the comfort trap in the script, Ray. But it's this notion. Basically, it summarizes to a nice little quote, which is never waste a good crisis, right? And I think that's what these guys did. And it would beg the question, if your company's not in a crisis, could you do this kind of transformation? And the answer is it might be a lot more difficult. But I think Owen and his team kind of recognized the crisis. And didn't waste it, right? They said, you know what? We need to make hard changes and our back is against the wall. So, I think one of his quotes was literally like, there's almost nothing to lose anyway, right? So, the Cortez thing is appropriate, right? There's no going back. We're against the wall. There's nothing we can lose. So, we may as well kind of reinvent this thing from scratch, right? And I think it was that courage that enabled a lot of this story. And I worry For people who aren't in such crises or crises that they're going to kind of land on half solutions that get them nowhere. So, I think there's 2 takeaways from the tale. One is never waste a good crisis. And then 2, if you're not in a crisis, if you're just kind of muddling along, maybe you better start to think of it as a crisis anyway, right? Because if you don't think your company is going to have a real market in 3 years, it is a crisis, right? It's just a slow-moving one.

Yeah.

And then take appropriate dramatic reinvention steps.

Yeah. And I think with him, with his illness he went through, coming back in, AI just getting launched by the ChatGPT, I think it was almost a perfect storm where he said, I don't have anything to lose. But I honestly think a lot of the CEOs I talk to right now and their SaaS companies that are down to 8% to 12% growth rate, but they're generating 20% to 30% EBITDA. I actually think a lot of them feel like they're in crisis because it's just not as much fun. So, I would say if you think there's a breakout play, maybe you just look in the mirror and say, I am in crisis because I'm not having fun. With the multiple compression going on, the M&A market's not going to be nearly as lucrative. I don't think the VC investors are going to be happy. I think most companies who have seen those dynamics Are in a crisis. They just need to admit it.

Yeah. And let me wrap this up, Ray. I think a nice bow on the episode is more and more people are blogging, tweeting that there's effectively 2 paths out there for SaaS companies. One is dramatic AI reinvention, and this could be your role model tale. And that's why we told it today. The Intercom AI transformation has been truly epic, and it was done to a good-sized company. So, it wasn't 20 guys who had to pivot. It was a couple hundred. And the other path is basically rule of 70, right? But that's where the road is going, which is either you reinvent for AI and you go through a transformation where you come out on the back end with a native AI solution, outcome-based pricing, and you look like Intercom, or you go the other path and say, okay, we're just going to be a money-making machine and we're going to shoot for like rule of 60.

Right.

Or I think I told you this the other day. I had a friend who runs a PE-backed company and the founder of the PE firm was saying, rule of 70, rule of 60 is not good enough.

Yeah. And I would wrap up with, I love another thing that Intercom is doing. They're basically saying their Fenn agent has eliminated the need for almost 7,000 customer service agents across their customer base. So, they're putting some real great messaging behind Hey, this is real cost takeout. And the other thing, I just gotta add this, they just announced their customer agent. Notice I didn't say customer service. So they're going from inbound customer support calls to outbound customer engagement and even upselling. So it's really interesting how they're continuing to increase their target addressable market with this new product.

Awesome. Well, thanks for joining me today, Ray.

It was fun. Thanks, Dave.

The Metrics Brothers AI, SaaS, and More is a production of The Metrics Brothers Growth and CAC and a member of the Benchmarkit Podcast Network. By accessing this podcast, you acknowledge that The Metrics Brothers make no warranty, guarantee, or representation as to the accuracy or sufficiency of the information provided or the humor content of the jokes told. Ahem.

Ray?

Information, opinions, and recommendations presented are, according to our spouses, probably wrong and provided for information purposes Only. This podcast should not be considered professional or, for that matter, unprofessional advice. We disclaim any and all liability for any direct, indirect, indirect, misdirect, incidental, special, ordinary, consequential, inconsequential, or other damages arising out of any use of, or God help you, reliance upon the information presented here. Ray Grothreich is based in New York City and available on X @RayReich. Dave Kellogg is based in Silicon Valley and available on X Kellogg. Schedectity, which is French for unspellable, is not our actual production location. Thanks for listening.