Intercom’s AI Turnaround + Why Software Companies Must Train Their Own Models | Eoghan McCabe

The month after I came back, chat GPT launched and so the whole world changed really quickly and Dez, my co-founder, came to me a weekend or two after and said, >> "Hey, the AI group, they think that

there's something interesting that can be built here." We realized we could build an AI thing that would do service instead of humans. And the more we pulled that thread, we're like, "Holy

[ __ ] this is going to disrupt entire business."

entire business." >> Exactly. Everyone who sells seats to

>> Exactly. Everyone who sells seats to humans that do service are in big trouble.

>> Welcome to the peel. I'm your host Turner Novak, founder of Banana Capital.

Today's guest is on McCabe, co-founder and CEO of Intercom and Finn, the AI customer service company.

>> I left the CEO role in the summer of 2020. Frankly, I thought the company was

2020. Frankly, I thought the company was about to be sold. This was an extremely candid 2-hour conversation on how Intercom was one of the first late stage software companies to successfully

rearchitect itself to be completely AI native.

>> We had five quarters of sequentially declining net new AOR.

>> They also just announced they've built their own customer service focused AI models and own shares why most software companies will have to do the same.

>> When you own the full stack, it allows you to be a lot more nimble. Apple own

the hardware, the operating system. They

even own the chips. It just allows them to be a lot more fluid.

>> We also talk about the future of software. What he learned coming back to

software. What he learned coming back to run intercom a second time.

>> And you'll never have worse regrets as a founder than if you make mistakes and you went against your intuition.

>> Why many AI companies don't actually have negative gross margins despite the popular narrative. How Intercom and Finn

popular narrative. How Intercom and Finn built the first outcome based pricing model for AI.

>> Messed our pricing up. so badly in the past.

>> Why it's so much harder for customers to buy AI software.

>> You used to be able to sign in to the new app, look around and say it works, it looks great. You can't do that with AI.

>> The importance of branding new AI products.

>> If you heard that Cisco are making the most interesting AI coding platform, you would say >> I would say there's zero chance.

>> It's probably not true. And we get in the wayback machine talking about the pain raising intercom's initial million-dollar seat round and how venture capital has changed.

>> It was a Skype call. It was 26-year-old me, 12 guys in suits.

>> A quick thank you to Owen's co-founder, Dez Trainer, for his help on this. And a

reminder, I publish new episodes of the Peel every week exploring the world's greatest startup stories just like this one. Tune in next week for a

one. Tune in next week for a conversation with Dan Federer, who runs private market investing at the University of Michigan's endowment.

Let's talk to own after a quick word from RAM. If you're running a finance

from RAM. If you're running a finance team, you know how much time gets wasted on expense management. Chasing receipts,

categorizing transactions, waiting for expense reports. It adds up quickly.

expense reports. It adds up quickly.

RAMP handles all this automatically.

RAMP is a corporate card and expense management platform that over 40,000 companies like Shopify, CBRE, and Stripe are using to streamline their financial operations. But here's what makes their

operations. But here's what makes their corporate card different. Every

transaction gets automatically categorized to match your receipts. No

more wondering what that $47 charge was 3 weeks later. You can set spending controls, get real-time alerts, and even block certain merchant categories. That

sounds pretty cool. It's like having a finance team member embedded in every purchase. The platform integrates with

purchase. The platform integrates with your accounting system in ERP, so everything flows through without manual data entry. Whether you're issuing cards

data entry. Whether you're issuing cards to a few employees or managing spend across departments, RAMP gives you visibility and control without the paperwork. Stop chasing receipts. Check

paperwork. Stop chasing receipts. Check

out ramp.com/theel.

Get $250 and see what a corporate card can actually do for you. Time is money.

Save both with RAMP.

>> Owen, welcome to the show.

>> Thank you.

>> Did I get the Did I get it right? Okay,

perfect. We were just talking before we started recording that I was trying to get the pronunciation right.

>> Good job.

>> Um, but I'm excited for this. You are

the founder and CEO of Intercom. You're

probably one of the only >> later stage larger software companies that has >> basically gone back and completely reinvented yourself to be kind of AI first. I'm just kind of excited to kind

first. I'm just kind of excited to kind of spend the next hour and a halfish just talking about it. So going back, I don't know when you usually like to start that journey, but when like kind

of the past couple years, how things going?

>> Yeah. So

there's a bunch of stuff I want to give context on. One is just the general

context on. One is just the general market at the moment. I mentioned to you earlier like latest stage software is actually in trouble. um you know look up your favorite uh large SAS business

that's public >> and you will see uh consistently declining revenue growth.

>> So there are some that are declining >> well declining net new error >> and then in the last quarter a number of big names went into negative net new error territory. So the error is now

error territory. So the error is now decreasing. So just to kind of set the

decreasing. So just to kind of set the stage that like latestage software the kind of win is just not there anymore that was propelling it.

>> Do you know what's causing that?

>> Um I actually don't. Um

uh I wish I didn't bring up this topic because now I look like a fool but I >> you know that it's happening. It's it's

it's happening and some it's somehow connected with like somehow SAS and software has gone

stale and the the midw explanation would be oh AI something something AI

>> but I don't actually see that but there is something common amongst these companies that are not doing so well which is that they are I it's hard to see the relevance in the age of AI.

>> Um, you know, a lot of them don't have a story for how they are relevant in the future. It's not yet the case that vibe

future. It's not yet the case that vibe coding has eclipsed the value of traditional software development such that all of these SAS tools are just no

longer valuable because they're commod commoditized. Commodified

commoditized. Commodified >> commoditized I think that's the word maybe. Well,

maybe. Well, >> let's just keep both of them in for everyone. Yeah. Um

everyone. Yeah. Um

I don't think that's true. Um because

vibe vibe coding is not quite there yet.

It's just vibe still.

>> Yeah. I think many people kind of made the observation like have you seen any actual production vibe coded apps that are you know more than just a screen or two >> but it's adding efficiencies etc. But so

I don't think it's that but um >> there's just a lot of loss of energy. I

mean maybe it's that everyone has all the software now like Certainly all these software categories, they're not new anymore. You know, AI came around at the same time that these

categories all became super mature.

>> If you look in any software category, think of like an ASA like I don't follow them closely, but I'd be in a tough spot if you ask me how how should they innovate in a major

way. Like what's new there? Like these

way. Like what's new there? Like these

companies, even intercom's been around 14 years. I think's been around 15 16

14 years. I think's been around 15 16 years.

to like I I just feel like probably the juice has been squeezed out of these categories. That's what I think

categories. That's what I think >> like everyone has all the features.

Everyone is like kind of >> I think that's kind of it. Yeah, I think that's kind of it. Um and certainly we

came out of a period postco, you know, insane fake growth. So there's

a rebalancing in some ways and a lot of companies have been shedding the fake users and revenue that they won in 20 late 20 well 2021 and a little bit of

22.

>> So it copied that a bit of a combination of the categories are all super mature.

There's not a lot of opportunity for innovation anymore and these companies have been slowly shedding the crappy revenue they won in the postcoid sugar rush. Maybe that's it. I'm sure some

rush. Maybe that's it. I'm sure some smart analysts have a better narrative, but the the reality is it's happening >> and it really feels like latestage software is kind of in trouble unless it

can find >> a an AI story to help it be relevant in whatever comes next. And is it because people are are adopting and using AI in in a way that's kind of net new compared

to just a new >> No, it's more what I'm saying is that for whatever reason I've attempted to give two reasons and maybe both of them

are not true, but for whatever reason um latestage software is slowing down and I don't see it and it's not about I don't I'm almost certain it's not about just accelerate just building more software.

M >> what's going to be required is that the people playing in those categories need to figure out how do they get disrupted by AI and build that >> themselves.

>> Correct. And that takes us to our story which is what you asked me about. I was

just trying to set the stage for what's happening more broadly. So

>> was there was there a moment when you realized this or >> Yeah, but it's a combination of things.

So you know the story kind of starts almost in 2019.

I got sick in 2018. you know, much later found out it was it started when I got a tick bite in the summer of in the start of June 2018.

And so I was super sick. Some days I mean couldn't get out of bed. Some days

couldn't like really see. It was like wild. And then

wild. And then revenue started to slow cuz we made some pricing mistakes chiefly because I was sick frankly. And then I was like super

sick frankly. And then I was like super unfairly attacked in the press and in my sick and deflated state. didn't know how

to defend myself. And so by the time early 2020 rolled around, I was just done with being a software CEO and

really down on myself and and was ready for a change. So I left the CEO role in the summer of 2020 and I promoted someone internally to be

CEO.

Um, frankly, I thought the company was about to be sold. Um

>> there was two big companies that approached us at the time. So I thought I played it excellently where I left and wouldn't had to work in that.

>> That's actually pretty smooth. Yeah,

that >> didn't work quite work out. Then I

thought we were going to go public. We

had a date in 2022 2022. Is that right?

2022. Is that right?

>> Oh wow. Do you remember a rough date?

>> Yeah, it I I want to say it was something like the 15th of July 22.

Yeah, it was the summer 22 >> but at that stage you know the entire of the technology market had technical term [ __ ] the bed.

>> Yeah and technical term >> Manny had dropped like 90%. No many

great names had dropped 80% 85%. So IPO

the IPO market was closed and frankly just really reopened.

Um, and so the com I was just unhappy with how the company was performing for some couple of years. Did we didn't agree with a bunch of different decisions, but

I was on the outside at that point. And

my critique didn't was didn't find any purchase with the board until the company started to perform so poorly.

>> Did you start to go through some of that like mature crusting?

>> Yeah, but in a really dramatic and violent way.

>> Oh, really? Like some of these companies that are finally now seeing negative net new A and negative A growth, they evidently took a lot longer to come down from the 2021 highs.

>> Yeah.

>> But we really fell off a cliff. And it

was because we just had a really dilute strategy. Didn't know who we were. And

strategy. Didn't know who we were. And

the revenue growth we enjoyed in 21 was really a lot of cheap expansion that we kind of beg, borrowed, and steal. we

were very aggressive with our customers >> and um and kind of upsold them on the things they were already buying rather than paying careful attention to what they

really needed in the future. And so I feel like the combination of everyone pulling back from all software but particularly

um uh that software which they didn't absolutely need um contributed to our precipitous drop. So we had five

precipitous drop. So we had five quarters of sequentially declining net new AR. I I came in at the end of the

new AR. I I came in at the end of the fourth quarter and >> this is 22.

>> This is 22 late 22. And so the company was in a bad spot like a weak dilute strategy. Didn't know

who it was anymore. Even the culture was weak. Um revenue falling off a cliff.

weak. Um revenue falling off a cliff.

And I I went back with a lot of personal energy to kind of you know write the wrongs I um had seen over the years but also with

the moral authority both as founder and as the new CEO who came in when the company was in a trouble spot to make a lot of changes.

>> And so part of the secret of our success and the dramatic pivot we made is that we had to make a big change. Um, and you know, I don't want to I don't don't want

to discount the value of the decisions I did make. It still takes bravery to to

did make. It still takes bravery to to bet the company on wildly new things and turn off revenue and make all the wild changes I made. But I do want to

emphasize that we didn't have a choice.

>> Whereas many other companies, >> big names who like I keep saying now are experiencing uh negative pay or growth, >> they were doing okay. they were doing

even o good good good enough and so they had more to lose. So, so it was those those two things and then the month after I came back, chat GPT launched and so the whole world changed really

quickly and Dez my co-founder came to me some like a weekend or two after chat GPT launched and said hey the AI group you already had an AI group in the company says that they think that there's something interesting that can

be built here and we didn't even use the word agent at the time agents have got cool in the last 6 n months but we didn't say agent >> but we realized we could build an AI thing that would do service instead of

humans. And the more we pull that

humans. And the more we pull that thread, we're like, "Holy [ __ ] this is going to disrupt.

>> Entire business."

>> Exactly. Everyone who sells seats to humans that do service >> y >> are in big trouble. And so

we both needed a new vehicle for growth in the business, but also saw significant threat. And that

together was an opportunity for us to jump on AI and kind of you know that was a short number of years before we end up now where you know we're in the service AI business where you know quickly we'll

be in the customer AI business we'll do more than service but amongst say service AI agents com whether you look

at the incumbents like Zenesk or the um startups like DecaGon were the largest by revenue um largest by customer account number one in the benchmarks

performance per performance benchmarks.

We win 100% of bake offs against our direct competitors. We're number one in

direct competitors. We're number one in G2. So, I just want to emphasize that

G2. So, I just want to emphasize that the pivot was not only dramatic, but to our relief and surprise, like insanely

successful. Um, and so we're still in

successful. Um, and so we're still in the process of processing that because it kind of crept up on us even in the

last 6 months. how well it worked out.

>> And so you win 100% of the deals that you go up against >> with any of our direct competitors.

These are bake offs where it's us versus say two or three other CSAI agents and the competitor the customer will either

run them all headto head do like an AB test or have a batch of questions and run through them and Finn consistently beats the others in terms of the

percentage of the customer queries it can successfully resolve. Um,

yeah, and of course we got to keep that up. And so that's part of the adventure

up. And so that's part of the adventure of the high.

>> Yeah, I mean it just doesn't stop. And

so we could talk about that separately if you're interested. But um, the whole the whole world of software development is very much changed the pace and intensity.

>> And one thing that you said when we were talking beforehand, you mentioned that basically AI is going to be better than humans at pretty much everything, >> right?

>> It's kind of one of those things. It

sounds like a very loaded take, but also like a actually like very down the fairway like no duh. Like software is always better than humans, right?

>> Um like is there a certain timeline you kind of think through of like how that will evolve in certain areas? maybe

customer service you're >> most attuned to. But

>> yeah, it like it's definitely happening today in a very real way and it's just going to continue bit by bit and then one day we'll wake up and we'll realize

that all of the hardest work is actually being done by AI. It's going to be extremely gradual. so gradual such that

extremely gradual. so gradual such that like I said very real economically valuable work is being done today by AI.

I'll talk about customer service in a second but look at Whimo. It's real.

It's all AI and it's better than humans.

It crashes substantially less than humans.

Um the drivers are infinitely less. Um

>> they're perfect. They're whatever you need from them.

>> Yeah. Imperfect. Let me put it that way.

>> Yeah.

>> Um for women who don't want to be in the car in a car with a stranger, who for people who just want quiet time, who privacy, who they want to make a

business call or they want to sit with their loved one, who I mean, it's really crazy. So whether it's driving safety,

crazy. So whether it's driving safety, personal safety, privacy, peace, consistency, I mean it's just better.

There are some ways in which it's inferior, kind of slow, etc., but these things will get solved. But

>> um it's better. And and so customer service, for example, on average, it takes a human service rep 9.8 minutes

to close a ticket.

And on average they leave um customers waiting 102.8 minutes before they start to respond.

>> So it takes them 102.8 to to respond and then within 9.8 minutes they close. So

it takes almost two hours. Correct.

Correct.

>> 110 roughly correct to get through it.

>> F our AI AI agent is essentially instant.

>> You know it's not instant because of some back and forth. You're going to chat with it.

>> Yep.

>> Clarify some questions. Um, but it's basically instant. AI is better than

basically instant. AI is better than humans. Or you take Finn answering a

humans. Or you take Finn answering a really high value question about pricing.

It's always going to give the same answer. Or Finn needs to look up some

answer. Or Finn needs to look up some highly sensitive data.

Um, it's always going to follow the procedures. When you ask humans to do

procedures. When you ask humans to do repetitive work that they hate, they make mistakes. I mean, all humans make

make mistakes. I mean, all humans make mistakes regardless, but when you ask humans to do really inhumane and inhuman work, they're going to [ __ ] up many times. AI is just better than

humans.

>> Not in all the ways today, but when you can apply it to work like this, it just ends up being better. And that's one of the really interesting things about AI.

Like, we think about it as this efficiency thing, like, oh [ __ ] all humans are going to get replaced because it's cheaper. It's not just that. It's

it's cheaper. It's not just that. It's

better and it's faster. So, we we joke and we had a a short-lived ad campaign that said better, faster, cheaper, pick three. Um,

three. Um, >> usually you can only pick two, right?

>> That's right.

>> Yeah.

>> And and that is how deeply wildly transformative AI is.

>> And it's one of those interesting technologies where cost curve coming down. I think if you're using the most

down. I think if you're using the most recent models, maybe that's like not quite true, but generally trend trending downward, but then also it's getting better over time. Like I think there's this classic with like, you know,

internet connectivity, you know, it just it doesn't really get better or Yeah.

>> Oh, you're saying internet connectivity doesn't really get faster. Yeah.

>> Or like with with cloud, it's like with the cloud, it's like the cost came down, but you didn't like get better cloud. It

was just like you're in the cloud and it's there. But with AI, it's getting

it's there. But with AI, it's getting cheaper and better at the same time. I

think it's having multiple plateau moments.

>> So, I think it plateaued for a little bit and maybe it's going to like accelerate again, but yeah, the the >> we all imagine that certainly it's about to get significantly better.

>> I don't try and predict that, but like probably it will, but yeah, I think you're I think you're probably right.

>> How have you guys approached that then with building out the product? Like how

do you do the trade-off of oh, it's not quite here, but maybe in 12 months it will be because that's kind of in the trend. How do you think that when you're

trend. How do you think that when you're when you're building stuff?

>> Like for us, we like say with OpenAI with their chat technology or image generation,

people are delighted for it to do the best it can because the best it can is pretty great and it didn't exist before even though it's imperfect.

with Finn with a agent that you're gonna have to talk to your customer like best effort's actually not good

enough. You know, it's not okay for you

enough. You know, it's not okay for you to ask as an end user, an agent for help with

resetting your password and it pukes up your credit card number in plain text >> or it refunds or it deletes your password. You know, I'm I'm making stuff

password. You know, I'm I'm making stuff up here that's kind of facitious, but it's not okay for a best effort estimate today.

And I promise I'm not using this as an opportunity to pump and promote Finn, but I tweeted that uh we are now doing a million resolutions a week, which

equates to 6,533 humans.

And I asked chat GPT to visualize that so I could have a nice little graphic for my social media post and I was like no problem. And I had it said it put on

no problem. And I had it said it put on the top 6533 and then a bunch of human icons and I'm like and I counted and it's like 400.

>> I was going to ask you about that. Yeah,

I saw the tweet. I was like this doesn't look like 6,000 people.

>> That's what I my my joke was too dry. My

joke was too dry. I said here's its attempt. Um, and that's that's okay. We

attempt. Um, and that's that's okay. We

laugh at it. But if it if it does that with your customers, it's not funny. So,

>> yeah, >> we have built a very complex system that uses many models to um only get involved when it is highly

confident it knows the answer and can help. Um and as such we're not on the

help. Um and as such we're not on the bleeding edge of any one model whereby when that one model gets dramatically better we get dramatically better. And

this is the misunderstanding between these cutting edge AI application companies. I think it's certainly the

companies. I think it's certainly the case for us that they indeed are not GPT rappers. Sounds a little defensive when

rappers. Sounds a little defensive when I go and say that, but >> you know, it's I think close to a maybe a dozen models of many different varieties from different vendors and in

different with different configurations just to make it highly competent at serving customers.

>> Well, yeah. You're not like a GPT rapper. You're like a GPT Optimus Prime

rapper. You're like a GPT Optimus Prime or whatever. Like it's like multiple all

or whatever. Like it's like multiple all woven.

Yeah. But there's many other types of models and they're not just, you know, Eld LLM. There's a lot going on there

Eld LLM. There's a lot going on there including some of our own uh LLM technology or machine learning technology.

>> So, so you've experimented with some of your own.

>> Yeah. So, we're working on that and you know by the time this is out we will most likely have announced that um we have made significant advances with our

own models. Um they'll be the world's

own models. Um they'll be the world's first CX models, proprietary CX models, um that do the type of work they do. Um

trained on proprietary CX data and we're quite confident that they're not only get going to increase the performance of Finn yet again, help us continue to beat

our competitors in those bake offs, but also allow us to own the full stack. You

know, when you own the full stack, it allows you to be a lot more nimble.

Like Apple own the hardware, the operating system, the iMessage app, the iMessage protocol. Um, they even own

iMessage protocol. Um, they even own like the custom emoji things.

>> As bad as they are.

>> Um, they own this whole thing and it just allows them to be a lot more fluid.

They even own the chips that these things run on. Um, yeah. And so when you're when you're not held back by third party vendors, um you can move faster and get ahead of the game. And so

we're excited about being able to do that too >> because that's kind of in the current um I don't know like meta around all the everything AI is that all the values acrewing to like certain layer. So if

you don't have something in that layer, you miss out on capturing a lot of the value.

>> Well, I think that if you rely on the generic models, you are subscribing to eventually becoming generic. Um, and so

I think that unless you have your own innovation and your own technology, you can't get beyond that. And like you said, indeed the value will go to layers

in the stack where the innovation is and we'll get to a point where it's not the base level models anymore.

>> So do you think so I guess first off any software company you need to reimagine yourself for AI. Do you also need to start training your own models then most likely or >> I I don't know like there's so many

different types of software companies and there's software companies that serve older esoteric industries

that probably aren't in a rush for AI >> and then there's industries or so or areas where the

the AI is just not that obviously valuable right now >> like for customer service everyone saw that coming. But I will say that if you

that coming. But I will say that if you buy the idea that AI will eventually eat all white collar work, which it seems to

be good at, then it's going to um disrupt most software because most software is built for adding efficiency to white color work,

>> most business software.

But when when somebody hears like all these the the model companies like I look at open AI and you know if you if you kind of like take the current narrative is they're spending billions and tens of billions of dollars to train

and and run these things right >> isn't it super expensive or is it not as expensive as people would think? um

certain types of AI work is not as expensive as you might think, >> but some of it is and it has required that we've had to, you know, allocate significant parts of our balance sheet

to it, but it's not the tens or hundreds or billions that get thrown around. Um,

at least for the type of work that we're doing. Yeah, I guess part of that too is

doing. Yeah, I guess part of that too is like if you just if you kind of go under the surface a little bit, it's like Azure is not Microsoft's not actually giving OpenAI $10 billion. It's giving

them credits to run chat GPT which is like you know people pounding that all day every day. It's like one of the most frequent kind of use cases and it's just cloud cost or cloud spend essentially at the end of the day.

>> Exactly.

>> Yeah. I think that's and that's where a lot of the magic and innovation is happening in uh AI at the moment. It's

at runtime. H. So, did you when you kind of decided, okay, we need to just rip everything out and start over like what was the process just like how did you how did you figure out how to approach this and what to do?

>> Well, I had the great benefit of being an outsider um cuz I was chair chairman of the board, but I wasn't in an operating capacity. Mhm.

operating capacity. Mhm.

>> So, you know, as an outsider, you can make all of these broad statements that on the ground are not true in 10 different ways, but

at a high level um can be powerful and quite meaningful.

And you know, the things I saw from the outside were that we needed a lot of clarity in our strategy.

We needed a lot of focus in our company.

and and strength and resilience in our culture. We needed to be a lot more

culture. We needed to be a lot more customer oriented in our commercialization strategies, not force everyone to talk to sales. Our pricing

was super complex. Um, and so I came in with a lot of those ideas written down. I had a form of a manifesto and a 10-point plan that I

acted on. I haven't had that clarity

acted on. I haven't had that clarity since. you know, once you're like in the

since. you know, once you're like in the trenches, then it's all gone. So, I I I'll never be as effective and as strategic a CEO as I was before I was

CEO again.

>> It's just a it's just a tricky thing.

That's why breaks are so important for leaders.

>> You know, even one day away for me, I find that like, damn, I get so much clarity that I didn't have when I was just smashing my head against the wall.

You see that a lot with young leaders. I

made that mistake, too. they're just

grinding seven days a week >> every week of the year and it's actually quite difficult to have clarity on what the most important work to do is if

you're doing it anyway. Um, so that was part of it. And then the other part was, um, you know, I went in deciding that I was going to very much bet on myself in

a way that I hadn't in the past. Like

obviously I had >> in a very practical way when I when I co-ounded the company. But, you know, it's easier when you're younger to question yourself when there are smart people around you. Coming up with, like

I said, those 10 reasons any given decision might be bad. And so sometimes I'd hesitate on decisions. Sometimes I'd

kind of like pair scale down decisions >> and >> this was back in the 2018 2017 version.

>> Yeah. Yeah. Before intercom started in 2011. So I had a big long run before I

2011. So I had a big long run before I left. And you know I did what I see so

left. And you know I did what I see so many CEOs do which is just try and satisfy multiple stakeholders and play it a little more safe. And I just

decided to never do that again. and you

know I now make rash decisions and um you know go with the first idea that came to mind and I've learned that most of those decisions are really good decisions >> you know probably a bunch of them are

not or probably they could be optimized if I like sweat over them for two more weeks maybe they couldn't maybe they'd get worse but now I'm just making mostly fast and loose decisions the hard

decisions I'll procrastinate on and then when it's like overdue then I'll make the fast and loose decision you I should have done it two weeks ago.

>> Exactly.

>> But you know I I really have found that that that um you know there's a lot to be said for founder intuition because the intuition contains so much you know

tacet practical uh undefined undescribed understanding and knowledge that goes a long way. And so when I'm talking to

long way. And so when I'm talking to young founders now and I was speaking to one yesterday and she was trying to land on a a name for a company. She gave me a list of 10 names. She said, um, you

know, what do you think? What do you think of these names? Such and such a name is my favorite, but a number of people have said it kind of has negative connotations. And I'm like, if it's your

connotations. And I'm like, if it's your favorite, it's your favorite. Choose

your favorite. She's like, "Oh, yeah, you're right." She needed that like

you're right." She needed that like little bit of push. Just like, "Go with your good."

your good." >> Yeah.

>> [ __ ] the Midwits. They'll always have all these reasons your strong ideas are wrong, you know, and and and you'll never have worse regrets as a founder

than if you make mistakes and the mistakes are someone else's idea, not not your own.

When you like went against your intuition, >> time and time again, great artists and creatives and interesting people, they got the right idea, but they don't

[ __ ] lean into it. Um, so that's one of the big things I learned anyway was to bet on my intuition.

>> Yeah. I feel like I've I've kind of run into similar things actually where it's like you >> you didn't fully execute on or didn't fully lean all the way into the thing that you thought you

>> that you thought was right or like you you took someone else's advice on something that maybe they don't know what they're talking about or they're more removed from it

>> or Yeah. or like they're very smart such that they are absolutely not wrong with their criticism, but they don't have the broader insight that you have that tells

you that the criticism doesn't negate the power of the insight.

>> Like you you as the founder are the insight holder. You're the one that has

insight holder. You're the one that has to keep that flame alive.

>> It's not that's not their flame.

>> And so actually having these conversations can be very helpful. you

they can help generate ideas for how you can approve for your flame.

>> I'm like stretching the analogy now, but um but allowing them to smother that flame that you were excited about it tends to be a bad idea in my my

experience anyway.

>> And you were initially I know you just mentioned like a little skeptical about AI just when it when it first kind of hit. Was there a kind of evolution? Was

hit. Was there a kind of evolution? Was

it talking to the team that was like, "Oh, maybe there's something here." Or

did you still push back a little bit after that?

>> I just have learned to be skeptical of all hype.

>> And even still today, there's a lot of AI talk that is kind of uh propelled by group think and FOMO.

>> Like a lot of people talking about the power of AI. Like you'll see these, right? You

of AI. Like you'll see these, right? You

see these posts on social that says, "Holy [ __ ] I just started 10 companies and made $50 trillion over the weekend."

>> The vibe coded.

>> Yeah. By asking Chat GPT this one [ __ ] thing.

>> Yeah.

>> Like >> the power of AI.

>> Yeah. You're like the power of AI. Read

this 15 part thread >> and then sign up for my course.

>> Exactly. And it's like no.

>> Yeah.

>> And obviously mine was like silly and facitious but >> they're like half believe. They're

significant. They're sufficiently

believable that everyone else is like, "Oh [ __ ] everyone else is doing cool things with AI." And if I am not also, >> yeah, you're going to get left behind.

>> I'm going to get left behind. I'm going

to look like a loser.

>> Yeah.

>> So, I'm I better say some cool stuff about AI.

>> So, there's a lot of that.

>> And as transformative as AI is, and I'm certain it's more transformative, >> I'm I'm I believe it's more transformative than the internet. our

current understanding of it, I'm also quite certain is quite infatile >> and is bolstered by a lot of hype. And

so that that's all, you know, I just felt the hype, but I'm like, wait a sec.

Let's see what this can really do.

>> So, you mentioned something that was interesting. You said that you think

interesting. You said that you think that AI is going to potentially be more transformative than the than the internet. Why do you say that? M

internet. Why do you say that? M

just because it's it it's flagging itself right now as like a really big deal

bigger than cloud and mobile I think you know has the potential to do actual work that's a new type of thing it's a new type of technology >> y

>> and always at the start of these new big deals by default we underestimate their potential because we can't imagine the

ways in which they'll get used and and be deployed. So I used to say at the

be deployed. So I used to say at the start of intercom and and I still think it's true to today and this is at in 2011 that we still haven't seen the degree to which the the

internet is going to transform the world >> and and that's probably still true now like you can ask yourself what is the internet like chat GPT like leans on the internet is it a product of the internet

like the AI itself like is arguably quite different and separate from internet technology but we probably we know I'm certain that we haven't seen the full effects of the

internet on humanity >> and so just kind of pattern matching I just think that this is clearly a big deal it's bigger

than other trends it's different in that it does work that humans used to do it doesn't just make them more efficient or connect them like the internet did and

it's only getting started and so we're lightly highly lightly underestimating it Um, that's all.

>> And it seems like the biggest kind of like the takeaway is this. This is the first time that the internet or or software does work for you. Like it's

never really been the case.

>> Software. Yeah. I mean, it has a little bit before, right? You can have we used to have software that would manage certain types of automation and >> like early machine learning type stuff

maybe.

>> Yeah. So, it was doing little bits of work, but clearly there's a chasm that's just been crossed. And again, that's going to change even in a few years when

people start to figure out how to apply agents and let them be more agentic and define their work. I don't mean like really really exotic ways. I just mean when you're going to have agents in the

workplace that understand more abstract briefs and can figure out how to solve problems themselves and then get to work with it. That's a fundamentally new

with it. That's a fundamentally new thing. The automation we've had in the

thing. The automation we've had in the past has kind of been it's been very deterministic. Even if you use machine

deterministic. Even if you use machine learning to kind of make it a little bit more efficient, it's been quite deterministic.

>> So arguably, it's been doing a little bit bits of work, but not it could be 100,000x the amount of work that software used to do.

>> Like you know, yeah, it's like it's at least 100x could be it's probably a thousandx. I mean, it's just so much more work. Not only is it going

to do all the work that humans used to do, it's going to do work that humans couldn't do, didn't have time for, or it was not economical to spend money to

have humans do. There's going to be so much more work done in the world. So

much more. as power gets cheaper, as computing gets more effective, as AI and agents gets more powerful, um we're going to just deploy these

things to problems we haven't even thought of yet.

They'll deploy themselves to problems that we certainly haven't thought of yet.

>> The agents will start deploying other agents. Well, maybe. But that again,

agents. Well, maybe. But that again, that sounds more like the Matrix. But I

just I mean like we'll have abstract agents in business and other places and that will know how to get to work.

>> Yeah.

>> So, so you mentioned something super interesting right there about the um the way software is going to change like the pricing specifically. that you

pricing specifically. that you mentioned. I think that was kind of a

mentioned. I think that was kind of a big thing that you guys kind of you mentioned earlier that you messed that up back in kind of the maybe a couple years ago, right?

>> Um and then your co-founder Dez when I told him that we were talking he's like, "Oh, you need to talk about how you figured out this new sort of outcome based pricing like this AI native pricing model." Totally.

pricing model." Totally.

>> I think you were the first one to do the first. Yeah. Yeah, we were the first.

first. Yeah. Yeah, we were the first.

Yeah. So, was it hard to to do that or >> um again part of the story is that we messed our pricing up so badly in the

past that we developed these very principle the rules for how pricing needed to work at intercom going forward >> and

created the the new FIN pricing in that image. So,

image. So, >> historically intercom had very complex pricing because we did so many different things. We sold sales support and

things. We sold sales support and marketing. So we tried to like serve

marketing. So we tried to like serve many use cases. Then you know it was very salesforward. So it was kind of

very salesforward. So it was kind of aggressive and we'd like twist people's arms and it was kind of unpredictable because it had all these metrics that reacted to usage.

>> And so we developed these principles that included stuff like you know transparency, predictability etc. But one of the core principles is a clean

map from value to price. And so we try and always practice value based pricing.

And the ultimate in value based pricing for a thing that does customer service is to charge for the service itself. And

the unit of service for us is a resolution. When it resolves a customer

resolution. When it resolves a customer issue to the liking of the customer, then we will charge you 99 cent and never before. So if Finn tries and

never before. So if Finn tries and fails, no charge. If Finn tries and actually learns a lot from the customer but can't finish the job and then passes it to the customer service rep and then

the customer service rep has all that information to get it started, get him or her started. I shouldn't call it humans. It's we're now mixing agents and

humans. It's we're now mixing agents and humans.

Um we still charge you nothing.

>> So yeah, we were just crazy principled about it. Originally,

about it. Originally, it cost us $121 per resolution, but we were losing money.

>> We just had the conviction that we would that surely we'd be able to make this cheaper, and we made it dramatically cheaper. Uh, so we have a very healthy

cheaper. Uh, so we have a very healthy margin now. But, um, yeah, we just

margin now. But, um, yeah, we just believed in the power of 99 cent. It

just felt very fair. And, you know, by the way, we pay $26 at Intercom when we do customer service.

We paid $26 for every human resolution.

>> So some people will do 15. Some people

will get it as low as five, but a lot of those people who think it's very cheap, they don't look at the fully loaded costs of, you know, benefits and office space and all the things.

>> Um, but if you can if you can answer a resolution instantly for 99 cent versus $26, you know, we knew that would be a a smash hit. But everyone has now done it.

smash hit. But everyone has now done it.

So basically everyone's doing resolution based pricing one way or another >> and it's like that across not just customer service but like a lot of different categories in AI. It seems to be like you you pay for the work that the AI delivers.

>> It's a little different than paying for the outcome.

>> So paying for the work is the AI did a bunch of [ __ ] >> and it's terrible >> and you have to pay for it.

>> Yeah.

>> Whereas we're saying the AI did a bunch of [ __ ] You only pay if your customer is happy with the work we did. It's a

very high bar. It's a very high bar.

>> Yeah.

>> What we like about that is that >> it's hard for the customer to argue with.

>> The only the only friction you have there is that the it's a new metric.

>> Anytime you introduce a new metric you you introduce friction because they have to learn seats. You don't have to explain anymore. But

explain anymore. But >> say V. This is part of the fun of playing in a new category.

>> So then how did you do the customer education around that? Like did you roll it out and it was super simple and everyone got it or was there like >> people got it quickly.

>> Okay.

>> Really quickly. Um

yep. It was it was remarkably effective.

Um but I was going to say it that the the the reason we like it is that the second reason we like it apart from the fact that there's a little friction um

uh when we uh explain this to customers and it it feels good that we're charging for value is that we then are incentivized

to increase our resolution rates. And so

if we can make our system more effective at doing work for them, we earn more too. So, it's kind of got a builtin

too. So, it's kind of got a builtin perpetual incentive and reward system where we continue to get paid for that.

If we just charge for work done, >> then we working all day just >> doing [ __ ] >> Yeah. Just stuff out there.

>> Yeah. Just stuff out there.

>> Why not?

>> So, we like that model. Um, some people have tried to charge for work. I think

it's okay, but I think the braver harder thing is to charge for real value and outcomes.

>> And one thing you mentioned was you're charging 99 cents. It used to you cost you a $121, so you're losing money.

>> That seems to be kind of the the current narrative around AI products is like, oh, they all have negative gross margins. They're all losing money. This

margins. They're all losing money. This

is a huge bubble. But I saw you reply to someone on Twitter that was that was commenting on this and you're like we actually have comparable margins right >> to traditional software.

>> Yeah. Bestin-class.

>> So what's the disconnect there?

>> Um you know we were always in the chat business. We

had a messenger early on and we started to build bots early on.

>> And this is Finn is a third generation of agent or bot.

>> I say that to explain the reason that we had an AI group from the start >> and we now have an AI group that's over 50 people. We have a research group

50 people. We have a research group within that. These are real AI

within that. These are real AI scientists, researchers and engineers.

Crazy team all colloccated in Dublin at the moment.

where we like relocate people to Dublin.

Amazing [ __ ] team. And the same reason that we're able to earn a strong margin on 99 cent. And by the way, people like I shouldn't say the name cuz

I'd have to double check it, but some of our competitors are charging two and three and $4. So we're also the cheapest. But the same reason that we're

cheapest. But the same reason that we're able to to charge a health or to to charge a health healthy margin even with such a cheap resolution is the same reason that we beat all our competitors and trade-offs.

And that reason is that we've got real AI, real AI teams, real AI technology, real AI scientists and researchers. And

I think that that's where frontier AI application companies are going to go.

And and that's a really interesting challenge because if you haven't noticed, there's a really serious war for talent. Um, but if you want to be a

for talent. Um, but if you want to be a real AI company, you need real AI talent. I think if you're making AI

talent. I think if you're making AI applications, again, in longtail, unsexy, unobvious esoteric

areas. I don't want to like pick on

areas. I don't want to like pick on anyone, but imagine if you're dealing with some sort of like plumbing management, you know, system. there's

maybe less competition.

>> There's less competition and you can probably use like the generic models for like a much longer period of time, >> but customer service or like customer agents is arguably the most obvious

application of AI. I mean, customer agents are going to do all the customer interaction in the near future. They're

going to do trillions of dollars of salaries. It's, you know, potentially

salaries. It's, you know, potentially the most obvious area.

then there's going to be a lot of competition and then the frontier companies like intercom and fin are going to need to do real AI. It's a long-winded answer but

real AI. It's a long-winded answer but that's the actual truth.

>> So is it that you train you trained your own models that you get local but we also just use the models in a far more sophisticated way. It's not hard to

take Sonnet or GPT and wrap it in some really light vibecoded software and have it answer some questions from customers. It's

extremely hard to do that at an average resolution rate in the 60s, which is the best in the industry, and beat all our competitors, and not hallucinate. It's

extremely hard. And it's just like a lot of very complicated tuning and configuration of the this multifaceted system. It's just [ __ ] hard.

system. It's just [ __ ] hard.

>> So it's if I have like a vibecoded customer support, >> I say vibecoded out to be facitious. Art

and peppers are not vibe coded actually.

>> But >> but if I were to go say I'm just going to try to compete with intercom and vibe code this thing. You could make a cool demo >> and well will it probably be that in reality on the back end I'm just running tons of cycles and it cost me $8 versus

you had it >> potentially that but it's just going to be again another technical term dog [ __ ] like in the market it's going to fail hard >> this is the one of the big things

software companies need to wrap their heads around in the pivot to AI you can't validate it um empirically

whereas you used to be able to sign in to the new app and click around and say it works. It looks great. You can't do

it works. It looks great. You can't do that with AI. You need to actually run scientific studies to see how it performs with real data and in our case real customers.

>> And part of our secret also is that we have billions of data points from our tens of thousands of customers from the help desk that we also sell to do rapid

high volume testing. It would probably be because you're you're not just getting software. You're basically

getting software. You're basically getting an employee that's doing work.

So if you hire an employee, you're not just going to like to show up and you say, "Oh, look good. You're hired." You

want to see a proof of work. You're

going to check references.

>> Well, it's that it's not even one employee. It's a team of employees.

employee. It's a team of employees.

>> So you're going to like really extens.

It's almost like if you're going to acquire a company of a thousand people, you're going to want to know deeply >> entire team. Yeah. You want to see like resolution rates, time to response, but

also like the feel of the answers like okay such and such an agent actually had an acceptable resolution rate but it's

super verbose and it takes customers a long time to read the questions and they find it really robotic and it has impacted the seesat score. So there's a

lot of art to it, too. But chiefly,

because I'm kind of messing my argument up now by saying that there's art when I was trying to say the science, chiefly, it's about running through the tests to see how all these different tweaks

impact the outcome. We've now run well over 300, we're probably approaching 400 different uh AB tests run to very high levels of

statistical significance to see how all these different changes of the FIN system perform in the real world.

>> So if somebody has lower or negative margins, it's probably because they haven't >> maybe that's part of it, too. But I

think that like like so I don't know this but the narrative publicly I believe is that a

lot of these coding uh co-pilots or coding platforms AI coding technologies have negative gross margins. That's what people are saying.

margins. That's what people are saying.

I don't know if that's true or not.

>> Um and that's a little bit more of a raw exposure to the generic models. At least

that's what it appears. And I don't know, you're a lot more close you're closer to the action. And I feel like your your your

ability to charge a margin above what you get from Claude Code or just Claude itself is going to be a little limited.

Um I'd have to like really look at these businesses and see the numbers better to give a better answer. That's a guess here. But um

here. But um there's a lot of companies out there that it's kind of insidious actually.

They've got negative gross margins and so much of their software is kind of like a tire kicking software. Uh sorry,

so much of their revenue is kind of like a tire kicking revenue and that people are signing up to try these new tools because they're really exciting and cool and they've got a lot of dead revenue.

Whereas companies and products like Finn where we only charge for usage and sorry not usage value and outcomes it's not possible for us to have dead revenue. We

can only get revenue when we actually deliver value in the world. If we stop delivering it then our revenue goes down.

>> And so there's kind of a two categories of AI companies now. There's those that have this kind of apparently negative gross margin revenue and a lot of dead

revenue and then there's those that are doing real work in the real world. And

that's an example of what I'm getting at when I'm talking about the fact that our understanding of AI today is still very naive. Like in just a year, we're going

naive. Like in just a year, we're going to be like, "Oh, wow. That was silly of us." And like, "Wow, that company we

us." And like, "Wow, that company we thought was really hot is not that hot, actually."

actually." >> Yeah. Or the one we thought was dead and

>> Yeah. Or the one we thought was dead and screwed is 10 times hotter now. Yeah,

>> that also >> cuz I think one one window of that argument is let's say you pay a really good engineer 300 grand a year all in.

>> Sure.

>> And then cursor's 20 bucks a month.

>> Sure.

>> Do you really think that cursor is not going to be able to charge instead of 20 100 or a,000? Like

>> if it's actually as good as it's supposed to be and somebody's paying a thousand times more like can you just increase the price?

>> But the question is like how much is it better relative to its competitor and how much are they willing to charge?

It's fair.

>> And it's going to be tricky unless one is vastly superior for them to be able to maintain a significant margin above cost.

>> Um I think the other thing that they were doing was it's a it's like a fixed subscription price, but then they had variable cost. So if somebody uses it,

variable cost. So if somebody uses it, >> it just it's the cost is out of control.

So it was just like the wrong pricing model initially.

>> I'm not sure how to think of it, honestly. I don't know. I'm not I

honestly. I don't know. I'm not I haven't like studied it very carefully.

But >> like my example where I said that we know that we pay our service reps, our humans $26 per resolution, but we sell

Finn for 99 cent per resolution. We do

that because we know that the argument, the following argument won't work. Hey,

you pay 26, we'll charge you half 13 bucks. We just know it doesn't work

bucks. We just know it doesn't work >> because we know that people know that AI is a lot cheaper. And if we're charging 13 bucks and the other guys are charging

two bucks, I don't know, they're going to start to look attractive.

Uh, you know, so yeah, the margin game is going to like what I think's going to happen is that there'll be a lot of players in all the categories. It's going to be hyper

categories. It's going to be hyper competitive.

>> Mhm. And then one will get an edge and then if they manage to continue to play their cards right, they'll com that edge will compound and other players will

then end up in certain sweet spots or certain areas. Some will be acquired.

certain areas. Some will be acquired.

There'll be consolidation, but there's not going to be 10 coding platforms. There might be two really big ones.

>> Mhm. Um but this will naturally get figured out but only when there are a much smaller amount will they be able to charge

>> a margin and and and and if they really do have negative gross margins right now that's the bet that other investors are making >> that these guys seem to be out and ahead of everyone else once they kind of have

that dominance and superiority in the market they'll be able to catch up. It's

an interesting game and it's a game that used to be played in the consumer world when people lost margin to customer acquisition and a lot of it didn't work out.

>> A lot of it some of it did some of it and massive mass to massive success too >> like Instacart and those types of companies Uber >> Uber Eats I think they all had these dynamics. So So if you can and I named

dynamics. So So if you can and I named all the winners so if you can end up as the winner then it it worked out. It was

good. And it's kind of like, you know, let's just say you're one of those ones like, "Oh man, our gross margins are negative. We better slow down. Like, we

negative. We better slow down. Like, we

better pair it back. Somebody else is just going to charge past you." And if they have the right, >> you know, investor set on board, employees, the right product, like they're going to smoke you. So, you kind

you kind of have to sort of play the game to the extent of that it's out there.

>> Well, part of the game is like a game of chicken. Like who's going to

chicken. Like who's going to not put who's going to not continue this very crazy game >> and continue to raise increasingly large

amounts of money, dilute themselves, set expectations in the hope that they become a winner.

So eventually people will drop out or be forced to drop out because they can't raise. That's how winners emerge also.

raise. That's how winners emerge also.

They just can't continue to finance their >> losses.

So is that a strategic thing that you've kind of thought about and how you position in the market or >> for some reason it's different and and you know we've been making a positive

gross and a strong positive gross margin on Finn almost from the start only at the start we were making a loss. Yeah.

>> And so that hasn't been a dynamic for us and we have not been cash constrained.

Um, and I don't know if I had a little longer to think about it, I'd probably do a better answer, better better job at answering, but doesn't seem to be a a dynamic.

>> I feel like in a sense, you're in a pretty interesting position where you had like the scaled customer base already.

>> I think that could be a really big part of it >> because you didn't have to go and say, "Oh, we need to spend 100 million on >> essentially customer acquisition to scale up really quick."

>> So, that's certainly a big part of the dynamic. And then you also had an

dynamic. And then you also had an existing AI team like was like, "Oh, we got to go hire 50 AI researchers." No,

we we have 50 people on our AI team already.

>> We had a customer base, distribution, AI team, and a lot of cash in the bank.

We're cash flow positive. So, all of those things are helpful, >> but a shrinking business. So, it's like, okay, we need >> Yeah. a a a business with a declining

>> Yeah. a a a business with a declining growth rate. We had not yet started to

growth rate. We had not yet started to shrink.

>> Oh, you didn't? Okay. No,

>> but the growth the the net new RR was was decreasing. We pulled up before

was decreasing. We pulled up before revenue started to decline. We had

>> uh slowing uh growth rates, but we never hit that bottom thankfully.

>> Okay. Yeah, that's But so it seems like you you almost had like the perfect setup of doing a

like lifealtering or like company altering strategic even though it it kind of seemed like a terrible situation you were in. You actually had an interesting setup to face this course because you had

>> the sort of uh the team, you had the distribution, you probably had I'm assuming a lot of your investors were like, "Oh, we like AI is great. Can we

get some of that AI dollars?" Of course.

>> Yeah.

>> Like >> the following two things are true. And I

mean that strongly, so much so that I resent anyone that claims that that's not the case. It's true that this opportunity was handed to us on a

platter.

It's just unarguably true. Customer

service, you know, we were in the right space and there's just such an opportunity there and we just went through all the

advantages. It's also true that it took

advantages. It's also true that it took a willingness to be kind of [ __ ] crazy to bet the farm on this new fancy

thing. I mean, we were the first to have

thing. I mean, we were the first to have a billboard with AI on it.

>> We were the first to, you know, I said we're going to spend 100 million of our own balance sheet. We're the first to say that they were spend their own real money on AI. We're the first to just of a late stage company to say that we're

going to pivot AI. So, and and we weren't the only company in trouble. So,

it still required bravery um and craziness, but yes, the opportunity was also presented to us. It was ours to screw up for sure.

>> Yeah.

>> And both could be true at the same time.

>> So something interesting that you mentioned a couple minutes ago uh a little earlier was you're not really thinking about this as just specifically customer service. It's kind of a broader

customer service. It's kind of a broader opportunity than that. Can you expand on that?

>> Yeah, I mean, you know, we think that everything in a business in a business that doesn't involve making a product or delivering a service is going to be done

by agents. um you know whether it's deep

by agents. um you know whether it's deep in a company doing accounting or whether it's on the front line with customers

helping them you know buy your product.

you know, there'll still be for some quite some period of time humans in all these departments, but um there's going

to be agents doing the a broad spectrum of not like nonstrategic work, the stuff that doesn't define and make the company

different. Um

different. Um and I don't think you're going to have multiple say customerf facing agents. I

don't think it makes any sense to have an agent that chats with your customer on your website to help answer questions about the product and then another agent that helps it get onboarded and use the

product more successfully and then another agent that answers questions when it gets stuck and then another agent that upsells the customer. Doesn't

make any sense. And certainly in the real world, customers don't see a distinction. When you

distinction. When you are in a boutique store and you have a question about a product, you'll ask the person behind the desk.

>> If you bought a product and it was broken, you'll go back and ask the person behind the desk. If you

uh as that business want to upsell a customer on like a more valuable item, it's going to be the person behind the desk.

>> So the customer doesn't have any distinction. Um and so I just think

distinction. Um and so I just think making a bunch of different agents is going to be wildly crazily ridiculously complex and as such people

will want one to rule them all. And

we're calling that thing a customer agent. and and the goal is to deliver

agent. and and the goal is to deliver concierge experience to every single customer. White glove service for anyone

customer. White glove service for anyone who shows any level of interest in your business throughout the customer life cycle.

>> Um just like agents are better than humans at service, they're going to be better than humans at all of these things. Um yeah, there won't be a single

things. Um yeah, there won't be a single customer that won't get the attention they deserve. If you really wanted to

they deserve. If you really wanted to treat them with the dignity and respect you felt that they had given that they're paying your bills and potentially making you rich,

>> um you certainly wouldn't make them wait multiple days for an answer.

>> Mhm.

>> You'd be on it. Every single whimsical question they had, you'd be there for them. And in the future, customer agents

them. And in the future, customer agents won't just, like I said, solve problems, but they're going to be experts, consultants. Think of the

experts, consultants. Think of the clothes like the the the the use case of buying clothing online.

If you imagine the agent as being a really an AI agent as being a really effective version of the thing that the humans do today, hey, do you deliver to this place or can you change the delivery address or I need to do a return?

>> Y, >> you're thinking of it all wrong.

>> Customer agents are going to be the very best stylist in the world, right?

they're going to be experts in their field. Like, hey, I know that you chose

field. Like, hey, I know that you chose those pants and those shoes and you told me earlier that you're shopping for this wedding you're going to this summer in Italy.

>> They could totally work. Um, people

usually buy these other shoes with that suit. And there's I actually wanted to

suit. And there's I actually wanted to show you this photo of this one celebrity that wore this shoot to this, sorry, this suit to this show. Um, and

they wore these shoes and I think it looks awesome >> and um, it's up to you, but if you want if you upload your can you send me your a photo, let's see what it looked like on you.

>> Like I I think it looks pretty dashing.

Um, so anyway, I don't want to do a hard sale or anything, but you know, if we wrap this up now, I'll give you 5% off this package, too. You're I guarantee people are going to be like, you know

what? Sold. You see how people talk to

what? Sold. You see how people talk to our agent Finn all day long. They're

thanking it. They're making little jokes. Finn is making jokes back. People

jokes. Finn is making jokes back. People

love it.

>> And so if it can give you that experience, >> I think people are going to love it, too. So that's that's clearly the future

too. So that's that's clearly the future of all different types of businesses is >> customer agents doing all the things incredibly well that don't differentiate you from your competitors. Well, you

could even say with that like the the customer or the the customer agent could say people just really want this like new suede suit like everyone is asking

the suede suit we don't have one maybe we should make a suede suit.

>> So on the back end so you're talking about like an operations agent which obviously makes sense too.

>> Yeah.

>> So customer agent is all the customerf facing stuff. Then you got the

facing stuff. Then you got the operations agent which is on the inside.

>> Do you think that will be separate? and

TBD.

>> Um, I think people will have their own operational agents >> and that's I think that there's going to be companies that make kind of like glue agents or like everything agents.

There's a cool company called Lindy >> um that's that's doing these types of things. There's probably many more.

things. There's probably many more.

>> Um, and then there's going to be specific agents that we want probably for legal stuff, accounting, customer stuff. Um, and then TBD, who's going to

stuff. Um, and then TBD, who's going to do the internal operations? And you

could imagine that internal customer operations be done internally. And we're

working on certain versions of that where it'll work with your team, ask questions of the humans when it doesn't actually have answers and build up its repertoire of knowledge. So I think there'll be some overlap, but TBD what

the operations agent looks like and who will build that. So, how do you think the way that companies are designed and run is going to change just how how we

see this kind of agent uh almost like an agent for each department? Like as that kind of evolves over the next >> decadeish or so like is it going to be >> you know it's a single person and they

have 20 different agents that just represent every single different function of a company like is that the extreme or >> like I don't know. I think that it's not

just going to be like a manager and all these IC agents. I don't think because there'll be an amount of the work that the agents just can't do that humans will still want to do

>> like always even in 10 years.

>> I think there'll even be like a preference in many instances where they'll want to actually differentiate themselves from the shitty companies that only use agents

>> and say, "Oh, we messed up. like we get on a quick call and they like apologize and and like it's going to be a really interesting experience.

>> We'll get to a point where you're like it's like a peak behind the curtain.

It's like it's like you'll tell your buddy like the craziest thing happened to me today. I was like on this store blah blah blah blah blah and something [ __ ] crazy happened and like I then

out of nowhere ended up talking to a person and he's like based in San Francisco and he was just like this guy in a flannel shirt and apparently he

runs the thing. It'll be like archaic and quaint you talk to humans but it'll make sense. It'll be remarkable. Yeah.

make sense. It'll be remarkable. Yeah.

But that's in the outer >> reaches of of this future world. for the

next 10 years, the agents won't even be able to do a lot of the human stuff.

>> Um, there's a long way to go to cover 100% of all demand for service.

>> Um, and so my point is that it won't just be managers and IC agents. It'll be

a mix between managers and IC humans with the agents. And maybe what it'll look like is you have like an an IC type manager who can do some of the human

work and then has peers who work alongside it and also work or him or her them and and also do their work.

>> Understand? Huh?

>> So, but this is just one of the examples of like smart assets like me can get on podcast and say how they believe the future's going to look. It's all

[ __ ] We haven't a clue.

>> Yeah. And you're probably going to be biased towards people using Finn versus not.

>> Well, of course they not for sure. But

like >> Finn aside, my idea about how humans and agents are going to work together, that's just [ __ ] [ __ ] that I almost I mean almost just made up now.

Like >> I can imagine it, but I don't think we we have a clue. None of us have a clue.

>> You should be very skeptical of people who are like very very confident about what the future's going to look like.

>> Yeah. So one thing that I have noticed is that you have not really said intercom much. you specifically when you

intercom much. you specifically when you created the AI native version of the product you called it Finn.

>> Um was there an like a certain significance around that decision?

>> Yeah, very much so. Um

like historically brand prestige meant something important.

Um, even Apple, for example, apart from maybe recent years, so it could be a bad example, have historically and still consistently make good products and and their

reputation for being hardware people at the very least um is relevant today because people are still building hardware. We want new hardware. So,

hardware. We want new hardware. So,

Apple's brand and reputation as a company that makes great hardware over the last 20 to 30 years is relevant today.

In when a new technology or trend emerges, not only can that brand maybe not

translate, but it can be a kind of a a ball and chain and a and a counterwe, >> something that works against you. And

we all have without saying it out loud realized that AI is kind of like a new person or sorry a young person a new company's game. Like we kind of imagine

company's game. Like we kind of imagine that all the new hot AI is going to come from new hot companies. And you look at them it's like a bunch of kids working

365 days a year on negative gross margin companies. Just kind of joking about

companies. Just kind of joking about that one but um and that's our image for like what what that looks like. And so

if you have like a prestige brand from a previous generation that looks very differently from what your expectations are for what good looks like in this new world.

>> Yeah. It can work against you and and that's our perception with you know a brand like Zenesk, Salesforce or maybe even Intercom or their previous generation companies and no one expects

that they're going to be the leaders in AI. Yeah. Like if you heard that

AI. Yeah. Like if you heard that Cisco are making the most interesting

coding AI coding platform, you would say I I I'm going to need to see some I want to see that.

>> I would say there's zero chance.

>> It's probably not true.

>> I would say right >> well it's like like with this is actually a bad example of your point because IBM made Watson this new thing.

>> True, but Watson wasn't that great actually. I mean, it was new, but it

actually. I mean, it was new, but it wasn't the modern AI.

>> That's fair.

>> Right. So, but but you know, it can happen, but it's it's pretty rare. And

so, for us, we're proud of who we are.

We're intercom. We're unique in that we have the best help desk and the best agent for our part of the market. And

that's why we're winning a lot of customers from from people like Zenesk because people want something that works together.

>> You can't really get that with the standalone guys. Um, but we we really

standalone guys. Um, but we we really need people to know that the Finn thing thing is a new thing.

>> And so we want it to kind of like win on its own. And you look at our billboards,

its own. And you look at our billboards, for example, don't say intercom on them.

>> Like we're spending tens of millions of dollars on advertising that has no intercom on it. We send all the traffic to fin.ai. Only on the bottom of the

to fin.ai. Only on the bottom of the page does it say an intercom product.

We're not trying to hide the fact that we're intercom. We're just trying to let

we're intercom. We're just trying to let people see it and you for what it is because it is new. I mean, the AI group was like six or seven people when we started working on Finn. They're 50 now.

You know, we were spending a couple million dollars in AI um back then.

We're spending, you know, 100 million plus over the next year or two. Like

there's so much that's new here and we've made a dramatic pivot and we need to package that in a way which people understand. Hm.

understand. Hm.

>> Probably it's probably kind of common sense, but it's a kind of a wild thing to do when you have a respective brand, but it is actually working. And now we

have a great channel on intercom.com. We

sell a ton of software that way. But now

we have a new channel, Finn.ai. We sell

a ton of software that way.

>> Um, I don't know. We'll see. It's

complicated.

>> You mentioned you're spending tens of millions on billboards. Do billboards

work? I think they work when you're trying to establish uh a new leader of a new category in people's minds.

>> You know, just brand recognition, you can brute force it.

>> I think you can. It doesn't mean that they'll develop preference for it, but I think they have a way of there's an insidious way of that

positive exposure to brand a brand influences um people's perception of the company. If you've heard it many

the company. If you've heard it many times, you're kind of like, I've heard that somewhere. That's an interesting

that somewhere. That's an interesting company. I think I need to check them

company. I think I need to check them out. And yeah, we we we're we're

out. And yeah, we we we're we're fighting the inertia of the the the the the

star that is the hot white star that is intercom and and therefore it makes sense to spend a lot. Now I think those $10 million are are billboards and they're also digital advertising etc. So >> oh fair. So it might not all be I this

is you know I shouldn't know how much you're spending on all the various things but >> I have smart people that do that but I do believe it works to some degree.

>> Yeah >> and and you said something really interesting in uh in the past you mentioned when you came back to intercom that you just didn't like the brand at the time.

>> Well I didn't like the brand expression.

>> So yeah. So what was >> I thought it was like very robotic and corporate and [ __ ] M >> we always were brave and different creative and we see ourselves as that

kind of company. And so when a brand shows up differently, it's going to make everyone feel uncomfortable. It's like

you having to go to a party dressed in I don't know you. I actually don't know you well enough to finish that sentence, but dressed like a [ __ ] dork.

>> That's what it felt like.

>> Yeah.

>> Yeah. maybe slightly different topic, but you um talking about some of the AI stuff, the sort of if you look at like the two kind of like CEO archetypes, there's the manager CEO and there's the

founder CEO.

>> You probably fall more in the founder CEO bucket.

>> Well, I am a founder.

>> The founder CEO.

>> Uh how do you think about those two archetypes as we kind of go into the next couple years? Well, I just think that I know that AI is so [ __ ]

competitive and the people working on it are young and very hardworking kids in their 20s

>> and we worked super hard back in the day.

Maybe they're working harder. I don't

know if that's true, but you like there's no time out time timeouts in in this game. Like you just got to push aggressively every day of

the week and work at a cadence that is quite unfamiliar in the latest stage software days. In the latest stage software days,

days. In the latest stage software days, every year or two, you check in in your strategy.

You'd do a quarterly exec meeting. You'd

review the goals for that quarter and set the KPIs.

Uh after having reviewed the goals from the last quarter, maybe once a year, you'd have the innovation and strategy team present to you >> some new innovations.

>> They tell you what's coming.

>> Yeah.

>> The R&D team would say, "Here's what we've been working on for the last year."

year." >> That's good to hear from. Yeah.

>> Yeah. And you'd like be like, "Cool."

Mhm. I like that. Oh, cool. cloud. Okay,

great. Let's make sure we have something for our cloud strategy. I'm really like being a [ __ ] ridiculous. Yeah, people

like, oh, maybe we need the NFT strategy or >> whatever. Maybe. Unfortunately, many

>> whatever. Maybe. Unfortunately, many

people had an NFT. Um, and there was this like slow cadence and it maybe, and I say maybe, and really I mean didn't,

fit the professional CEO who was part of an organization and they were really good at that cadence, structure, order, harmony.

But in the AI world, it's like imagine you set your new strategy and your goals for the quarter. It can be day three on

the new quarter. Someone releases a new model and your competitor releases a new product and the other competitor just

announced their new financing and the product you all were really hyped about kind of didn't work and you need another quarter and you realize that your old brand that everyone loved

is kind of [ __ ] holding you back. You

need a new brand >> and you can't get the fin.AI domain and it's really the only only one. You're

going to have to talk to the [ __ ] guy who owns it. He wants a lot to be paid for it.

>> Did this actually happen?

>> This actually happen.

>> Wow.

>> And and it's like this constantly dynamic wild thing. And um yeah, professional CEOs I just think anyone who wants to work in that cadence is

just going to be left in the dust. And

so I just think that this time demands the founder MMO, which is flying by the seat of their pants and, >> you know, rolling their sleeves open,

getting involved and making quick and fast decisions and not being afraid to make a mess.

>> Well, yeah. When you when you think about that an annual product plan, let's say we're sitting there, >> you know, August of 2024,

>> I think Bolt launched in October of 2024, lovable somewhat recently after that. Then you had Vzero from Verscell.

that. Then you had Vzero from Verscell.

You had Replet launch.

>> Yeah, we had like Windsurf and Cursor.

You all these companies basically break the record >> for fastest AR growth course. You know

what's what's the joke? Accounting rules

aside, whatever. Did you see the one video that went viral? Like,

>> oh, that's actually one you reply to.

>> That's what I replied to.

>> Yeah. It's like, you know, gross margins aside, >> yeah, >> extreme growth. So,

>> you professional CEO, like you check back in October of 2020, >> plan doesn't work.

>> Yeah. And so part of what you also need as employees who are willing to work in that chaos >> and you know I've always got feedback that like oh we're making too many

changes and you know >> is that the hardest part of working for you is that you're always >> No, we can get to what the hardest part of working for me in a second but I

always got that feedback and I actually took it as a negative and was quite insecure about it because I'm not a hyperorganized person. I'm not the

hyperorganized person. I'm not the professional CEO operator.

>> Yeah. you know, I'm the person that hires that person. Um,

but now I know it's a feature, not a bug. Actually,

bug. Actually, >> in the AI game, you have to change. We

would be [ __ ] dead if we didn't rip up all of our rules and all of our plans and all of our strategy and ICP and everything else multiple times in the last couple years.

>> Yeah. Because there's been a a quite a few startups that have emerged trying to do customer service, right?

>> Yeah. Yeah. Yeah.

>> You have to stay adaptive. Yeah.

Speaking of that, actually, if you were a startup right now, if you had to start over, like if you're like you're not allowed to do intercom.

>> Yeah.

>> You're not like you have to no customer service allowed. What kind of

service allowed. What kind of opportunities do you think are out there for startups right now? Like is there like a framework that you have?

>> No. I like

starting a new is so difficult particularly now that things are moving so quickly because you need to start something long before it's valuable or

cool like well before the category takes off.

>> So so much of the early stage thing is a crapshoot because of that you know which you probably know.

>> Yeah. Um,

so for me, when I think about starting new businesses, it's always just needs I see in the world that I will be really excited to fill.

>> And I do think that that works in technology today, including in AI, if you stay close to fields of personal interest, and of

course you can try and build stuff in fields that are not interesting to you, but I hate that approach personally. I

hate that throw [ __ ] at the wall approach and see what sticks. It's just

really really hard to build for people you don't understand or care about.

>> Yeah.

>> But if you can stay close to domains that are interesting to you and that you care about, really get to know them and then also be the person in that field that knows the most about modern technology and applications of AI. I

think that there's opportunities to build cool new stuff. Um, but so many of these AI companies are like really abstract also. Like

abstract also. Like >> how so?

>> They're building like platforms for AI companies that do a certain type of niche. Do you know what I'm saying?

of niche. Do you know what I'm saying?

>> I mean, I've definitely seen some of those. Like we

those. Like we >> But then you look at the revenue and their customer base, it's like, "Oh, it's kind of real."

Like, and I believe it's real for what it's worth. Now, maybe it's actually

it's worth. Now, maybe it's actually not. Um, I'm not going to name the

not. Um, I'm not going to name the company. Maybe it's like the optim the

company. Maybe it's like the optim the optimist in you is your like >> No, I'm not an optimist actually. I'm

totally all skeptical on the AI thing.

I'm not an optimist. I think at least half these companies are going to flame out. Like that's how it works.

out. Like that's how it works.

Particularly when everyone's fomoing into a space. Maybe twothirds are going to flame out. Like I kind of also everyone should go into it all eyes wide

open. But um I uh

open. But um I uh yeah, I do think that there are those abstract opportunities. If AI is going

abstract opportunities. If AI is going to be as big as AI seems to be, then building tools and platforms for AI is probably a thing. That these these second order things that I never would have thought of. You have to be a

certain type of nerdy individual to like get excited about that, too.

>> Yeah.

>> But, you know, when you get far enough away from early stage, all your early stage ideas kind of like kind of wither and die. You need to be a desperate

and die. You need to be a desperate founder to smell the ideas quite often.

>> Yeah. But and I feel like when you're just thinking about competition, it's like there's probably less incumbent competition in helping AI native companies, but there's probably a lot of startup competition in

the >> you know, how do you do eval better or um you know, I think there's there's this one I don't know if this even worked. It was like employee reviews for

worked. It was like employee reviews for AI agents or something like that. Like

it's >> you can get pretty esoteric, >> but maybe it does work. Yeah.

>> Yeah, maybe it does. Yeah, maybe it does.

>> Uh, one one question that I know I wanted to ask you about was you are originally from Dublin, Ireland, obviously also based here in SF. You

probably were one of the first scaled and successful kind of distributed teams. >> Is that still something you guys lean into a lot? Um,

>> so we're like semi-distributed in that we have multiple offices, but we try to not hire remote. Um, it's just kind of a a painful fact of our existence. like it

was an advantage in so far as it gave us access to talent markets that we wouldn't have if we were just in >> San Francisco and we have benefited and continue to benefit quite a bit from being in Ireland and London

>> and being one of the more interesting companies to build actual technology in >> uh in those places but being far away from the humans that you're

collaborating with and kind of spending most of your life with and on such a deeply uh emotional and a profound adventure with sucks. It's horrible. I

don't recommend it to anyone. Um,

>> and a lot of people sleepwalk into it because, you know, they're kind of like, oh, my founder, my co-founder doesn't want to like move for whatever reason

or, you know, we can't find a designer here or whatever it is. I'm not saying you have to be an SF at all, even though there are great benefits to being here.

But um they they take the easy path of building a distributed team and then just make it ever harder to undistribute themselves.

>> So much so that they never become undistributed. And uh I just think it

undistributed. And uh I just think it sucks. I just I just know that in

sucks. I just I just know that in person's better. You know, there's, you

person's better. You know, there's, you know, the the days I crave were the days where we were in the same office or I was back in Dublin. You're going in in

the morning, you're meeting your co-founder or someone for a coffee. You

kind of can have um ad hoc meetings.

I don't know, it's just more fluid and fun. Like I spend

fun. Like I spend the majority of my day looking at a

little digital square in front of me.

And like for my one beautiful human life on this earth under the sun, I choose to

[ __ ] stare at this little digital square. I don't know. doesn't I don't

square. I don't know. doesn't I don't know doesn't seem like the best way to spend your days.

>> Probably not. I don't know if Brian Johnson would approve.

>> He He probably hates this idea. But I'm

not even talking about it from a wellness perspective. I'm just talking

wellness perspective. I'm just talking about like just richness of life experienced.

>> Yeah.

>> So I do it out of necessity, >> but I just think it's I know it's just so much more fun to meet people in person.

>> Like we all have various >> trade-offs we make. It's okay if you make the conscious trade-off. I'm just

telling people don't make it unconsciously.

>> Yeah.

>> Um >> and it sounds like if there's if there's multiple distributed team with multiple offices and it's like the people that are in different cities because of the like the remote thing cuz I do have

people in SF I meet in person.

>> There's a lot of people that meet in person in Dublin and in London and and we try to keep teams together certainly in the same time zone. The time zone makes it even worse.

>> Oh yeah.

>> Way worse. cuz Dublin's 8 hours ahead and so I'm just starting my day. I'm

barely waking up. They're at the end of their day. They want to go home.

their day. They want to go home.

>> Yeah.

>> It's like terrible mismatch.

>> How do you think uh the media has changed over recent years? I know you I mean we're on a podcast. I'm

>> I'm an influencer, I guess, at the end of the day. Like what's what's kind of the >> It's depending on who you're talking to.

It's like sometimes it's like an embarrassing thing. Some people are

embarrassing thing. Some people are really excited for it. Um yeah, for me it's like I kind of accidentally became one. So it's like you got to lean into

one. So it's like you got to lean into it. But yeah, um how do you how have you

it. But yeah, um how do you how have you kind of seen that evolve over time as a founder and how you kind of play that for your company?

>> More and more people evangelize the idea of founders going direct and it's a trend that's real. You've even got people like Mark Zuckerberg going on a

range of different podcasts speaking in long form in a very raw way. And I think that that's only great for people like Mark. I think without that he was

Mark. I think without that he was substantially less human to the world.

And so for founders and having this opportunity to go on different people's podcasts, have these long form conversations, publish their own podcasts or videos, even just their own

writing on Substack or even on Twitter or X. Like all these things are

or X. Like all these things are outstanding. I will begrudgingly admit

outstanding. I will begrudgingly admit that uh there is some at the very psychological stamp of

approval or or or value stamp that conventional media still delivers. And

when you're covered by the Wall Street Journal, I don't know. I guess it's got a certain sheen to it still.

>> Yeah. People people kind of lean forward when they >> I think so. Like now if you think carefully about it, is there a good reason for that?

Probably in part no, but in part yes, because to actually get covered by the Wall Street Journal, you need to be comprehensible and interesting to the mainstream.

>> And that is a kind of signifier that there's something that's broadly more interesting here than the nerdy podcasts.

>> Yeah. Um, but yeah, I just, you know, I just tell all founders to just tell their story early and often. Don't be

afraid to repeat themselves.

The new cohort are all vibes founders, at least the best of them. They're

really good at media.

>> Um, >> did they grow up on the internet? Like,

is that why or >> part of it, right? Well, they grew up with Tik Tok and they grew up with Instagram >> and the best of them learn how to promote themselves or at the very least saw how people who were good at

promoting themselves did it.

>> It's hard actually for a lot of founders. I'm not that type of person. I

founders. I'm not that type of person. I

don't want to do that. Like I don't want to create a bunch of [ __ ] videos of myself. It's not interesting to me.

myself. It's not interesting to me.

>> Yeah.

>> Uh very painful. But there's people who like doing it and they're good at it and they get a lot of benefits now.

>> Yeah.

>> And um >> Yeah. So, I I I I I think the game is

>> Yeah. So, I I I I I think the game is now a hype game. It's now it's now it's now a cell phone media game. It's a

direct game.

>> And the more you can tell your story early and often, the better. And and and I think the last thing I would say is the more vulnerable you can be, the better.

>> Really, why is that a big deal? Well,

just because with the vibe marketing comes the [ __ ] >> and there's a lot of overly slick content.

>> Yeah.

>> And kind of dry podcast interviews that are very difficult to listen to in my opinion.

>> That's true. If someone's listening this far into this one, you got to give them credit.

>> What are you saying? This is dry. Is

that what you're saying?

If you can be vulnerable and authentic in yourself, I just think it's a lot more interesting and relatable.

>> Yeah.

>> And it's hard, you know, everyone on the internet is trying to take quick and cheap shots. It's hard to take shots of

cheap shots. It's hard to take shots of people who are just fully being themselves and not trying to sell too hard. So, the prototypical future

hard. So, the prototypical future successful founder, and I do think that this is like a strong signifier for someone worth investing in, is someone who is able to tell their story

confidently, knows who they are. They're

authentic. They're willing to make mistakes, admit when they're wrong, and is not afraid to be on camera and just produces content and communicates. Like,

those guys go so far. So far. Um, some

of them go too far. And there's some companies I can think of which I won't name that maybe they're all hype. We

don't we don't know yet. But I'm I'm I'm almost certain that the future successful founders will be will learn how to do this well.

>> Well, it's that whole of like distribution's all that matters like it that's not entirely true but to the point >> well it's all it's all it makes a big difference if indeed software is getting cheaper to develop.

>> Yep.

>> Right. What's left then?

>> Yeah. And to your point of uh maybe someone went too far, we don't know if they actually have a product and if it actually works, but we we know that we should care. Like there's that's there's

should care. Like there's that's there's two different sides of that problem.

It's like you made a great product, nobody knows about it.

>> And there's the other side like you have no product, but everyone knows about your product.

>> Two different problems. Maybe one of them is easier or harder to to solve for or get around, >> but well, I think you need to do all the things. And we've got a lot better at

things. And we've got a lot better at building software. Like when I started

building software. Like when I started building software, >> things like Ruby on Rails just came out and it was a profoundly >> you didn't have cursor. Like you had to actually write the code, right?

>> Forget cursor. We didn't have I don't know what version of Rails were on now that make it even easier. Like these

frameworks only just came out that allowed you to save a lot of time developing basic CRUD apps. Um now

software is commodified and yes now we have cursor and everything else which are going to add great efficiencies. Um

but distribution is not commodified at all. It's monetized.

all. It's monetized.

>> Um so you can pay your way in and if you are a new founder with not a lot of money, you're finding the incumbents like me with the hundreds of millions of dollars or tens of millions of dollars

dumping it on billboards allegedly.

>> Allegedly.

So what have you got left? You've got

your own vibe in Janice Qua and magic to give to the world. And so maybe this take pro

the world. And so maybe this take pro proves wrong, but I I I really think that the new hot founders are going to be ones that can sell themselves well.

>> Yeah. Well, I think one thing too is when you think about um you like if you think about from yourself and you just think of like do do how do I feel about constantly putting myself out like I'm

tweeting 20 times a day. I'm making a bunch of Tik Tok videos, make like 10 or 20 a day.

>> You know all that you're doing and you just assume every single person out there is seeing and consuming all and they just think, "Oh man, Turner is just tweeting all day. Turner's making

TikToks all day." But the average person is going to see maybe one or two or three of those and >> they might see 10 or 20, maybe 50% of what you put out. No one will actually see all of it. So it's it's kind of like

putting shots on goal.

>> Yeah. So, and a lot of times you don't really know what's going to hit when you're making content. Like I I will tweet something that I thought was pretty clever and like, "Oh, this is going to be a banger." And it just kind of flops. And then sometimes I'll have

of flops. And then sometimes I'll have one where there's like a typo and it was like took me two seconds to make, but it gets a million views. I'm like, "Shit, I almost didn't tweet that one."

>> Yeah. Yeah.

>> So, it's it's kind of one of those like sort of the same thing of like a high shipping velocity on product is just with content. It's just you just put it

with content. It's just you just put it out there. you do it 10 times a day and

out there. you do it 10 times a day and >> the ninth one and the 10th one is the one that really hits and that's the reason people know about you. But if you chop those off and you only did eight >> or you only did one a day or you did one

a week or one a month. Like imagine if you did 10 a day the hundth time is the one that hits 10 days and you got you hit versus if you were doing one a month or one a week and it take you 100. It would literally take

you a hundred years.

>> It's true. Unfortunately, I'm hearing like the one a week strategy. It's cuz

I'm not one of these young founders that is like social native where I like have all these ideas for pity tweets and I'm really excited for the world to like me.

Like >> yeah, >> I don't I'm not that guy unfortunately.

So this is some of the advantages that the new kids do have.

>> Yeah, >> some people can work and tweet like Amjad from Replet apparently can work and tweet.

>> He's built a great product. The business

is on fire. The guy's out there in social media promoting it non-stop.

Apparently it all works. Kudos to him.

>> Well, and one thing to think of too is like you might think, oh, he's just tweeting all day or he's just all he's doing is just typing all day, but you're probably spending a lot of time in Slack. You're spending a lot of time in

Slack. You're spending a lot of time in email. So, you're just kind of shifting

email. So, you're just kind of shifting some of that to being a place that's one on many. So, it's like you can send an

on many. So, it's like you can send an incredible message, super insightful on Slack or an email to one person or if you put it on the internet, a 100,000 people will see it and you almost get

more credit for it in a way. So, you

know, there's downsides to putting the wrong thing out there on the internet, but there's also like unbounded >> upside. Yeah. You have to not overthink

>> upside. Yeah. You have to not overthink it.

>> You have to be willing to be wrong.

Yeah.

>> Yep.

>> Talking about putting things out on the internet. Uh I kind of wanted to just to

internet. Uh I kind of wanted to just to kind of wrap this up. I went to your personal website. Your website's

personal website. Your website's incredible. Like it looks like it was

incredible. Like it looks like it was made in 2003, maybe.

>> Yeah. It's it's designed to look like it was made in 1996.

>> Okay.

>> That they were the glory days for me when I first got online. Okay.

>> Geio Cities and all that good stuff.

>> Okay. I was gonna say Geio Cities. Yeah,

I had a Geio Cities website back in high school that I had.

>> It had the little um the sidebar ads thing. You know what I'm talking about.

thing. You know what I'm talking about.

I think you could pay for premium to get rid of it, but >> I was 12, so I didn't have that.

>> That was a beautiful thing.

>> Yeah. You said one of the things on your website. You said you took out a loan

website. You said you took out a loan for like $2.2,000 for your first flight to San Francisco.

>> Yeah. Yeah. Yeah.

>> Really? So, what's this? What's the

story with that? I mean, just that I didn't have a lot of money. That's

that's the that's the end of the story.

>> You were just hellbent on I'm gonna get to San Francisco.

>> Yeah. Yeah. I had one friend out here who I told me earlier that year like, "Dude, it's crazy out here. Um kind of like the

whole internet's being built out here >> and there's like so many VCs and you kind of have to be here." And yeah, the word was kind of getting out about San

Francisco. San Francisco wasn't

Francisco. San Francisco wasn't the center of Silica Valley in 2011.

That was just like a conversation.

People were saying, "Oh, it's starting >> to move to San Francisco >> from South Bay."

South Bay." >> Yeah, exactly. South Bay, Paloto and Certino and Melo Park and Mountain View.

That was where Silicon Valley was.

>> And so I moved here just as people were saying, "Oo, San Francisco's going to be the new Silicon Valley."

>> And yeah, I came here to be part of that.

had to be part of that.

>> Yeah. And another thing semi-related on the website, it said you it took you quite a while to raise the very first round for income. I think I saw that the pitch deck is on the website, which I'll

throw a link in the the description for people to see, but I think you were trying to raise 600 grand. I was trying to raise a million dollars >> and

I like just burned through every single investor intro I got. I just got noses everywhere. You see the thing, this is

everywhere. You see the thing, this is where I really sound like an old man, but like back then there weren't seed firms.

>> Like there was maybe three or four or five. So was this 20 >> 2011 like there weren't really seed firms like there were VCs

and I think kind of the way you did it was that you raised like friends and family round and then you went to the institutional people and then maybe you

raised a series A and the friends and family round was maybe your seed kind of >> like there weren't many people who were writing big seed checks. Now you see all these people, all these [ __ ] kids, um

I'm just kind of trying to joke here, trying to lighten the situation given how old I sound. Um announcing their seven, eight, n$10 million seed rounds.

There was one just announced today that I invested in and that was unheard of.

Seven [ __ ] million dollars for a seed round. Like if someone told you that in

round. Like if someone told you that in 2011, it would be actually funny. It

would be like it would be a funny joke.

>> You would have thought they were like literally like >> it would be a joke because it didn't happen.

>> Yeah, >> it didn't happen. And yeah, so anyway, I went to all the various angels I could get interest to. Got a bunch of nos.

Then I got a couple yeses. I like got 5k checks, 10k checks. I think I got a 50 or 100k check from Biz Stone who was

the founder of Twitter, one of the Twitter founders.

>> So, he was one of the big ones. And

>> like I met him, we took a walk around the mission for probably 25 minutes. He

shook my hand and said, "Yeah, I really like it." Yeah. Okay, cool. I'm in. Got

like it." Yeah. Okay, cool. I'm in. Got

like 50 100k like but really I really got the impression that like and I like Biz a lot. This is only a compliment to Biz. He ended up making like a bunch of

Biz. He ended up making like a bunch of just killer investments.

>> I'm like, god damn it. That guy really hasn't done his due diligence at all.

>> Yeah.

>> Like how what if he doesn't know investing >> exactly? But I mean it really worked out

>> exactly? But I mean it really worked out for him. But

for him. But >> um yeah, long story short, many painful depressing months. Came back to Ireland

depressing months. Came back to Ireland cap in hand like a [ __ ] loser to my co-founders. I had managed to scrape

co-founders. I had managed to scrape together 500k, much of which just came together in the last minute. But then Biz knew these

last minute. But then Biz knew these Japanese guys, Digital Garage, and introduced me to them. And I had a meeting with them during the day in

Dublin. It must been really late in

Dublin. It must been really late in Japan.

It was a Skype call. It was 26-year-old me. There was like 12 guys in suits.

me. There was like 12 guys in suits.

>> Okay.

>> The the the the lost in translation factor was very real. Did they know English? Yeah, they did.

English? Yeah, they did.

>> Okay.

>> But like I'm pretty sure that they barely knew what I was saying. I barely

knew what they were saying.

>> Y >> but it was something like I don't know how it went, but they were kind of like I'm not you know they said

something like very good. Uh we'd like to invest $500,000. Sorry. Yeah.

$500,000. And I was like okay. And you know that went on to be a

okay. And you know that went on to be a great [ __ ] investment for them also.

I mean a fair investment back then. Of

course seed rounds were we raised at a $6 million valuation.

>> So it's like maybe some dilution.

They're probably sitting around 100x at least, right? Something like that.

least, right? Something like that.

>> There's some dilution. I don't know if it's 100x but uh it was a good investment. I'll put it that way. So

investment. I'll put it that way. So

kudos to those guys. And and I got to like meet them over the years. We'd meet

in person. They were just the most polite, lovely gentlemen ever. So our

round was just really random. I didn't

deserve that million dollars, but it came together. And so I can't overstate

came together. And so I can't overstate my resentment for the kids who are raising $10 million at $100 million valuation >> in a week.

>> So why is that happening now? Um I think we're more aware of the potential of

uh these investment vehicles like you know u 15 years later we now see the plethora of unicorns and decacorns.

>> It's the first time is that we say decacorn.

>> Deacorn.

>> Terrible phrase but I'm embarrassed even that I said that phrase.

>> The most buzz word. I realized I've never said it before.

>> Yeah. You've never said it out loud.

>> Never said that word before. Um there

was barely billion dollar private companies back then. Yeah.

>> Like it wasn't a thing. The reason it was called unicorn was that it kind of didn't exist.

>> And now there's like little AI companies that are like worth a billion dollars 3 months after launch. I mean that literally happens now. And so

we weren't aware of and awake to the potential still of software, the internet, technology, cloud. And that's

why now when people see AI come around, are like they're like, "Holy [ __ ] we're not going to make that same mistake again."

again." >> Cuz these things are so much bigger than we realize. And and and just like, you

we realize. And and and just like, you know, SAS was bigger than we thought before, I I say it again that AI is probably way bigger than we could possibly imagine. And so, as ludicrous

possibly imagine. And so, as ludicrous as these rounds are, even though some will flame out and people will lose their money, some are going to make phenomenal fortunes.

>> Yeah. Because just like unicorns were so rare then and now we have companies worth tens of billions, maybe in 10 years it's not going to be rare for companies to be worth hundreds

of billions when they're private. It

might not be the thing that only five companies can achieve. Maybe there'll be 50 >> and so then it makes sense to lean in aggressively. But um not back in 2011.

aggressively. But um not back in 2011.

>> Yeah. Well, well, an interesting thing though when you think about the people investing that capital and how do you monetize it? Like you charge fees and

monetize it? Like you charge fees and you get the profits or the carries, right?

>> So in in public markets, fe there's fee compression. Sure. Like

so >> if if you're if you're running a public market fund, you probably can only charge about 1% of your capital. But if

you run a venture capital fund, private markets, you can charge 2%, right?

>> People charge 2 and a half or 3% a year.

So there's incentives for the capital to kind of be there when if people know the returns are there you >> you know you you can you you miss out on

with public you get uh pay your carry quarterly sometimes annually with private markets you need to actually exit >> so you don't get the carry quite as much but the incentives are that moves the incentives even more towards raising AUM >> sure

>> and getting that management fee so it's like >> you've got if the companies are doing well and they're growing into the the correct profile you have investors who will kind of step up and raise the capital to kind of solve for that and

keep them private, >> right?

>> So, I think the incentives are aligned whether you could say it's good or bad, like the incentives are set up to where that's probably going to happen >> and there'll be just a lot more money chasing this current capital. Yeah, for

sure.

>> Yeah.

>> And just going into once again many esoteric esoteric un unfamiliar unsexy categories. the the capital is

unsexy categories. the the capital is going to trickle down in many areas and directions.

>> Yeah. So, it's going to be interesting to see. Maybe VC firms will grow. Maybe

to see. Maybe VC firms will grow. Maybe

it turns out that the close to infinite VC funds will actually not be enough.

>> Maybe. I feel like that's a pretty hot take. I feel like most people say

take. I feel like most people say there's too many VCs.

>> That's what I would say.

>> Yeah.

>> Well, it's just really hard to differentiate now.

>> Yeah. What do you How do you decide like when you meet an investor? Yeah. I guess

you're a little different now from where you're at, but maybe if you can reflect on it and think about it. I mean, I guess at any spectrum, like how do you think about it? Like, do you look at like what what value will you provide to

me? Is it like a

me? Is it like a >> do I trust you? Like, would I want to get a beer with you? Like, how do you gauge it?

>> I have sharp hot takes on this. I'm glad

you asked. Okay.

>> Um, I think that value ad is [ __ ] [ __ ] I do think that some if they're particularly if they're operators can maybe help, but I think

it's really the job of the founder to run their damn business. And if they set up a situation where either they think or the investor thinks that the investor should be contributing anything,

something's going wrong.

>> Um, but for the most part, most investors are are not going to be able to help the founders. I mean, it's just a fundamentally different discipline. I

do think that they can help them thinking about future financings and about corporate matters, valuations, exits, and they can make appropriate introductions to other investors and

bankers and strategic buyers, etc. So, all of that makes sense. I think that's really cool. But I I'm I'm really

really cool. But I I'm I'm really uh allergic to the big firms that have apparently these suites of services. I

think that that's [ __ ] So, >> that's kind of one side of it. the the

the the the investors. And so I'm I'm repelled from investors that say, "How can we be helpful?" I'm like, "Come on, buddy. Like, you can't. Sorry."

buddy. Like, you can't. Sorry."

>> Yeah.

>> You typically can't. Not with me. Not

with the brand new founder. You can't

like >> you're I you're probably a great investor, but that's a different discipline.

>> Yeah.

>> So, I'm repelled from those types of people. I'm attracted to investors who

people. I'm attracted to investors who have the confidence to communicate directly and openly and actually know what they bring to the ca table and are

willing to disclose their own interests.

So when you get investors who are like I just want to be involved and like you know great founders and you're [ __ ] amazing and like oh I really actually love your company and you are

particularly incredible and I just oh the timing isn't right or something something >> I didn't say anything that was the most generic line of words I've ever heard.

That is 80 to 90% of investors still. I

like reconnected with a bunch of investors cuz we haven't raised since late 2017. So I've been like not exposed

late 2017. So I've been like not exposed to investors for >> oh wow >> eight years and I'm like oh maybe they've changed and I got the same lines

from a brand new cohort of investors.

But a couple investors would say things like um or will say things like um you know that opportunity is interesting

on the face of it. like that company but you know our return profile you know is you know 3 to 5x and at this stage we don't know if we can achieve that or just by the nature of this business

which is an arguably great business is not going to make as much sense as me allocating this capital to a slightly earlier stage company that's growing at this different rate and they'll kind of

like show their cards a little bit and have a grown-up conversation >> because then if it's authentic >> and feel. If the founder wants to feel hurt,

then it's kind of on them.

>> Whereas, if the investor is [ __ ] bullshitting them, which the founder will always feel, at least subconsciously, that's when I think the relationships go ary.

>> And so, the investors, while they think they're helping themselves from hiding their true motivations or or or hesitations about the business

themselves.

>> Yeah. Or keeping optionality of like my team to invest. Now,

>> I think that they're actually damaging the reputation with the best founders who will smell [ __ ] a mile away. So,

again, I've seen just time and time again the best investors, you know, the top 1% or 2%. They're more than confident to say, "This is a little

early. We'd like a little bit more proof

early. We'd like a little bit more proof on this. We like this and this, but this

on this. We like this and this, but this I'm a little concerned about.

>> Um, we'd like to get to know you a little bit longer." And it's just a chill thing. or this is not for us

chill thing. or this is not for us because we only do this other thing. But

as soon as you start laying all the like you're [ __ ] amazing and I do anything to work with you but for some reason I'm not investing in you. Absolute [ __ ] That's still 80% 90% of investors which

is really really surprising.

>> And it's just because sorry it's just because I think a lot of investors don't have experience managing people.

You know, when when you manage people and you have to give hard feedback and you have to build rapport with people who want to follow you and work with you, even though you told them things

that they didn't want to hear, you quickly learn um uh just the the um uh the authenticity pays >> and [ __ ] doesn't. And um I just

don't think that they've had that experience. I will admit that it's very

experience. I will admit that it's very easy for me to say that on the founder side of the table who doesn't have to maintain optionality and doesn't have to maintain their reputation amongst immature founders who are going to

badmouth you no matter what you say.

>> Yeah.

>> So I actually know it's [ __ ] hard.

But I will say that despite how hard it is, some investors can do it. And I

think they're the [ __ ] legends. I've

met some of them and I'm just so impressed with them. like I would if I was raising money again for this business, another business, I would go straight back to those people because they're real.

>> Um, and I I'd love if the world kind of knew more about them. M

>> um and one of the in insidious and shitty parts of just the overall dynamic here is that there's always a new batch of founders that don't have the experience that I have that don't know

it can be a different way >> and they'll fall for the [ __ ] and they don't have the the standing and ability to tell investors to go [ __ ] themselves if they misbehave.

>> Yeah.

>> So the bad patterns perpetuate.

>> So should I tell someone to go [ __ ] themselves or But no, I'm just kidding.

Or do you >> No. Well, unless if there's a reason to.

>> No. Well, unless if there's a reason to.

Yeah. But what I'm saying is if >> like what are the various reasons you won't invest? Sometimes you just think

won't invest? Sometimes you just think you just don't see the opportunity there.

>> Yeah. This is like giving a data point.

It's just like >> Yeah. Like

>> Yeah. Like >> like here here's literally the conversation I would try to have.

>> And if you had 100 of these conversations, I guarantee it's going to resonate with all the right people.

They're going to come back to you. Yeah,

>> I would say um I'm trying this new form of feedback and communication with founders that basically no investors do because

they're all trying to maintain option value and it encourages them to [ __ ] founders >> and I really don't want to do that. And

I'll admit that I've I've I've been there a little bit because sometimes it just doesn't pay to be really real. But

I'm not going to do that anymore. Uh,

and I'm going to try and build a reputation that is based on authenticity.

And so here's what I think your business I think you're very compelling. I think

that you really believe in what you're doing and you're smart and I've enjoyed my conversations with you and I really hope that despite what I'm about to say, you'll talk to me again. I think that

the way you're currently configuring the business, it's just a little odd for what we've seen before. And that could mean that either I'm super wrong and

it's a incredible opportunity and I'm really close-minded and you're going to make a fortune without me. Or maybe if I'm lucky,

without me. Or maybe if I'm lucky, you'll come back in the next round.

>> Or it could it could possibly mean that you need to make a little adjustment and you'll figure this out. And I've seen it before with great founders like you who are slightly misconfigured, who go to

market in the wrong way, learn and change, have incredible success. And so

for me, who can only make a small number of investments, I have this limited size fund. These are

the check sizes are right.

You know, I I'm I'm really only going to make one investment every month. I can't

get comfortable when I think that that's the wrong thing. And either you're right or I'm right. doesn't really matter. But

that's where I'm at. And I hope that >> I hope that me being real with you will win some trust such that when you're raising again, you'll come back. And

maybe it won't. And I guarantee that the best people will be like, "Okay, I've actually never heard that before."

>> Yeah, that's pretty hard to argue with.

Yeah.

>> Like, thank you. Some people will be like, "Fuck you." Whatever. But they're

not the right people at all.

>> The the best people will be like, "I really like that authenticity."

>> Because if you talk about the other 80 90% They're they don't even get this might be the first time someone actually said that to them.

>> I guarantee if you do that, you'll be most times the only person that does that.

>> I guarantee it. Cuz the types of investors I meet now are way more experienced. They're later stage great

experienced. They're later stage great firms. So, they've learned to do this.

The early stage guys haven't even learned that. So, I think it's a secret

learned that. So, I think it's a secret weapon. Just radical transparency,

weapon. Just radical transparency, directness, authenticity. you could

directness, authenticity. you could build a whole brand around that.

>> Um, in my opinion, but again, easy for me to say from this set of table. Wish

you the best of luck with that.

>> Yeah, thank you. I mean, I think at the end of the day, that probably sounds easier than >> It's hard >> building an AI company from the ground up.

>> Well, no, I think it's all hard. I think

it's all hard.

>> Listen, your dollars are the exact same color as everyone else's, >> but somehow people have to choose your ones somehow. And there are now

ones somehow. And there are now countless idiots with green money.

Countless.

>> Yeah.

>> You and I could never list them. Chat

GPT could never list them. And so you really got to be careful when you're building that reputation. And so that conversation is easy to have when a smart ass is talking on a podcast.

>> But in real life, it's difficult. But

having said that, I I think it's an opportunity.

>> Interesting. Okay.

>> It's good to reflect on. I I do try to give as much feedback as I can, but to your point, it's it's hard to do it at scale like in a kind way everyone. Totally.

>> Things definitely do fall through the cracks.

>> Yeah, but that's for the legends of the legends.

>> Yeah, I like that though. It's good

motivation. It's good food for thought.

>> Yeah.

>> Well, it's been a lot of fun.

>> Yeah, really fun.

>> And I hope you had fun. A quick thanks to RAMP for supporting this episode.

Head to ramp.com/theappeal

for $250 on your first set of cards. If

you missed it, make sure to check out last week's episode with Albert Isoo from Level Ventures on how to actually generate alpha and venture capital. Tune

in next week for Dan Federer, who runs private market investing for the University of Michigan's endowment. If

you enjoyed this conversation, please like, comment, subscribe, name your next internal AI model after me. If you don't want to miss a future episode, subscribe to my newsletter, The Split, linked in the description to get each episode plus

transcript emailed directly to your inbox every week. Thanks again for listening. See you next time.

listening. See you next time.

[Music]