Rebuilding a Company for the AI Era

Shensi, welcome to the show.

Cool. Thanks for having me.

Yeah, thanks for, thanks for, uh, being on. Um, I think probably the coolest thing, most interesting thing that we can talk aboutright now: you sort of feel like Merge companies had multiple founding stories. Most founders don't get multiple founding stories.

So what's, what's the second founding story of Merge?

Yeah, um, we started in 2020, which was probably— I know everyone says, like, whenever they started was, like, the worst time, but yeah, every single year something bad and crazy has happened, um, since 2020.

Really?

Um, yeah, I know.

What have been the bad things?

Uh, like COVID. Um, we had a fun-based, like, on Zoom when, like, it was weird to think about now, but at the time, people weren't used to fundraising on Zoom. Um, and so when we would do calls, it was, like, very, very awkward, and people weren't quite used to, like, having that kind of interaction.

Now it's like, whatever, but, like, back then it was— everyone was like, oh, this is, like, weird. I think we were like one of the first fundraisers to happen over Zoom.

Really? So when was this?

Um, like, June 2020.

Okay, so what made it so awkward? Like, did VCs— like, was it you don't know who should talk, you don't know, like, how to build rapport with people?

I think, I think it was partially that. Zoom was still new. A lot of people we hadn't met yet in person.

Mm-hmm.

Um, and so our first interaction was, like, doing that.

Yeah.

Um, yeah, it was just, like, not— I think people felt like there was, like, a different human component that was really missing to Zoom. And, I mean, it is true, but now I think people are so used to it that it doesn't really impact things.

Yeah. And then 2021 was probably— there was just way too much money sloshing around.

Yeah, exactly. Like, everyone had so much money. Everyone was so competitive. It was, like, so aggressive. You had a ton of companies just, like, randomly buying and, like, changing ideas all the time.

Yeah.

Um, and then 2022, um, a lot of our customers were starting to die. Not yet, but, like, were starting to die. And then 2023, they just, like, all died to, like, all the companies that—

Really?

Yeah, that were, like, our initial customers, except for, like, a few. Like, obviously Ramp, I think they're okay. Um, and so—

Yeah, they seem to be doing pretty well.

I think they're doing okay, yeah. And, like, same with, like, a few of the other logos that we ended up closing back down, like Brex, um, Drata, Vanta. But, um, yeah, and so it was just a lot of, like, ups and downs and us having to figure out, like, okay, so what is— where is the industry going?

Like, where, like, what's happeningright now? We— in the beginning, when ChatGPT came out, we started having, like, a few AI companies that started onboarding, but none of them were really, like, production ready, and none of them were really getting a lot of traction or a lot of customers, just a lot of sign-ups.

Um, and so we— we didn't have enough conviction on, like, how— which direction we could go in to support these companies, because a lot of them just looked like SaaS companies with a little bit of AI on top.

Mm-hmm.

And so, um, but then I think, like, 2024, we started realizing, like, there was a really big opportunity, and the market might be a little bit too early, but the market needed integrations. Like, AI agents are really useless without context.

Mm-hmm.

Um, and at the time, I think people were really just thinking, like, they could do it themselves, and they just needed a few core integrations in order to have, like, an offering that was enough. Um, but yeah, like, 2020— end of 2024 and beginning of 2025, like, was really when all of it started hitting, which was really exciting for us.

So what was that like then when you're— you're basically making the decision of, like, we have to go all in on something that we didn't know was a thing when we first started the company? Like, I mean, it didn't exist when you first started it.

Yeah. Um, it was definitely scary. Like, I felt like an— I felt like it was really— I was really confident that it was a direction where the industry was moving in, but the— but the product roadmaps weren't reflecting it.

So I was kind of like, are we— when is this going to happen for us? Like, when are we going to start getting all the hyperscalers? Like, when are all the large, like, AI companies going to realize that they need some help with these integrations?

Um, and it did require a lot of, like, networking. We had a continued meeting with these companies, really, like, sharing why we were theright partner to work with.

Mm-hmm.

Um, and it— yeah, it was— it was a lot. We also had to actively invest in areas that weren't quite ready yet or mature enough and, um, take resourcing away from some other industries that were definitely slower growth or were probably not going to do super well over the next few years.

Um, that was very, very painful for us because I want, like, all of our customers to always love us, but especially when there's a potential market opportunity and you're the first to do it, you have to do it.

Others is really irresponsible.

Mm-hmm. What was the hardest part about making that transition?

Yeah, I think it was just finding the resourcing. Um, because we had a lot of customers that were very dependent on our product, and taking resources away from that product would be really de— like, it just felt like you were kind of, like, screwing your customers over.

Um, but we wanted to also launch, like, our own AI products at the same time while, like, adjusting our core product, um, to, like, serve a lot of AI use cases. And so, like, figuring out how we can make the— like, make the resourcing work, how to figure out, like, how to, like, move this engineer here for a few weeks, like, do that.

Mm-hmm.

Um, and then let alone, like, building an entirely new product that was, like, I don't even know how we'll find the resourcing for it. So what we ended up doing, which ended up actually becoming a blessing, was we had myself, my co-founder, and one engineer focused on building the new product.

Mm-hmm.

And so we were— we— that was it. Like, and, uh, my co-founder and I, we— we would only be coding at night, so it would be, like, like, from 5:00 to, like, you know, 11:00 PM. Like, after all of our meetings were over, we'd be coding.

Mm-hmm.

And then on weekends too. Um, but because we were building zero to one, we learned so much about AI and, like, what was possible.

Mm-hmm.

And when you're building, like, a ready existing product with AI, the— what you feel, like, is possible is just significantly lower than doing zero to one.

And it's probably helpful that you were actually doing it yourself because it's really easy to just see an article about AI and be like, "Allright, guys, let's do AI."

100%. Yeah. We— and we learned so much during that time period. Um, and it forced our team— because we knew so much and we got, like, fully AI-pilled.

Mm-hmm.

Um, it made us really force our entire team to really invest in it. Um, so I think if we hadn't done that, honestly, we probably wouldn't, like, have the kind of AI fluency that we have in our organization now.

Was it harder to sort of build this product and scale it the first time or the second time that you kind of went back and refreshed?

Oh my gosh. Okay, so the first time, um, we were in stealth for nine months before a single customer onboarded. Um, and we were doing research for, like, six months before, before that. So it— it took, you know, a long time.

Yeah.

Um, and our second product, we— from, like, day one to launch was around, like, six months, and the third product that we launched was three months. So AI is just getting better and better and better. And then also, I think it is the skill set for a company to learn how to launch new products as well.

Mm-hmm.

So we're also getting better at that too. Um, but yeah, like, it's— it's so good now, and it's so, so much easier to launch new things. Um, and the ROI of doing so is really, really high, but there is opportunity cost.

Like, you have to kind of, like, analyze, like, there's so many different ideas. Like, what do I end up doing?

This episode is brought to you by Numeral. Numeral is the fastest, easiest way to stay compliant with U.S. sales tax and global VAT. It's easy to set up, and they automatically handle all registrations, ongoing filings, and their API provides sales tax rates wherever you need them with all the integrations you need.

Their solution combines AI-driven automation with human expertise to manage global sales tax compliance end to end. Numeral supports over 3,000 customers, including companies like Brex and Character.AI, and they pride themselves on white-glove, high-touch customer service. Plus, they guarantee their work, and they'll cover the difference if they mess anything up.

If you want to get compliant, check out Numeral at their new domain, numeral.com. That's N-U-M-E-R-A-L dot com for the end-to-end platform for sales tax and VAT compliance.

This episode is brought to you by Flex. It's the AI-native private bank for business owners. I use Flex personally, and I love it because they use AI to underwrite the cash flow of your business, giving you a real credit line.

The best part is 60 days of float, double the industry standard. Flex has all the features you'd expect from a modern financial platform, like unlimited cards, expense management, bill pay that syncs with your credit line, and their new consumer card, Flex Elite.

Flex Elite is a brand-new, ramp-like experience for your personal life. A credit card with points, premium perks, concierge services, personal banking, cards and expense management for your family, net worth tracking across public and private assets, and a whole lot more fully integrated with your business spend.

One card for your businesses, one card for your personal life, one card for everything. To skip the waitlist, head to flex.one and use my code TURNER to get an additional 100,000 points worth $1,000 after spending your first $10,000 with Flex Elite.

That's flex.one and code TURNER for $1,000 on your first $10,000 of spend. Thank you, Flex, and now let's jump in. So I know you have a lot of customers that use it, like Netflix, Uber, OpenAI, Perplexity, uh, JP Morgan, BCG.

I'm probably missing some. I know you have a lot of customers, a lot of people that use it. So what is Merge? We're just one who just, like, five minutes in, like, what are these people talking about?

Yeah. Um, so we provide connective infrastructure for AI. Um, so we have connectors for products and agents. So if you want to connect to, like, G Suite, Microsoft Suite, um, Linear, all, like, productivity tools, um, instead of having to build these integrations one by one, we have synced connectors, and we also have live tool call— live tools, um, available, uh, via MCP.

And then we also have a smart LLM router as well. So now a lot of companies want to be able to have model sovereignty. They want to integrate with all the different models. Um, every day there's, like, a hot new model coming out.

Um, maintaining and connecting to all of them in-house is a lot of pain. Um, and so we end up having, like, this layer that makes it really easy for people to build next-generation AI products.

Is it really that hard? Because you just argue, like, say, "Hey, Claude, just go, like, do all this stuff for me." Like, why do you— why do you need to use something for integrations?

Oh my God. Like, we try doing that all the time with our own products. It's just not— it's not good enough. Um, there's always, like, edge cases, like.

So what's— what's so hard?

I mean, for connectors and integrations, like, there's just so many edge cases for what could go wrong. Um, there's also a lot of context that is just not online and not available anywhere. It's only available if you have industry knowledge, um, and if you are just, like— if you have, like, the data points from your customers for what could happen.

Mm-hmm.

And again, it's just not online anywhere. So an agent could try to do it on their own, but, like, unless you have, like, a real-world experience, it's just hard to replicate it. So one thing that is also very human about connectors is that increasingly a lot of access to sandboxes, which you need, um, which is basically like an instance where you can test against and start building your connector.

Yeah.

You can only get access to it from a partnership. And a lot of these partnership conversations require, like, meeting with someone, having, like, some kind of agreement, um, them knowing who you are and making sure that, like, you're not sketchy.

Um, that's not— an agent can't really do that.

Mm-hmm. So in a sense, then you are, like, that partnership's layer.

We are, yeah.

For your customers. So just, uh, we want to use this thing. We don't have to go talk to a million people and spend all this time. It's just use Merge, and it spins up really quick, and we get it, and we don't— we can get goingright away.

Yes. And especially when there's, like, API changes, there's new things that update, or, um, like, one customer has, like, a crazy instance that's a little bit different from everyone else. Uh, we are able to account for all those edge cases too.

The interesting thing is that you guys actually came across this originally because you were kind of building these integrations yourself, sort of, like, back you and Gill when you first started it. So I think you were getting dinner one night, and then what happened at dinner?

Yeah. Um, so we were getting dinner, and we were just talking about work. Um, and at the time, I was working in a cybersecurity company that needed a lot of ticketing integrations, and we actually didn't have any.

Mm-hmm.

Uh, we were losing a lot of deals to our competitors just purely based off of the number of integrations that we did not have and they did have.

Mm-hmm.

Um, so we ended up trying to avoid building these integrations as much as possible, having, like, CSV upload— or, sorry, download and upload, um, launching our own API, hiring contractors to build out these integrations.

Mm-hmm.

And it was just not successful. We ended up having to hire a full team of engineers purely to focus on these integrations.

Mm-hmm.

Um, but I have, like, an engineering background, and I was like, "This can't be like that hard. Like, come on. Like, you know, let Mama go after it." Like, I know this is going to be much, much better than you guys can.

Um, but my co-founder, Gill, he was head of engineering at a startup, um, that was in the recruiting space, so very different from cybersecurity.

Yep.

And, um, he was like, "No. Like, this is much, much harder than you expect because there's so much data that we have to sync. We can't even get started without getting partnership access or having an established, um, sand—

Mm-hmm.

Like, partnership and sandbox account from these different ATS providers. And, um, there's just a lot of edge cases that can go wrong that are not documented in the API docs. And so what ends up happening is there's customer support.

Um, customer success doesn't know what the answers are. They always have to get engineering involved. And so we built these integrations after a lot of effort and a lot of hard work and a lot of, like, partnership, um, time, but our engineers have to keep babysitting them, and we can't really move on.

Mm-hmm. So then that comes into, like, what we talked about a couple minutes ago where you almost have this, like, you're almost like a partnerships team.

Yeah.

In a sense.

We do. Yeah, we are.

So it's almost like you're, like, a forward-deployed partnerships team.

Yeah. It's— it's— it's actually one component of the product that was always, like, a nice benefit where, like, especially if you want to, like, immediately get started, um, you end up having to, like, go figure out, like, "Okay, how do I get, like, a client ID, client secret?

Like, where in the UI do I find this? Like, who do I reach out to?" Um, and originally with our unified API, it was, like, a— like, a helpful part, but, like, now, especially because AI is so good, it's increasingly a more and more important part.

Mm-hmm.

And a lot of these, um, SaaS platforms and, like, uh, I guess, like, sources of truth are increasingly becoming a little bit more protective of, like, who has access to their data and how so.

Yeah.

Um, so yeah, we've been able to help facilitate a lot of that too.

Oh. And then, um, how— how long did it take to get all this going? So you just said before it took nine months the first time you built it. Like, did you start coding on day one and it just takes forever to build, or was it, like, a research project?

Like, how long did all this take?

So we had six months of research before we, like, started writing any code. Um, and then.

What was— what was so hard about the research? Like, why did it take you six months?

Because we were back then— it was in San Francisco. I feel like this is, like, the olden days, but back then, Zoom— people didn't meet on Zoom. So you— we had to meet with people in San Francisco for coffee.

Each meeting had to be in person. Um, I also had to need to, like, find the people to meet with, so I had to, like, do cold outbound and, like, beg people to meet with me.

Mm-hmm.

A lot of these people were people we'd never met with before, and so we had to, like, ask them questions to just validate that it was a product that they'd be willing to spend money on in the future.

Um, and also, in retrospect, I'm like, "Wow, it was very generous of those people to spend this time with us because, like, it was just— we had nothing. Like, it's not even, like, we had a product that we could demo at the time."

Yeah.

Um, but yeah, it just took a lot of time, like, to make sure, like, we weren't going to be quitting our jobs that we really liked for something that was just going to flop.

Yeah. How did you convince people to talk to you? Because to your point, like, you were nobody, some, like, random.

Yeah.

You probably, like, didn't even have a startup yet.

Yeah.

You were, like, you had this idea. So what was, like, the general framework you used for reaching out to people, convincing them to talk to you? Like, what kind of questions did you ask?

I— I think we were— honestly, we probably could have done a better job. I just did a shit ton of outbound, um, like, ton of pipeline generation, like, being, like, "Please re— please meet with me. Please meet with me.

Please meet with me." Um, so it wasn't strategic and very bad compared to, like, what we end up doing now. Um, but yeah, it was just like, "Oh, we have this idea. Like, we'd love, like, your background's really interesting.

We'd love to get your feedback and advice on, like, what we're building. We'll meet you anywhere." Yeah.

That's probably, like, a secret that you have to go to them.

Oh, 100%.

Like, you just, like, go to your office. Like, the meeting room that is beside your desk, we will meet there.

Yeah.

Like, the minimum possible effort from you to do this, we'll do that for you.

100%. Like, there's no— and then after, oh, yeah, like, there's no way these people are already doing us a favor, and then we have to make it harder for them. Like, no way. I think that's something that, like, people don't really understand.

It's like, this person is actively doing you a favor, so you have to, like, make it easier for them.

Yeah.

Um, but yeah, if the meeting went well or we really had chemistry with them, then we'd ask for an introduction to someone else after.

Yeah.

And so that way we were able to expand our network.

Yeah. It's always like a— I don't know, I feel like a good cheat code almost. It's just, like, every time you're like, "Hey," like, I always ask when people come on the podcast, I'm like, you know, "Hey, this was great, whatever, blah, blah, blah.

Here's all the links published. By the way, anyone else you'd recommend to come on?"

Yeah.

I feel like that's a good way too because I think the guy I'm recording withright after this, um, Michael Tenenbaum, and Figure.

Oh, he's sick. I love him.

Yeah.

He's great.

So, um, I'd heard a couple people mention him just, like, over the years.

Yeah.

Like, "Oh, this guy's, like, the best non-founder, CEO, operator." And, um, I had Sam Blond at Monaco on, and he's like, "Oh, you got to get this guy on." I was like, "Huh. I've heard of him multiple times."

Like, yeah, that's the issue to get yes.

Yeah. You should ask him about, um, so we had, like, this event at, like, the Warriors game. Okay, so Michael, uh, Michael went to Columbia, and then, uh, Jeff Charles from Ramp also went to Columbia. And I am so stupid.

I, uh, we had, like, an event, um, so I try to not have, like, competitors, like, at the same event that we organize because, like, I've been there before, and it's, like, fucking weird. And, like, it just makes you— you can't, like, be as, like, relaxed as, like, when there's, like, a competitor there.

Yeah.

Um, and so I, yeah, we hosted, like, this, uh, Warriors versus Knicks, uh, basketball game, like, box suite, and I invited both of them. And I was like, "Fuck, like, I can't believe I messed this up." So I gave Mike— I gave both of them a heads-up, and Michael was like, "I don't care."

Because Michael was, like, about, you know, whatever. He was about to go to, um, Figure anyway.

Yeah.

Um, and, um, and Jeff was, like, so excited. Jeff was like, "Oh, I really want to meet him."

This is when they were Ramp and Brex?

Yeah.

Okay.

Um, and Michael was so fucking scary in this meeting with Jeff.

Who?

He was like, "Where are you from? Okay, I've heard about you." I'm like.

I've heard of Ramp before.

And it's so weird because Jeff is also scary as fuck,right? Like, so, but, like, I mean, Michael's, like, the GO.

Yeah.

Like, he cre— like, he's just, like, so he created the billboard, basically. He revived the billboard in, um, in San Francisco.

He created the— he revived the billboard. So it's just.

I'm sorry, not created. He revived. It is his fault.

His fault.

Yeah.

Okay.

But he, yeah, he's just brilliant.

Anything I should ask him?

Um, you should ask him about that interaction.

Interac— what happened when you went to the Warriors and Knicks game?

By— hosted by Merge, and you met, um, and you met the Ramp guys. He has a secret friend at Ramp. He told us.

A secret friend?

At Ramp.

Oh.

But he won't tell us who it is.

Oh, interesting.

Yeah, yeah, yeah.

It's almost like a spy. You guys had a spy one time.

When you can build anything, Amplitude lets you know how to build theright thing. Use human language to get complex answers about your products. No more manually selecting events or building charts or dashboards. Just ask. Use agents to sense changes in customer behavior, decide what's causing them, and ask you if it's okay to fix it continuously in the background while you work.

Get the answers you need while building directly in the tools you are already in, like Claude, Cursor, Lovable, and more. And for the first time, understand if your agents actually work. Measure quality, debug failures, experiment, and measure their ROI with agent analytics.

Amplitude. With AI analytics, all you have to do is ask.

This episode is brought to you by Merge, the connective infrastructure for production AI. The hardest part about building an agent is everything around it: connecting to the tools your team and customers rely on, letting agents take action with theright permissions, and keeping everything reliable and cost-efficient once you're in production.

Merge handles that all for you. It connects agents to thousands of tools, handles permissions and LLM routing, and lets teams move faster without building it all themselves. OpenAI, Dropbox, and Ramp all use Merge to move faster and build AIright.

Visit merge.dev/turner to start building for free. That's merge.dev/turner to try Merge for free.

Oh, yeah, we fucking did. Yeah.

What happened with the spy?

Um, oh, so I'm not 100% sure he was the spy, but I think he was. Uh, so, um, there was this guy he was interviewing at Merge, like, sick engineer, like, strong guesses all around. Like, everyone was like, "This guy is so good at coding.

Like, we have to hire him."

Yeah.

I was like, "Oh my gosh, I'm so excited." So I looked him up— I looked him up on LinkedIn, and I was like, "Oh my God, great resume." He worked at Loom. Um, and my friend Vinay is the founder of Loom, so I was like, "Okay, like, if he's good, like, I want to, like, just backchannel.

Maybe, like, Vinay can say something nice about, um, us and, like, you know, what we can do to, like, try to recruit him." So I texted Vinay. I was like, "Hey, like, do you know this guy? Like, I saw he was at Loom for, like, five years."

And Vinay was like, "I've never seen this guy in my life."

And he was— he's, like, the CTO.

Yeah.

He was, like, the head of all the engineers.

He was like, "This guy never worked at Loom." I was like, "Are you fucking kidding me?" And he was like, "Yeah." Like, "I've heard— I've gotten this backchannel before from one other person, and this guy is just straight-up lying."

And I was like, "Who is he?" And Vinay was like, "I don't know. I don't know." And so then we were like, "Should we confront him?" So our recruiter was like, "Hey, um, we're friends with, like, the Loom founders.

Like, are you— but they said they don't know you." And the guy was like, "Oh, that's weird. Like, he— they should know me. Like, I'll— I'll, like, you know, circle back with, like, another back— like, um, another reference."

Mm-hmm.

And then his LinkedIn was just deleted.

Wow.

And we— we don't know where he is.

So you're saying he was a foreign spy or maybe from, like, a different, like, a competitor?

No, I think it was a foreign spy.

Really?

Yeah.

Because you see those stories where they're from a different nation, state.

And I heard they're sick engineers too.

Yeah. And they do those— I've seen, like, the— they do, like, the Zoom interviews, but they, like, have all the answers kind of.

Yeah.

Like, they're, like, getting fed the answers to the questions and stuff like that. Was it that kind of situation or?

No, I think he was actually, like, a sick engineer.

Oh, wow.

Like, this was— I think this was before AI was, like, really, like, good for, like, you know, like, cheating and coding interviews.

Mm-hmm.

Um, yeah, he was actually a sick engineer. But we're in— we're in person, so it wouldn't be like he could join— like, try to join us by being, like, a remote.

Mm-hmm.

With, like, a fake face thing. Um.

That's fair. Yeah.

But.

But you guys are really big on in person in the office,right?

We are, yeah.

Is that a reason because of the spy? Like, we can't do remote anymore?

Well, it's a nice, um, it's a nice positive part, like, is that it's a little bit less likely. Or if they are, they're, you know, yeah, I don't know. It's just a little bit less likely. Um, but yeah, it's just— it's really fun.

And the team is a lot closer. It helps a lot with just, like, communication and being creative. Um, yeah. We also have, like, a lot of couples that have formed from experience at Merge, so.

Really?

Yeah.

How many?

I think, like, seven or eight.

Holy shit. How many employees?

Like, 140.

So if you join Merge, there's, like, a 6% chance you'll get married?

I guess so.

Are these marriages or just dating?

Dating, but very serious.

Okay.

Yeah. Also, a lot of times, it's, like, really, like, high performers dating each other because you don't want to date, like, a fucking loser. You want to date, like, somebody who's really good.

Yeah, that's true.

Yeah.

I mean, I feel like that's— that's one of the, like, the hard parts of, you know, like, modern dating or whatever, where you've got, like, the dating apps. I'm married, so, like, I've actually— I've never really used the dating apps.

So, but, like, you get matched with someone, and, like, it all seems good, and then your— your interests and, like, your ambition, like, everything is different.

Mm-hmm.

That'd be tough. But yeah, I guess, like, you show up and, like, you're both at Merge. You're obviously, like, great.

Yeah, I filtered.

Yeah.

Yeah.

Great filtered. Yeah. Interesting, huh?

And you know, like, if they're good at their job.

That's true.

And you're also, like, friends with the same people. I'm not, like, you know, encouraging workplace, you know, but, like, it's just, like, and it does, um, yeah, it makes me happy that, like, you know, we had, like, a real impact on them.

Like, obviously, in their career, but, like, it's— I don't know. It's nice. Yeah.

I mean, that's a— that's potentially a really good pitch to work at Merge.

My husband.

Like, you might just say anything else.

My husband said the same thing, and I'm just like, "I don't know. I feel like it's, like, assumable." Or it's, like, scary. Um.

I don't know.

Actually, one of my team members— okay, this is really bad, but she was just like, "You need to— you don't need to hire, like, some hot people for me because I'm single and, like,right now, the success rate, like, the probability of someone finding their future husband at this company is, like, pretty high, so you need to hire some hot people for me."

Like, she just started hiring only hot people.

Yeah. Only hot people for me. I was like, "Okay. I'll try." Can't be a filter, though.

Okay, so aside from being really hot, how do I get a job at Merge? Like, what— what do you usually look for when you're hiring people?

Yeah. People have to be, like, really enthusiastic. They have to care a lot. Um, they also have to be, like, really, like, smart or, like, you know, want to do a good job.

Yeah.

Um, but yeah, I think, like, cohesive, like, part of our culture is, like, people care a lot. People are really nice.

Mm-hmm.

Um, yeah, and people are, like, enthusiastic. Like, I think the worst part of, like, going to a job or, like, talking to someone is if they just are, like,ugh, and, like, it's super draining, and, like.

Yeah.

After talking to them, you're just like, "Wow, I'm so tired," versus, like, meeting with someone you're, like, really energized and, like, you feel like, you know, you're, like, excited.

So what— what do you usually look for when you're— when you're interviewing? Are there, like, certain patterns of, like, people who act that way?

Yeah. I think people who complain, like, the entire interview and, like.

Mm-hmm.

Things happen to them versus, like, them being in control of their life. It's, like, very palpable. Um, because also, like, if they complain about every single job they've ever had, they're going to complain about you. Like, you're not going to be, like, an exception.

They're just not going to be happy. Um, so that's something that I do look for a lot too. Um, like, I think it's totally fine to have, like, a job that you didn't like, but I think it's, like, how you talk about it.

Um, that's really important. And sometimes, like, I've heard people be like, "Oh, like, that— all those people were just, like, so bad, like, blah, blah. I hate all of them." I'm just like, "Whoa." You know?

Yeah. I mean, I had my very first job. I worked in a factory.

Whoa.

It started at 6:00 a.m.

Yeah.

It's like the shittiest job ever.

Yeah.

Honestly. But— but it did teach me, "You got to wake up at 5:00 a.m. and go to work and stand on it."

Now you're just like, "This is so easy compared to that."

Yeah. Now I just, like, tweet all day and, like, record podcasts all day.

It's like men used to work in a factory. Now they just move their fingers.

Now they just podcast.

Yeah.

Society. That's why— that's the downfall of society.

Yeah.

That's why.

Yeah.

I need to go back and work in the factories again.

Yeah, yeah, yeah.

Um, I— but it actually did make me realize, like, "Wow, I need to take college seriously." Like, I need to, like.

Yeah.

Get a real job because.

Yeah.

Because it was— it was basically like a, uh, freshman year after college.

Oh, okay.

It was just, like, because a friend's mom, like, worked at this factory, and I needed a job for the summer because I needed to make money.

Yeah.

And I was like, "Eh, can't be too bad. I've never had a job before, but, like."

Yeah.

"I know I'm going to try hard, and I do well."

Yeah.

"Sure. Give me— give me a shot." And then I was like, "Wow, this sucks."

That's great.

Like, it sucks with your retirement.

Well, what was your hourly wage?

I think you got $9 an hour.

Oh, that's pretty good for factory.

Yeah. Yeah. And then you got, uh, overtime if you worked on the weekend.

Okay.

For, like, so I just always did overtime when they offered it.

Yeah.

And yeah, so I don't know. I— I can't remember exactly. I think I was getting, like, $900 a week or something like that. Like, a thousand dollars.

That's pretty good for, like, a college freshman. Yeah.

Yeah. It was pretty good.

Yeah.

Um.

I— I didn't work at a factory. I worked at, like, a Froyo store where I was getting, like, verbally abused by this, like, Korean lady. Um, but.

A coworker or a customer?

It was just a coworker.

Okay.

But I was, like, in high school, and, um, it was, like, $8 an hour. And I, um, but every time I, like, I— I love mak— I was like, "Oh my God, this is so cool. I'm, like, making money."

Yeah.

But, like, standing, um, was really, like, standing for that long was something that, like, really I— I, like, like, really, like, came into, like, my memory forever. I was like, "Wow, I, like, love sitting."

Yeah.

Um, and so, um, now, like, I— when I have a desk job, I'm just like, "This is, like, fucking awesome." Like, I love.

Like air conditioning?

Yeah. I love sitting. I love, like, not burning any calories. Like, I just, like, moving my fingers. Like, it's.

Yeah.

It's just, like, I'm so— I'm so lucky. And I like it. And I like it.

Yeah.

Yeah.

And— and then I— I kind of went on this, like, I don't know, like, arc where I was, like, uh, there was, like, a job portal at school where you could, like, apply for internships.

Yeah.

And I was like, "Oh my God, I can make, like, $10 an hour, $12 an hour." So I just— I just kept doing more internships every semester.

Yeah.

I'd get a new one where I made more money to the point of, I think my last semester I was making, like, $22 an hour.

Wow.

Which is.

Wow.

I mean, yeah.

That's good. Yeah.

Like, a 21-year-old, 22-year-old kid.

Yeah.

And then— and then this was obviously— it was— this was in West Michigan, in Grand Rapids.

Mm-hmm.

So there's not, like, a Google where you can make, like, hundreds of dollars an hour as an intern.

Mm-hmm.

But it was, like, pretty awesome. Like, I was making more than my parents were making.

That's cool.

In college, which was, like.

That's really exciting. Yeah.

It was kind of crazy, though, because you, like, I don't know. You— you— like, it's like the most amount of money you've ever had. Like, I— I was getting paid more money every week than my mom had, like, ever had in her bank account, like, at any point in life.

Yeah.

Um, it's crazy. Well, on this vein of, um, sitting in the office, air conditioning, a lot of founders hire an EA or multiple. I actually know someone I met. They have four EAs.

How do you— how do you feel about having an EA? You don't have an EA. This episode is brought to you by Monaco. Monaco is the first revenue engine built specifically for startups. Monaco's AI-native platform replaces your legacy CRM and sales point solutions.

It has everything startups need all in one place. Monaco helps build your TAM, generate demand, run outbound, capture every interaction, manage pipeline, and automate follow-ups all in one tool. They pride themselves on an effortless onboarding, white-glove activation, and get you to value in days, not months.

And the product practically runs itself with built-in agents that are always working for you. Start growing your revenue faster with Monaco. Try it now at monaco.com.

I don't. Yeah. Um, okay, so I used to be a chief of staff, and so I have, like— and I used to manage an EA. So I have a lot of experience, like, understanding what the workflows of an EA and a chief of staff are.

Yeah.

Um, I think if you have, like, a big company and you are very successful and you're very busy and traveling a lot, fuck yeah. Get that EA. Get that chief of staff.

Yeah.

If your company is, like, 10 or 20 people, and this is different for VCs, by the way. VCs have, like, a very busy schedule where they're traveling a lot and, like, have a lot of meetings. But a lot of founders are like, "You're coding.

Like, you should be meeting with customers." Like, there's, like, Calendly, and there's, like, Google Calendar. You don't really need an EA. It's a— it's kind of like a vanity thing. It also slows down, like, the time for booking, like, a meeting with a customer when you're so small and, like, you really need to just get that on your calendar immediately.

Mm-hmm.

Um, and also, it just means you're, like, not really fast at your computer. Um, like, you should be able to learn how to, like, respond to emails fast. You should learn how to, like, respond to your team and, like, book things on your calendar and, like, book travel.

Like, it's a good skill set to know how to be fast at your computer. Um, it's definitely a hot take, but yeah, like, if you're, like, really fast at your computer, like, you can go for a long time without an EA.

Like, EAs are very— like, good EAs are a huge game changer. Totally agree. But, like, a lot of these people that I'm seeing with EAs and chief of staffs and, like, some kind of, like, fucking founders' associate at the same time, like, "What are you— what are you doing then?"

Yeah.

Like, "What— what are you— like, are you really that important and that busy where you need someone to do all these things? Or are you just, like, really fucking slow at your computer?"

Yeah. Is— is there— does there need to be, like, a formality, though? Like, let's say I'm— I'm a customer.

Yeah.

I email you or something or whatever the interaction is, and you say you need to be really fast. Like, do you— what if you just, like, lowercase, not a full sentence, just, like, "Yes." Like, what's— what's this, like, spectrum of you need to make sure you're, like, being fast and quick at people, but then, like, the formalities of, like, making sure you're presenting things?

Like, is that part of the EA? Like, it, like, makes you seem more legit, maybe, or?

I don't think so. Because also, I don't think so. Because, like, if I'm respo— so I've also— again, I've interacted with many different EAs.

Yeah.

If the EA is, like, really bad, it makes you look bad.

Yeah. Like, they don't respond for a day.

Yeah. Or if the EA is, like, a total bitch and, like— or, like, you said— sorry. Or, like, you said, like— you said, like, your availability, and they just, like, totally ignore it, and they said, like, "It's another thing."

It's just, like.

Yeah.

It creates friction. And especially if you're in a situation where you, like, really want to meet with this customer or, like, you really want to meet with this, like, important partner.

Yeah.

And you're— like, you have not hired well or, like, your EA is not, like, representing you well, that can really hurt the relationship.

Yeah.

Um, and so, yeah, it's just better to have that in your control. It's another mental load, like, that you end up having to, like.

Yeah.

Um, hold onto and, like, think about continuously.

I mean, I think a lot about it too is, like, I— I try to actively not schedule a lot of meetings and not schedule a lot of things. So I feel like when you outsource it to someone else, there's, like, less sacredness around it versus for me, like, if I'm going to spend time talking to you to, like, set up a meeting, like, I really want to talk to you.

Mm-hmm.

And, like, I— I don't do, you know, 10 hours back-to-back meetings all day. I guess maybe I'm in a position where I don't have to, but I'm— I don't really want to do that. But, like.

But you know what you can do with an EA, though? You can have the EA blow someone off and make it seem like it's not you.

That is true.

That's something that, like, yeah.

I guess it's a shield. Yeah.

It's a shield.

Interesting. I mean, that's why a lot of VCs hire associates.

Yeah.

Because you can just be like, "Oh, man, I'm traveling. I'm just slammed, like, here's Shensi on my team." Like.

Yeah.

"Finding time to chat."

Yeah. Oh, another thing about— about another thing about, like, needing an EA, it's like if you don't know shortcuts, like, on your laptop, like, you don't know how to, like, go to the beginning of a sentence or, like, switch tabs without, like, touching your mouse, you don't deserve AI or an EA.

Like, at minimum, you should know how to go between Slack channels with a short— like, with shortcuts.

Oh, yeah. That's fair. So actually, how do you— I don't actually know. I don't use Slack a ton, so I have never had to learn the shortcuts.

It's— it's just, like, using, like, the shortcuts for, like, Chrome and, like, Move.

Like— like it's like Control J or something.

Yeah, exactly.

Yeah.

Control K.

I mean, Superhuman is a VC.

Exactly.

I know all the shortcuts.

Yes. I think Superhuman has done— has added so much value for the people who never would have learned shortcuts, like, ever.

Oh, yeah.

Um, but yeah.

Well, it's like the whole— because you did investment banking?

I did. And not to brag, but, you know, I'm on the Trading the Street um leaderboard.

Are you really?

For formatting.

Okay.

Yeah.

And so for people who don't know, there's, like, Olympic-level competitions of, like, how fast you can build an Excel model.

Format and get it.

Without using your mouse.

Yeah.

And you can get really fast.

Yeah.

And so you're basically saying, like, using AI is just like using shortcuts.

I think so. Yeah.

Like, um.

Well, obviously, like, to a different level, but I just think.

Yeah.

I just think that, like, shortcuts are a mentality.

Yeah.

Like, if you are using shortcuts, your mentality is, like, "I need to be, like, really productive, and I need to be really fast at, like, whatever I'm doing."

Yeah.

Um, so obviously, then it extends to, like, AI. Like, I just expect people who are good at shortcuts are probably thinking, like, "Okay, how do I use AI?"

Hmm. Yeah, that's fair.

Yeah.

I mean, that's smart. So then personally, how do you use AI maybe, like, uh, inside of Merge?

Yeah.

Like, what kind of things have you done that may— maybe these are super obvious. Maybe they're not that obvious. It's unique to you guys. And then personally, like, how do— how do you leverage it to— to get more done?

So I don't have anything personal outside of Merge. That was, like, a stupid question. Um.

I mean, like, as a CEO.

I joke. I joke. I joke.

I mean, there's some people that's like, you know, they're like, "Oh, like, where should I travel? Like, we're going to have to think, like, what Netflix should I watch with AI?"

It's like level one. Yeah. Um, yeah. So for— okay. So I mean, it's basically just, like, anything that I see at the company that I'm just like, "Why is this taking so long?" Well, I'll just be like, "We— I'll just build it," or I'll have someone else build it, or I'll, like, give a very detailed prompt on exactly what to do.

Um, so yeah, one example is, like, so we had these, like, dinners that we used to organize for our customers and partners.

Yeah.

Um, and we would do, like, one or two a quarter because it was just such a pain in the ass to coordinate. Like, it would require, like, our whole team to go through LinkedIn, like, find who to invite.

They would invite them, and they would have to update the sheet. And then we'd be like, "Oh my God, the dinner's not full." Everyone would freak out.

Yeah. You got to, like, organize the dinner too.

Yeah.

Like, communicate with restaurants of, like, availability.

Yes.

Like, that's also a list.

And the seating chart. Like, there's just so many things, like, for the dinner that was just like, "Okay, we can only do, like, one or two a quarter because this is just too much time." And what would always happen because we couldn't fill the dinner is that— and I didn't want to be, like, embarrassed.

It's like, I would end up spending, like, all my time, like, the— through two to three weeks before, like, going through my LinkedIn, like, "Who's in New York? Like, who's in San Francisco? Like, who have we talked to recently?"

And, like, inviting them manually.

Yeah.

Um, and so I— I hated that process because it was something that I always needed to get involved in, and we just, like, couldn't scale it. But the ROI of them was so high. Like, people love these dinners, and we would have such a good time.

And, like, I love seeing our customers.

Yeah.

And so, um, one day I was just like, "You know what? I'm just going to spend, like, two days. I'm going to build, like, a dinner bot." And so now what this bot does, it doesn't do everything. Like, the taste of the restaurant, like, it can't do that.

Like, it can't decide, like, what's a good restaurant. Like. You know what I mean? Like.

Like McDonald's.

Yeah, yeah.

Chipotle dinner.

Exactly. Or, like, a, like, fucking, like, $30,000 dinner. Like, no.

Yeah.

Right? So, um, it— like, so— but it will serve, like, here's the top restaurants, like, maybe do an outreach.

Yeah.

Triple by— uh, Triple Seat, I think, like, doesn't have an API either, so we can't automate that part. So I do fill out a form. But whatever. Once the restaurant is picked, it does, like, the calendar invite. It, like, looks through Gong for every single person we've met with before.

It goes through, like, our Salesforce and Apollo, goes through our customers and contacts.

Yeah.

At those, um, and, like, who's in the city. And it just sends invites, like, as me. And then when they respond, it syncs the updates and responses. And then before the dinner sends the reminders, the day of the dinner, it reminds my sales team to, like, send notes to the people who are going so they were, like, know what to talk about.

Yeah.

Does a seating chart. And so we went from having, like, one or two dinners a quarter to now I think we have, like, three to five a month. Um.

Cool.

Yeah. And it's awesome. Like.

So it kind of maxed your dinner, uh, throughput.

Yeah. For sure. And it's been so helpful for us.

Yeah.

So that was something where, like, I just, like, decided to build it out to, like, help the team out. And it has, like, had a pretty material impact. But yeah, I think, like, me having that experience, like, building my own agents and, like, knowing when I can tell people, like, what to do.

Mm-hmm.

And if someone's like, "Oh, this is too hard," I'm like, "It's just not." Like, "I— I know what you need to do here." Um, versus if I didn't know, then I'd be like, "Okay." Then it would just be the end of the conversation.

Yeah. Uh, I think it's like a good framework where you're just like, um, you can use AI to do anything that— that takes a long time. And the thing I always do is any— like, I use AI a lot for the podcast, like, for all, like, the prepping and then all the post-production stuff.

And anytime I change something, I always just say, like, "Hey, I changed this. Just, like, do it theright way next time." Like, figure out what I did different.

Yeah.

And just fix it. And I mean, that's a pretty— so I feel like what do they call it now? Looping? They call it.

Yeah.

Product. Like, if they're.

Yeah.

Using AI to do loops.

Yeah.

To, like, build everything.

Pretty good. Yeah.

Yeah. Do you guys do that?

Yeah, a lot. Yeah. It's like.

And what's your extent? Like, are you— do you do any AI coding.

Oh.

Still?

I only AI code.

You only AI code.

Like, it's the point where, like, I'm on an airplane. This is really bad, but it's the point where, like, I'm on an airplane. There's no Wi-Fi. I'm like, "Fuck. I don't— I don't know what to do."

I had no Wi-Fi on my flight here today.

Oh, really?

I just.

Whoa.

Set.

What'd you do?

There's some music. I— well, I played— um, I have a game on my phone called Slay. It's kind of like Risk.

Oh.

It's just like a.

Called Slay?

It's called Slay.

That's so je.

It's a game from, like, the early '90s that I played as a kid.

Yeah.

That I found on my phone. But, like.

You can get them sponsored the podcast if you, like, love playing this game.

Uh, it's a free game.

Oh, it's free?

And I don't know if you can even pay for it.

Oh. How long is the flight?

Uh, it's like a— it's just like an hour.

Oh, from Michigan?

From Detroit, yeah.

Oh.

I think it was, like, maybe an hour and 15.

Oh, really?

Yeah.

I didn't know it was that close.

It just depends. Yeah. Sometimes it's closer to two. Sometimes it's closer to.

Oh.

One and a little bit. Yeah.

Okay. Well, welcome.

Yeah. Thank you. Thanks for having me here.

Well, wait. Are you in Western Michigan? So do you.

Uh, it's— it's Detroit, East Michigan. I live in Ann Arbor.

Oh, I thought you were living in Western Michigan. Okay. Okay.

Yeah. I grew up in West Michigan.

Oh, okay. Okay.

And then we moved to Ann Arbor, uh,right before COVID. It was really good timing.

Okay.

Um, because yeah, my in-laws are there, and then with kids, it's like.

Oh, nice.

You know, that's— that's why we live in Ann Arbor.

Are you going to make your kids go to University of Michigan, Ann Arbor?

Um, I'm actually not convinced I'm going to make my kids go to college.

Yeah.

I'm just like.

I think it's going to be, like, not. Yeah.

Yeah. I mean, everyone has different opinions on this, but I'm just like, I don't know if they're open.

That's very San Francisco of you.

Of being open to them not going.

Yeah.

Yeah. Well, I don't know. I mean, they're— my oldest daughter's more into, like, art and dancing and, like, singing. So I don't know. She might want to go to, like, a.

Dance school. Yeah.

Art dance.

Yeah.

School. I don't know. She also m-might want— she also, like, really good at math. She really likes reading.

Mm-hmm.

So she might be like— she might be an English major. Maybe she gets, like, a PhD.

Yeah, yeah.

Right? So I don't know. I'm just— I'm of the view of, like, I'm not going to force you to do something because that sucks.

Yeah, yeah.

I don't want to force my kids to do things that I want to do. So one thing I want to talk to you about, um, when you think of, like, marketing as a— as a brand, as a product, as a founder.

Mm-hmm.

A company, what's been your approach with— with Merge? Like, how it's— how's it evolved over time?

So I think marketing— I think every function has gone through, like, an AI transformation. And I think this year, marketing went through its own.

Really?

Like, a really, really dramatic change. Um, it became really clear that marketing was shifting a lot, especially, like, the past few months where, like, there's all these, like, taste AI companies, like, all these design AI companies. And it's because I think, like, the— like, code— AI coding got so good that the bottleneck for pretty much most companies was just— was marketing because.

Really?

Like, you'd end up building so many features, and you wouldn't be able to market them at the same pace because there weren't a lot of tools ready for that. Um, and so we've basically changed marketing to become, like, an engineering function.

Um, so all of our team members have to, like, build agents for everything. Like, the dinner bot, that's, like, one thing,right? Like, field marketing should— everything that's really recurring and everything that kind of sucks, we should be using AI for.

Yeah.

Um, a lot of our ads, like, we generate with— them with AI. A lot— like, a lot of our, um, we have, like, a recurring— like, recurring posts or, like, content that we generate based off of, like, things that we produce in our product, like, you know, and it just pulls straight from our GitHub and then generates the social posts.

And we obviously, like, eyeball them, but it just releases them too.

Okay.

Um, all of, like, the data analysis for, like, demand gen of, like, our different ads, like, channels, all of that also now is, like, fully we use AI. Uh, we have, like, a calendar, like, all that's, like, auto— like, everything that goes into the calendar is automated too.

And so.

Yeah.

Yeah, we've gone through a pretty big transformation over, like, the past, like, two months. Um, and it— it's— yeah, it's been really exciting.

Is it working? Or is it, like, too early to know or?

I think so.

Interesting.

It's hard to tell because it's a lot— it's definitely lagging.

Interesting. So then what are you looking at as these, like, leading indicators of, you know, AI native marketing machine?

Yeah. So, like, sign-ups, website visits, um, conversion, obviously, like, early revenue or, like, POCs and, like, opportunities that have been open in, like, Salesforce.

Yeah.

Um, so that's been, like, the leading indicators. But yeah, it's still only been, like, one or two months, so we'll see. Um, but definitely, like, I've been seeing, like, a lot more activity. And activity always— not always, but, like, hopefully ends up leading to more, um.

Hmm.

Other things. Yeah.

You mentioned something, uh, customer sign-ups. So I know in the early days of Merge, you were kind of more SMB-focused.

Yeah.

Like, originally. So why did you start with SMB?

We started with SMB because we were infrastructure, and no one bigger wanted to take a chance on us. Um, it was very scary. Like, we could have died, like, any minute. Um, and they would have been screwed. And also, um, if we had, like, a security incident or any— or, like, downtime— it was just— it's so many risks with being an infrastructure company.

And so for us, like, we really had to climb the logo ladder over time and, like, move to bigger and bigger companies.

Mm-hmm.

Um, as we got bigger— I'm sorry. As we, like, um.

Product?

Yeah. As our product matured. Um, and then around three years ago, we made a very, very concerted effort to move enterprise. So that meant, like, a complete, like, change of who our team was, how our team was segmented, how we approached our marketing.

Um, like, how we didn't— we didn't use to have, like, decks for, like, sales meetings. And then we ended up, like, having decks. And, like, we had a CRO who was really good at enterprise mar— yeah, enterprise sales.

And, like, that helped transform our company too. Um, so yeah, it was a pretty big transformation. And but it took— it was— it took a, you know, a few years. Uh, but last year really, like, was the year when, like, everything came in.

Like, we had all those logos that you mentioned, um, like, coming to us. And it was just really exciting to see, like, all the prep and, like, the work that we'd put in for the past few years. Like, it was really finally coming in.

So then what was kind of the evolution of the product then? Because when you first started, it was literally, like, accounting and HR.

Yeah.

Integrations with accounting and HR SaaS tools. And this was back in 2020.

Yeah.

So how did— how did that kind of evolve over time? And how did you de— decide what to do?

Yeah. Uh, so the goal is always to be cross-category. We just picked one or two categories in the beginning, um, as, like, a wedge into the market and to introduce ourselves.

Mm-hmm.

Um, but we picked them based off of, like, what would overlap, what had big, like, TAM opportunities. Um, and especially as the market continued, like, changing. So, like, fin— at financial services, like, fintech was getting really hot. Like, when we launched our accounting API, then we started noticing a lot of, like, um, like, sales AI and, like, uh, productivity, like, tools, like, starting to form.

So we invested in, like, ticketing and CRM integrations. Um, but yeah, we always just kind of followed, like, where the market was going. When AI— uh, when we saw AI search was going to be, like, the next frontier of what, like, really needed integrations, we started focusing a lot on productivity and, like, sources of truth, like knowledge-based chat.

Um, yeah, like, ticketing, file storage. Um, so yeah, it's always been, like, TAM-driven, like, customers. I think that's, like, the nice thing about B2B. It's a very— it's very, like, customers tell you what they want, and you just really have to listen to them.

Versus consumer, you have to be kind of, like, you know, philosophical. Like, I think this is what the consumers want. Like, the taste.

Yeah.

The discourse.

It's hard.

Versus B2B, it's just like, yeah, there's, like, a billion in pipeline that wants this.

Yeah.

Like, build the thing.

Yeah. I've always— like, I've always really looked up to, like, consumer founders because I feel like it's— it's much harder. Um, so.

Yeah. Uh, and then so the kind of, like, the evolution of the AI product. So when the very— what was the very first AI feature that you initially launched for Merge?

Um, so there were— so there was, like, a product that was for AI companies. And then we had, like— we had our first, like, AI products that were just, like, very AI— very AI.

Yeah.

Um, but yeah, like, our unified API, we started noticing, like, for AI search, like, we needed to have file storage integrations. We really good to, like, Google Drive, Dropbox, Box. Um, those platforms have so much data that people want to be able to search across.

Mm-hmm.

Um, and also, like, having knowledge-based, like, Confluence, Notion, um, chat, like, Slack, uh, Teams. Like, a lot of companies need those integrations, especially as you, like, want to make it very easy for people to do, like, semantic search in your product.

And no one is really able to do that. Um, so we felt like it was a unique wedge for us. And so we started going all-in on that, like, early last year. Um, and then— but at the same time, we were like, okay, we need to be very aggressive about, like, this— this move into becoming, like, an AI company.

So we were going to launch our first AI product, which was Merge Agent Handler. So these were, like, tools for agents.

Mm-hmm.

Um, so being— so being able to, like, make tool calls, um, live tool calls instead of, like, our synced integrations was, like, kind of a slight adjustment from our first product, but a completely different product. It required us to, like, really build something zero to one, do completely different marketing, um, learn, like, how to sell this new product, have different pricing.

Um, yeah, it was— it was a lot of— it was a lot of work.

Different pricing is scary because if I'm a public company CEO listening to this, I think of, like, different revenue structure and, like, different, you know, distribution strategy and, like, sales. Like, basically, like, you have to, like, completely reinvent the company when you change the pricing.

So how did you approach new pricing model?

Yeah. So can you— this wasn't, like, a full dramatic shift. Like, the first product, we kept the pricing consistent.

Yeah.

Um, the second product, we layered on, like, a different pricing model, and we had to, like, train our sales team for how to talk about it.

Mm-hmm.

And then when we did our third product, it was a different pricing model again. So we didn't shift the first two. We will go get to a point where we'll have to consolidate all three because it's just— it's too confusing for, like, our sales team.

Mm-hmm.

Um, we'll also probably have to, like, repackage what, like, the products look like so it's a little bit easier to talk about. Um, but yeah, like, it— it's been a— it's been a really big shift. And I think another thing that was really interesting for us to observe.

So Gil, my co-founder and I were, like, we're startup people. Like, we remember what it was like when it was very early to, like, get something from zero to one.

Yeah.

Um, and our team— or, like, a lot of our team members had joined when, like, Merge already, like, that first product was pretty established. We had customers. We have, like— we're thinking about scale. Um, and so they tried to apply a lot of, like, the things that were working for that first product to our second and third product.

And so we really had to, like, tell the team, like, we got to, like, move out of this. Like, we can't think about things that will scale. You guys are protect— trying to protect us from, like, the downside.

That's probably not going to happen.

Mm-hmm.

Um, because, like, we'll be very, very lucky if this happens. Like, we'll be so lucky if we have, like, you know.

Someone uses it in the first place. Yeah.

Yeah, exactly. And so that was something we really had to, like, iterate to the team. Like, let— don't, like, don't be so focused on that. We need to just, like, move really fast so that we can get those first few customers.

Um, and so they— luckily, they were very understanding. We were able to do pretty well.

Yeah.

Yeah.

And, um, I think the kind of, like, the second then is the Agent Handler product.

Yes.

So this is basically, like, it changes which model the.

Oh, that's Gateway. That's the third product.

This is Gateway.

Third product. Yeah.

Third one. So well, I— I mean, I think that's an interesting one too where you— you basically route the AI direction flow through the cheapest model or the most effective, highest ROI model. So.

So the beauty it's whatever you want.

Okay.

Yeah.

So— so then, like, why is that so important? Because shouldn't you just, like, have one vendor that you— like, one that you use and, like, you always get the same result? Like, why is it so important to be able to switch between different models?

So I think, um, before, when pe— because AI was still getting so much better at every single, like, release. Um, and whenever there was, like, a new model release, like, you were kind of like, wow, like, the last version was, like, totally shit compared to this one.

Like, why would I ever use the last one?

Yeah.

Um, like, the new one actually works.

Yeah.

Um, and so there— we had, like, a couple years where that was really, like, the case. And so every single time there was a new model release, everyone would just default go to the latest model.

Yeah.

And it would just be, like, a one-size-fits-all. But I think now, um, there's so many new model releases. Like, each frontier, like, lab is releasing models, like, every month and a half or so instead of, like, every six months.

Mm-hmm.

Um, and there's also a lot of options. There's a lot of open-source options. There's a lot of, um, there's a lot of, like, cheaper options also within, like, the same, like, within the same lab.

Yeah.

And the costs are really starting to become more and more prominent. Um, Fable coming out made it really at the forefront of how expensive this could be and how companies really need to start focusing on their optimizing their AI spend.

But I think we're seeing, like, Coinbase, DoorDash, they're very advanced companies thinking a lot about, like, how to optimize their internal, like, engineering spend and, like, how to make sure they're not, like, overspending on, um, certain workflows. But for a lot of, like, smaller companies and also companies that are not that advanced, having that kind of functionality and focus internally is just really, really hard.

Um, and it also even if it's, like, embedded in your product, like the AI optimizations, it takes a lot of resources. And it could be, like, a full team fully focused on, like, how do I route? Like, when do I make sure one plus one doesn't go to, like, Fable?

How do I make sure, like, hello, doesn't end up costing us, like, 3 cents? Um, those are things that, like, a lot of people need to focus on in-house.

Mm-hmm.

But it's what everyone's building over and over and over again. And so it was a natural extension of our existing products. It is a unified API. It's one API to route to all the different model providers.

Do you know how many different model providers or model options there areright now? Like, is it.

I think we have hundreds.

Oh, inside of Merge?

Yeah, inside of Merge.

What about, like, total that you could use total?

Oh, I don't know. I don't know.

Like, there's also.

There's also complexities because it's not even just, like, well, okay, well, people also have in-house models that they're probably not disclosing to.

Oh, okay.

So it's in— it's infinite. Like, people are starting to do post-training.

Yeah.

But we're definitely moving to— well,right now, this year was the year where everyone started realizing there's a best model per task.

Mm-hmm.

Next year, people are going to start training their own models. Um, and but it'll be, like, very good for one specific thing that's unique to that company.

Hmm. Where are the areas you think people are lighting the most money on fireright now with wrong model usage?

Yeah. I think it's, like, it's, like, in product where they don't allow their custom where they're not, like, smart routing their customers, like, uh, prompts, where they're just, like, like, everything they send because the customer's like, "I want to use, like, the latest and greatest.

I'm just going to pick Fable 5." And then they just end up, like, using so much money for, like, the stupidest stuff. And so they're not, like, smart routing it when they need to be.

What do you think is the biggest risk to Merge today? Because in theory, every single software could throw out an MCP, and there's no need to use Merge. I mean, in theory,right?

Yeah.

Technically. Is— I mean, is that it, or what's.

No, it's not. So, um, I think a lot of MCP is still pretty new. Um, and there's— there are some— it's— it's just a protocol,right? So, like, a lot of companies are releasing their own MCP servers. There's public MCP servers.

Um, but, like, for most companies, their public MCP servers were, like, side projects or, like, a marketing thing. It wasn't meant to be, like.

Or they didn't actually work.

A lot of them don't.

Really? I didn't.

Yeah.

I guess I didn't realize it was that.

We've been doing some studies, and the success rates of most public MCP servers is, like, 60%. We build all the MCP— like, the tools for MCP, like, in-house, and then we test them really extensively. So the success rate for ours are, like, 99.9%.

We're— we— we really, really bad all test them. Um, and also, like, there's different— like, you're going to have a really dumb agent that ends up, like, making it not successful. So we account for things like that.

Okay.

I know. It's really— it's really.

What's like— what's like the dumbest agent activity?

Yeah. You're— so you're basically like, "Here's the tool, and here's the information for what to say." Like, "You need to send us the date in blue." And it just is like, "I'm not sending you the date. I'm just going to send it in, like, red."

An agent will do that even though it's.

The agent will just be, like, fucking stupid. Yeah, we've seen that with, like, older agents. Like, some agents are just, like, really bad at tool calling.

So instead of upgrading to Fable, they're still on, like, GPT-1. Like, they're just, like, so dumb they can't even.

Yeah. And you'd be surprised. There's, like, some, like, kind of later-stage models that are just, like, still really dumb. Um, but yeah, it's not— it's not a risk for us because, like, there's other things that we cover in our platform, like DLPs.

We have a security layer that looks for making sure, like, PII isn't getting sent across. Um, you can add, like, custom rules too to make sure, like, if you have certain policies for your— for your company or you're in a regulated industry.

Mm-hmm.

You can block that data from getting passed to a third party. Um, and so there's— and, like, authentication, refresh tokens. Like, there's just governance, um, making sure, like, all agents only have access to certain connectors. A lot of those things are out of the scope of the MCP protocol.

Yeah.

Um, that we also include in our platform.

Why— why isn't that part of the MCP protocol? Like, was it just, like, when Anthropic kind of, like, invented this, did— was that not something they were thinking about?

I mean, MCP is already such a big— why are you shitting MCP?

I'm not shitting on it, but.

It's just— it's just out of scope. It's just out of scope because it, like, all those, like, the governance layer, like, the observability, like, logging, DLP, like, those are things that not every single company needs. Like, and also they might not think about.

Um, and it's— it's just a huge build.

Hmm. Well, and it's prob— yeah, to your point, it's probably, like, some of that stuff is more of, like, enterprise-grade.

It is. Yeah.

You need versus— I mean, the— the goal of MCP is just get people using Anthropic products. And it's usually just— I mean, the most average is, like, a developer in their bedroom making a side project on AI. That's, like, the— probably, like, the median project.

Yeah.

So it's just— it doesn't have to be that robust.

Yeah, it doesn't. Um, but what was interesting, so we actually start— when we— when MCP first came out, we actually were like, "Wow, this is actually going to become, like, a big security issue." And so we started off trying to build a security product for it.

Mm-hmm.

Um, but when we were going out to market, we were talking to everyone. No one cared about the security because there hadn't been— been, like, a— a vulnerability yet. Like, there hadn't been anything bad that had happened.

Yeah.

So now everyone was like, "Okay, this is nice, but, like, it doesn't matter yet because there's not, like, a big issue." And but we— what we really want are the connectors. And we're like, "Okay, I guess we're going to, like, build the connectors then."

Yeah.

So after now, the connectors are really good. Like, our connectors are very good, but, like, you can't have a security incident if the connectors literally don't work.

That's fair enough.

So now that we have, like, everything end to end where our— our connectors are really good, and we have the security layer, and then we also have the observability layer.

Yeah.

It makes it so it's really enterprise-ready.

Yeah, that's fair. So— but— but and you've pretty much gone all in on enterprise. Like, you still have self-serve. There's still, like, a lot of smaller companies that use Merge. But what was the decision to go all in on enterprise?

Because I think you mentioned you did it in, like, 2023-ish.

Yeah, yeah.

2022-ish. So why do it?

There's so many reasons. Um.

Did you have to? Like, was it, like, an existential, like, "We've got to do this"?

I think it became pretty clear that, like, we— well, with our first product, it was just a lot of effort for each customer. Like, the amount of time that we spent with each customer and, like, making sure they were successful and the amount of, like, requests that they would end up having, um, was just a lot.

And if the customer didn't make it, it was just a lot of work for someone who just died.

Yeah.

Um, so even if they paid us a lot upfront and they died, it just was very painful for our business.

Yeah.

Um, and so moving enterprise allowed us to have, like, longer-term relationships with companies that would probably be okay in, like, 10 years. Um, it allowed us to invest in them, them to invest in us, um, and also just made our company a little bit more predictable.

On the downside, we ended up becoming, like, more of a whales— whales business. So we ended up having.

Yeah.

Like, certain really large logos that we, um, clo— like, we close. Um, and there's, like, more timing risks in, like, our forecasts and our predictions. And, like, if something slips, it, it, like, hurts us more. Or if something comes in magically, like, it's really exciting for us.

Yeah.

Um, yeah. So there's definitely pros and cons. But I think, like, the pros for having enterprise customers is that you're building together. It's a longer-term relationship.

Hmm.

And you know that if there's something wrong, they're still committed to working with you versus, like, if it's very PLG and you're a comm— and you're more of a commodity business where everyone's just switching because you're the cheapest provider.

Yeah.

It's just hard to have customer loyalty.

Yeah.

It's hard to have, like, a second chance too if something goes wrong.

Really? It's interesting when you see, um, like, I don't know, like, market share of models, how it just fluctuates so much.

Yeah.

And even when you look at, like, maybe a year ago, all of a sudden it felt like out of nowhere, like, everyone's using Claude. Everyone's using Claude, Claude Code.

Yeah.

And then, like, a couple weeks ago, I was talking to someone in the latest YC batch, and he's like, "Everyone's using Codex."

Yeah.

I'm just like, it's just crazy that it just fluctuates so much.

I know.

And people just change.

People are, yeah, very unloyalright now.

Yeah.

Yeah. So that's why, like, sometimes, like, if your mo is, like, the customer relationship, like, it's helpful.

Yeah. I mean, that's a lot of things when an investor's evaluating, like, how good a business is. It's like, do you have sticky, hard-to-acquire customers? I mean, every business, it's just valued off the cash flows. Like, do you have customers that will pay you a lot of money that will not churn?

Yeah.

Like, that's essentially what a business is. It's just, like, a collection of customers paying you money.

Yeah, yeah.

Right? Really, at the end of the day.

Yeah, yeah.

We're, like, not romanticizing this at all.

Yeah.

But— but— but, like, a lot of investors will think about that. It's like, do you have big customers that will pay you a lot and, like, you've acquired these, like, really hard-to-get.

Yeah.

Businesses to the point it's like they won't churn.

Yeah.

So that's— that's valuable. Um, and so you think more startups should go after enterprise. But, like, it's hard,right?

It's hard. And also the revenue growth is, like, it's going to be slower than, um, than, like, what you're seeing with, like, Lovable. Like, Lovable is PLG, and so it's very easy to have, like, a ton of companies just, like, randomly sign up and, like, not.

Yeah.

Lovable doesn't have to do as much, and it just scales really quickly. With enterprise customers, like, each customer, like, you have to have a developer relationship. You have to, like, build together.

Yeah. The dinners. The—

Yeah, security review, like, legal. It's just— it's bigger.

Mm-hmm.

Um, yeah, so it depends. It depends. But then also, like, I've seen, like, some companies, like, they just get, like, some really, really sick, really big logos, like, immediately. That helps them a lot too. Um, but.

Yeah.

Yeah, it's— I don't know. If— if it's really hard, it's hard for someone to copy.

Yeah, that's true. I— I mean, I'm definitely— I'm a proponent of you should just do the harder things because, to your point, it's like, it's harder to copy.

Yeah.

Like, if you— you know, the— the whole meme of, like, someone will vibe code and just, like, steal a thing. Like, it's kind of true. Like, they kind of— you— you kind of want to make sure you're not in a position where people can just, you know, quote-unquote, "vibe code" your product and.

Yeah.

There's, like, a million competitors that pop up. So it's like, what can you do that's harder?

Yeah.

That, like, makes it that less likely and less possible.

Yeah.

Um, I feel like part of it too is also there's always so many other things, like, whether it's the customer relationships, like, the partnerships. Like, there's— I think it's actually not that likely that someone just vibe codes and copies the product.

No, it's such a pain in the ass to have to maintain something in-house.

Yeah.

Like, I would never vibe code Salesforce in-house.

Yeah.

Like, ever.

That was— it was so interesting that that discourse is probably, like, was it 2024?

I think so.

That was, like, the big thing?

Yeah.

That everyone's just going to vibe code.

Yeah.

And all the software's going to just come through.

Well, also, like, what kind of— like, okay, I don't know if it's going to be canceled, but, like, what— what— who has, like, the time to, like, build and maintain this, like, application in-house in addition to their core product?

Yeah. Well, you just said you make the— the events thing.

It's not a product.

But your events thing. That— so that— so this is not, like, you're selling it to people.

No.

You're using it internally.

It's just using it internally.

What if I make a marketing event field work tool that then I sell to you? Like, would you buy it from me or, like.

Yeah.

So then how do you decide then? Is Shensi's.

Depends on price.

Okay. So was it— you built this internally because it probably didn't exist.

It didn't exist.

And, huh, so then if I'm, like, thinking about, um.

I would wait if, like, but at the time, if it existed already, I'd buy it.

Interesting.

I didn't want to build it.

Yeah. How much would you pay for something like that, though?

Probably not a lot.

Yeah.

Because the— because it's not, like— but— but there are— but, like, Salesforce, we pay a lot because it's way more, you know?

Yeah.

So it depends on, like, the ROI.

But— and but then with Salesforce, I feel like there's people that say they, like, vibe coded replace the Salesforce. Why do you still use Salesforce, and why have you not vibe coded it?

Because it's a source of truth that connects to everything. Like, it really— everything connects to it. Um, we've maintained it for years. We have a lot of automations built on top. There's certain things that— yeah, there's just so much that we do in Salesforce that we can't— we just can't replicate if we built in-house.

Yeah. It's probably, like, the more, um, integrations that you have and the more, like, different things you connect with.

That's why—

The more defensible you are.

That's why I always tell a lot of these companies that are, like, a source of truth. I'm like, the more companies that are built on top of you, the more you're sort of— Lynnear does a really great job of this.

Yeah.

Like, they've created an ecosystem around their product. Um, that makes you really, really sticky. Ashby is also doing a really good job with this too. So many companies integrate with them, and, like, the more legacy companies are like, "Oh, well, like, you know, they don't understand."

Like, these newer companies, like, don't understand, blah, blah, blah. That's where we're, like, closed off and, like, we have secret sauce, and people are really dependent on us. So, but, like.

Yeah.

That's— that's fine for now, but that degrades over time.

You've kind of taken the mentality of, like, all the different partnerships, all the different integrations.

Yeah.

Like, you are kind of— you're— you're, like, kind of that layer, like, personified in a way.

We are. Yeah. We do. But the thing is, like, whatever someone wants, we'll follow. Like, if— if a company is like, "We want to be— we're openright now, but we're going to be closed off next year," we'll work with them on that.

Like, we are always going to follow what the partner wants and make sure that they're, um, that, like, we facilitate that. And we really want to be, like, that layer in the middle and help just make companies, like, integrate with them more easily.

And so you— one thing you mentioned is you still do a lot of outbound, like, as the founder.

Oh, yeah.

Manually. So, like.

A lot, yeah.

So why do that? You could, like, vibe code some or use some, like, AI automation stuff.

Oh, yeah.

So.

So, so it's— it's different type of outbound that's not necessarily, like, automated.

Okay.

Um.

So what's an example?

Yeah. So, okay, so for recruiting, I do a lot of sourcing. Like, for engineers, um, I use this recruiting tool as an AI recruiting tool. I, I do use AI for that. Like, I'll just basically just be like, "Okay, this is what I'm looking for.

I'll have— I'll have an agent that just, like, automatically sends outreach for candidates based off of the criteria that I've done." Um.

Oh, really? So you don't even know that it's sending it?

Yeah.

Oh, that's interesting.

Well, because I'm hyper-specific.

Okay.

Um, but I can also just, like, this tool's really great. It allows me to also just, like, select all and send.

Okay.

So I can do that too. Um, but, um, for, like, dinner outreach, it's— it's— like, it's kind of like having a good party. Like, you want to have a good party, like, you got to know the vibes.

Yeah.

The vibes.

There might be, like, Shensi would not be a good guest at this party, so I don't want to invite her.

Yeah.

Or I do.

Yeah.

Like, I do. She would be.

Like, I know this person's, like, fucking hilarious. I'm— they're going to be in the middle.

Yeah.

Or this person sucks. I'm going to put them, like.

Like, it's our— it's like, oh, it's one of our biggest customers. I haven't talked to him in a while.

Yeah.

I should probably— I should probably be here, but, uh.

Yeah, exactly. So there's just, like, certain things that are, like, hard to, um, replicate with, like, AI or even, like, having another person. That's why it's, like, hard if we hire someone to even, like, do some of these things.

Um, but yeah. And then another thing that was outre— the outbound is more, like, a lot of networking. I hate to say it because, like, it's kind of, um, like, I think when we were smaller, networking was, like, not helpful.

But now that we're, like, bigger and, like, there's certain networking events that have been, like, super useful for us and, like, ROIs, like, created partnerships.

Yep.

Yeah. We— networking has actually been, unfortunately, very helpful.

Networking in— in what— what sense?

Like, um, like meeting other founders.

Okay. And you're just, like, manually, like, you come across someone, you just, like, DM them, like, "Hey, want to chat?"

No, not quite.

Or what?

It's like you're at a conference and, like, you meet them. Or, like, you're at, like, some kind of, like, founder event and you meet them. Um, yeah. And you just have to say, like, "psycho shit" so they remember you.

Wait, what?

Yeah, you just have to say because the thing is, I think, um, like, you just have to be memorable.

Yeah. I like that advice though or that tip of, like, you have to say something that is memorable.

Yeah.

And then it's easier to follow up.

Yeah.

Because, like, the follow-up's so important. When you read those, like, networking one-on-one guys' guides, it's always like, "Make sure you follow up after the meeting."

Yeah.

Well, there's been times I've, like, gotten a follow-up from someone and I, like, forget meeting them.

Yeah. I know because they suck.

Yeah.

Like, so you just have to, like, be really memorable. And yeah, so I met the runway ML guy, like, on the— when we were, like, going from, like, one founder event to, like, another founder event, and we were just on the street.

And I was like, "Okay."

Yeah.

I need to just, like, not be a bitch. I need to just, like, say hi to people. And so then I talked to him.

Yeah.

But I talked to him and then— and I was like, "Oh, I'm never going to be able to sell to him." Like, I don't even know why I, like, bothered. But I was, like, really nice to him. And I made a few jokes.

Yeah.

And I made a few jokes. And then, um, now there's a deal. And he remembers me because I was, like, the only person that talked to him. Like, everyone else was being, like, such a fucking cunt and, like, I was the only person, like.

Wow.

That talked to him on that, like, street.

This was when you were walking.

When we were literally just walking.

Yeah.

And I just talked to him.

Interesting. Actually, there's one, um, event at Michigan, University of Michigan that I went to, and, uh, just, uh, one of the— she was, like, a professor, also has a company that I talked to in between some events. And yeah, I just kind of saw her randomly walking through downtown.

And I was like, "Oh, I know that person." I just said hi to her. And, like, again, like, I remembered her because.

Yeah.

We talked at this thing.

Yeah, you have to be memorable.

Yeah.

Yeah. It's, like, really— so, okay, so this is, like, four years ago. Um, we were at Web Summit. And, um, I was, like, a baby founder then. Like, I wasn't, like, super experienced.

Yeah.

But I was, like, I— like, at the— like, at the time, like, Carta was, like, the hottest company in, like, the valley. Carta was just, like— and I remember being, like, "Oh, I'm, like, so bombed because Carta is, like, using her competitor and, like, I really want them as a customer."

Like, it would just be, like, a dream logo. Like, we've used Carta for a long time. I've always looked up to Henry Ward. Like, I really wanted Carta as a customer. And so then I look up and I see him.

And I'm like, "Oh my God, I have to talk to him." Like, I'm never going to see him again if I don't talk to him.

Yeah.

Um, and my co— and Gil, my co-founder, he's like, "What are you doing?" I was like, "I need to go talk to him." And I was like— I was, like, shaking. I was like, "Oh my God, I need to just, like, fucking man up and I need to, like, go talk to him."

Yeah.

So I go up to him and I introduce myself. It's so creepy because I read, like— I read you everything about him. Um, by the time like we.

It's like one of those things, like, you introduce and you, like, read their biography to them. Like, you recite every single information.

Oh, yeah.

That you know about them. Yeah.

100%, yeah.

So how do you avoid— like, what do you say when you first come up to someone like that?

Yeah. So I just had to be, like— I'm not usually, but I had to be, like, pretty charming. Um, I had to be, like, "Hi." And, like, like, and, um, like, think of things to say and, like, comment on, like, the event.

And, like, I don't remember exactly what it was. Um, but yeah, like, he remembered me. Um, and then I got a text or, yeah, an emailright after. I was like, "Hey, I'm doing a dinner tonight. Do you want to come?"

You got that from him?

Yes, from him. And I was like.

Wow.

I know. I was like, "Oh my God."

That's huge.

It was crazy. Um, and then so that night, we had dinner with Henry Ward.

Wow.

And it was— it was such a cool because I had looked up to him for so many years. Like, and, um, yeah, I just feel like he— he's just such a great founder. And I got a SNX to him for the whole dinner and I got black out.

Um, but, like.

So you don't even remember the dinner?

I don't remember most of it, but I had a great time. We had, like, a lot of— a lot of jokes. I, you know, we had a good time. And then they became a customer.

Wow.

I know.

So you have this thing called the Embarrassment Framework.

I do.

So what— what's the Embarrassment Framework?

So I always tell my team, like, what's more embarrassing? Posting something on LinkedIn about Merge or the company dying.

And so there's a clear answer.

Yeah. I mean, that would be terrible to actually have to post on LinkedIn.

Yeah.

But so that's the framework. It's just like.

It's the framework. It's basically— yeah. And, and, like, it's helpful for the team because I think a lot of them are kind of like, "Oh, like, it's so embarrassing." But I'm like, "Yeah, like, but, like, it helps us so much."

Yeah.

And, like.

Imagine if, though, you worked at that company and it died and it's just because you didn't post on LinkedIn.

Yeah.

Like, come on, just make the post.

Yeah. And obviously, like, okay, a LinkedIn post isn't going to save the company. Like, there's so many different things, but it certainly helps.

Yeah.

And it's free. And, uh, like, and it definitely propagates more to, like, the founders. Like, it is the founder's responsibility to do the thing that's embarrassing. Like, you need to send that email.

Yeah.

You need to, like, go up to that person. You need to, like, shoot your shot.

Yeah.

And there's so many times I've had to send, like, an email that is so embarrassing and, like, so horrible. And then I'm so glad I did.

Hmm.

Like, um, so Dropbox is one of our customers. And we went through deal process with them and they were like, "No, we're not going to move forward." I was like, "Oh, so sad." And then I kept— I, I stayed on it.

I was like, "Hey, like, how's it going? How is, like, building your house? Like, any updates?" And it was so embarrassing. Like, obviously, they didn't want to, like, fucking work with us,right?

Yeah.

Um, but then, like, I think, like, the sixth time I emailed, they were like, "Let's chat." And then now they're a customer.

Interesting. So how do you get past that embarrassment? Like, what's the— any advice for, like, doing these, like, uncomfortable things? Like, I get it. Like, the company's going to die, but you still have to do the thing.

I feel like your inner autist comes out.

So you almost, like, take down your guard of, like, like, social acceptability and just being like, "How's it going building this thing in-house?"

Yeah, pretty much.

Interesting. Okay.

Yeah.

How do you channel your inner autist?

I think you just have to give no fucks, but it's hard to get to that point. Yeah.

Yeah.

But you can be memorable over text too. Like, I show my friends, um, and my team members, some of the messages that I send to our prospects where I'm like, "Hey, Tim, why the fuck are you ignoring me?"

Or, like.

You actually will say that?

Yeah. Tim from Apollo has gotten so many times. It's like all blue, but I love him.

And, and that works.

It works because he, yeah, we love each other.

Okay.

Or maybe I love him. I don't know. But yeah, like, when we hang out, it's a good time. But he remembers me, you know?

Yeah.

Yeah. Um, also, one of my friends, um, she's— she's, like, a really, like, hot girl, but she's definitely autistic. She went to Stanford, but she, like, looks normal, but she's just— so, but it comes off like she's a bitch because she, like, because she looks normal,right?

Yeah.

Um, so she tells me that she, like, le— she knows she's, like, a little autistic and she just, like, leans into it when she needs to for business con— for, like, business purposes. And that actually really inspired me because I'm, I'm definitely not there.

I'm not like that.

I think everyone is, like, somewhere on the spectrum though. From zero to 100, most people are not zero.

Yeah. I'm not her level.

Yeah.

But, like, if I need— if I need to, I'll, I'll just try to do it.

It's kind of like I feel like, um, you know how, like, the first time you drink alcohol, you, like, drink beer and you're like, you get drunk and you're like, "Oh, this is what it's like being drunk." It's almost like you can— it, like, brings that inhibition level down.

Yeah.

And you can, like, tap into it later.

Yeah.

I wonder if it's, like, the same thing where you, like, you get an example of it from someone else and you, like, you learn, like, "Okay, this is how I do it." I don't know.

Yeah, she's really good. She's, like, very direct. Um, so yeah. Will this get me canceled? Will people be like the anti-autism? I don't know.

I mean, if they didn't get canceled— if you didn't get canceled before in the first hour and a half, I mean, this is— this will definitely do it.

No.

No, I don't know. I feel like, um, I feel like it's almost, like, uncancelable talking about, like, being— like, it's, like, acceptable,right?

Yeah.

Like, everyone's.

I think if it— I think if it doesn't— if it stays in the tech sphere. If it once it gets out, then it's kind of like.

Yeah, that's fair.

Yeah.

But then, you know, depending on.

Also, you never know when things will— okay, soright now it's not canceled out, but in, like, 20 years, someone's going to be like, "She, like, she said the A word."

Yeah, that's true. Yeah, that's true. It's— it's so funny when they'll dig up, like, old videos of Trump and he's just like, "Republicans are idiots." Like, if I were to run for president, I'd be a Republican because they're just really dumb.

Yeah.

And, like, I don't— he probably wouldn't say that today.

Yeah.

But, like, he said it in the '80s.

Yeah, yeah.

So, like, I don't know. Anyways, I, I feel like being canceled is kind of overrated. Do you— do you have a favorite, um, CEO, founder, or business, um, maybe, like, throughout history at any time that you get a lot of inspiration from or that you've always looked up to?

Yeah, I love Frank Slootman.

Frank Slootman. Okay.

I know that's kind of a basic bitch response, but.

So why?

Because Alpha is fuck.

Okay. So what's been the biggest thing you've learned from Frank Slootman?

Well, I've never met him, so it's a little harder. Um.

It's that parasocial relationship.

It is. It is, yeah. Um, yeah. I mean, I— yeah, I, I mean, I think he— he's really good at, like, saying the thing that's hard to say. Like, he's very, very direct.

Hmm.

Um, and he's not, like, afraid to push people, which I think, like, a lot of people forget. And he's okay with being disliked.

Is that a hard part about being a founder is, like, being disliked? Like, you say something and somebody, you know, they— they— they got quit.

Yeah, 100%.

Like, I don't know what for.

All the time. Like, you say something that's, like, true and correct.

Hmm.

Um, and for the good of, like, them, like, because you, like— maybe this is also because, like, I don't know, like, maybe, like, I'm a woman, but, like, I really care about people. And so.

Yeah.

Sometimes I'll, like, share feedback because I'm, I'm wanting them to, like, really do well.

Yeah.

And they'll take it personally in a way where they think I'm, like, trying to, like, hurt them.

Hmm.

Um, or it's just, like, malicious, but it— but it's not. It's for the purpose of, like, them and their personal growth. Um, but yeah, it makes me get disliked.

How have you learned to give, like, harder feedback as a founder? Because I'm assuming you have to give it a lot.

Yeah. I just am direct. Yeah. I just— it's just, like, you have to just say it.

Yeah.

Like, when you're not saying it, it ends up hurting the company and the team a lot more over time.

Yeah. I feel like there's been, uh, honestly, it's probably the most from my wife nowadays of, like, telling me something where, like, "You need to change this behavior" or something like that.

Yeah.

Or, like, "You do this thing." I'm pretty avoidant. Like, I have an avoidant personality.

Mm-hmm.

And, um, and she's always like, "You need to stop avoiding this." I'm like, "You'reright."

Yeah.

Like, I do.

Yeah.

So I think it's really— again, like, it's— it's hard because you.

Yeah.

You, like, don't want to hear it.

Yeah.

You, like, hear this negative thing, this thing that, you know, you messed up.

Yeah.

Or you need to be doing something better and, you know.

And the person— yeah, if the person knows and they clearly don't want to talk about it, then I'll just like, "Hey, I think we both know. Let's not talk about it, but let it can't happen again."

Is that something— is that a good way to approach it?

It's not.

It's like, "Oh, we both know."

If they know, then yeah, it's not really worth it.

Yeah.

Yeah.

That's true.

Because it's just, like, they're going to be harder on themselves than I could ever be. So.

Yeah. And that's the people you want is you want people who are like.

Yeah.

They realize, like, "I messed up. I really want to— I want to be better at this."

Yeah.

I'm really good at it.

That— yeah, exactly.

Yeah.

So, but yeah, I mean, it's hard and especially because the company's bigger. Like, I, I— I think the thing that was hard for me to adjust for was that the feedback got scarier. Like, I think the people.

Yeah.

That I worked with much earlier when, like, the company was smaller, like, we were just so close and there was so much trust from, like.

Hmm.

Two-way feedback that it wasn't really— like, I didn't even have to think about giving feedback and they didn't have to think about giving me feedback because it was just, like, it was just, like, how we worked together. Like, it was like, "Oh, this part wasn't super good," or, like, "Hey, like, this was a little delayed.

Like, you can do that next time," or, like, "Can you do this better?" Like, it would just be, like, two-way. But then, um, it was almost like once we hit like 75 or like 100 people, it just was, like, all of a sudden, like, scary.

Like, people would just be, like, really offended by things that I said. They didn't know me as well, so they didn't really understand. Um, and yeah, like, I always— I met, um, like, Howard Lurman from Yext and I— it was, like, around the time where I was going through, like, a really hard, like, founder transition where.

Hmm.

I just, like, wasn't used to how personally thing— like, or how removed people felt from me.

Yeah.

And he was like, "Yeah, it's, like, really hard at this point because, like, they don't understand you. In the beginning, they really want to understand you."

So when did that transition kind of happen?

It was like 75 to like 100 people.

Okay. And the— and the— the thing that you go through is that you just are not spending as much time with every single person, so no one knows you as well.

Yeah.

So what do you do? Like, how— what did you change after that?

I realized I needed to go through my managers a lot more and my executives and be a lot more reliant on my executive team.

Hmm.

Um, and that I had to spread out the responsibility of sharing feedback and, like, ownership with my executive team more. Um, and that helped a lot. So, like, we have a really, really strong manager bench. Like, our managers are very— like, I spend a lot of time with them.

Yeah.

Um, but, like, I don't give feedback to ICs anymore. I give feedback to their managers and their managers already know usually.

Hmm.

Um, and we'll be giving feedback to their team members. Um, same with our exec team. Like, our exec team will usually end up, like, talking directly with, um, the managers and then the ICs too.

So then your job— so then what is your job then as a CEO when you meet— like, it's my first day as an engineer. I'm meeting you.

Yeah.

I'm probably, like, pretty nervous, honestly.

Yeah.

I'm like, "Oh, man."

Yeah.

I mean, the CEO of this, like, hot company merger.

Yeah, something really bad happened. Um, so there's this guy who started and he— he was so nervous and he was, like, introducing himself to me.

Yeah.

And then on the second day, he actually called me another Asian girl's name.

Wow.

I know.

Yeah.

I was like, "I mean, I don't care." Like, for me, I'm like, "I don't care."

Yeah.

But I know he will never forget that.

Yeah.

I know he will never forget it. He will always think about it. I'm like, "That's so fun." I feel so bad for him.

Yeah.

But he's so nice.

Yeah.

But he— yeah.

So— so then, like, is your job as a CEO, like, it's mostly making people feel excited about what you're doing at the company or is it, like, motivating people or, like, sharing the vision? Like, it's— it's less about, "Hey, specific tactical things."

It's more, like, big picture. Like, is that more if I think about it or?

Yeah. So it's things that I can uniquely do. So it's, like, recruiting, um, showing the direction of where we're going, um, making sure, like, like, morale, like, making sure, like, you know, the team is, like, excited. Yeah, excited.

It's what you said, like, exciting, excited, like, know where— what we're doing. And a lot of those things can be aligned and, like, to get close together too. Um, and then, like, there's certain, like, strategic big deals and, like, sales things that, like, Gill and I need to be involved in.

And then product, obviously.

Hmm.

Yeah.

Do you— what's your level of involvement in sales? I mean, it sounds like you're still texting and emailing customers all the time, but, like, what's the level of involvement?

Yeah.

Have you— have you gone through any cycles of, like.

Oh my gosh.

More or less?

Yeah. Sometimes it's always, like, um, like in two to three-month stints where it's something will be, like, my 100% focus. So, like, earlier this year, it was like 100% recruiting with, like, some product.

Yeah.

And then at some points, it was like 90% coding. Like, I— we just really needed to get, like, a product out and I— I wanted to know what was going on. So I— I was coding so I could help contribute.

Hmm.

Um, and thenright now it's, like, mostly a lot of marketing.

Hmm.

Um.

Well, yeah. So then how do you know when to lean in and out of different things, especially, like.

Yeah. I— I don't— this is so bad, but I— it's just, like, gut.

So it's just your gut is like, "We need to get the word out. I need to do marketing," or, like, "We're shipping this thing. Like, I need to work on the product as well."

Yeah. I go to a spot where it's, like, extremely high ROI. Only I can do it.

Hmm.

Um, and if we don't, then if I don't help out with this, then we're— it's going to really impact the company negatively.

What was the hardest thing to— to give up doing, like, on a daily basis? Because I'm assuming, like, when it was you and Gill in the— in the room during COVID, like, you did everything.

Yeah.

What's been the hardest part to give up?

I love coding.

Hmm.

Yeah. I really love coding.

And so— and you obviously don't do that all the time anymore.

I don't get to do it. I'll do it for, like, for fun or, like, to help the team. Also, because I work a lot and I'm kind of a loser. So, like, when I've— I've a lot of— so I just have a lot of time to focus on merge.

Yeah.

So, like, I'll— with the time that I have.

Your free time is coding.

Yeah. Yeah. For fun. Like, when I'm watching Girl Housewives, I'll code.

Hmm.

You know what I mean?

What is it? Girl Housewives?

Real Housewives.

Oh, oh, Real Housewives.

Yeah.

What's so good about Real Housewives?

Oh my God. It's just you can, like, you can go in and out and multitask in a way that you can't do with anything else. Like, it's not just Real. It's like all Bravo shows, like Summer House, like Vanderpump, Southern Ch— like, all the shows.

Like, I don't really need to know what's going on because it doesn't matter.

Because just, like, some fight is happening.

Yeah.

Like, "Oh, this is entertaining."

Exactly. So, but I could do it while coding.

Okay.

And it allows me to be, like— it allows me to spend, like, versus if it's, like, silence and I'm coding, it's not the same vibe.

Interesting. Okay.

Versus, like, I feel like I'm, like, enjoying it. I'm, like, I'm watching TV while I'm, like, doing something very productive.

Yeah. Have you ever watched, uh, Traders? Have you seen that one?

Yeah, fucking love Traders.

I love that show.

I love it.

So interesting.

We should do, like, a VC Traders.

I thought about it.

We should film it, like, what Founders Fund did.

I thought about it. So, like, the— the issue or the thing I thought about is, like, it's a lot of work.

It's a lot of work.

Like, it's just— this is helpful because, like, I— I, like, learn a lot through this and it's just kind of fun to hang out. You get to meet new people. I make money from it.

Yeah.

Like, it helps the portfolio. Like, you have founders that you either, like, want to meet, want to help.

Mm-hmm.

Meet people. Like, there's just, like, this whole thing is, like, pretty helpful versus if I spent a month doing, like, a Traders show.

Yeah.

Like, it would be really fun.

Okay.

And I'm sure people would talk about it. Um, I don't know. I— one other one I thought about is making, um, parody music.

Oh.

Like, parody music.

Oh, that's pretty good.

So, like, one, um,

what was this? I think you— you know, um, by Drake, it's called God's Plan.

Mm-hmm.

Some guy made a song, it's like Claude's Plan or something like that.

Yeah, I saw that. So good.

Yeah. Th— that was really good.

So good.

One, one that I— one that I thought about is, like, um, the— who's the guy? Chief Keef. He's got a song, like, "I Hate Being Sober." Like, you could probably make a song, like, "I Hate Being Bootstrapped" or something like that.

Like, and he's, like, all the words are just, like.

Yeah.

Talking about, like, AI, like, token maxing.

Yeah.

I don't know. Like, you could probably just take popular songs and just change the words, like, the same pattern, same flow, and it's just, like, about something that's, like, startup tech.

Yeah.

VC related, and people would identify with it and get it.

Yeah.

And it's just unique and different.

Yeah.

Because, like, the— I think, like, one of the most important things with, like, generally marketing is, like, you don't do the same thing as everyone else. Like, you figure out, like, what is, like, a new thing to do or try.

Yeah.

It's like we kind of go through cycles with, like.

Yeah.

Launch videos, everyone starting a podcast.

Yeah.

Et cetera. So it's like.

Also AVG marketing.

AVG marketing. Yeah. Have you guys gone through any AVG marketing waves at Merge?

Unfortunately, no.

Unfortunately?

Unfortunately, no.

Does it work? Like, should you?

It probably does. Yeah.

Hmm.

All those posts have gotten, like, hundreds of thousands of views. I don't have a single post that has hundreds of thousands of views.

That's not true. You have a couple launch videos.

Okay. Yeah, that's true. So maybe I accidentally was my own AVG. Yeah. I don't— yeah, I don't know. I mean, it does— I mean, eyeballs works.

Yeah. But you're also, like, literally finding love at Merge. Like, the employees are literally getting married. So that is, like, a good— I feel like you're creating something.

Yeah. We're creating ARR and love.

Yeah. Well, this has been a lot of fun. Thanks for coming on the show. Thanks for doing it.

Oh, yeah. Thanks for having me.

And thank you for listening. Shout out to this episode's sponsors, Flex, Numeral, Amplitude, Merge, and Monaco. If you enjoyed this, please like, comment, subscribe, and share it with a friend. Make sure to check out the back catalog with over 100 episodes with investors like Gary Tan, Vlad Gill, Jason and Eric at Benchmark, and the founders of companies like Robinhood, Sweetgreen, and Mercury.

Tune in over the next few weeks for conversations with Michael Tannenbaum, CEO of Figur, the first blockchain-based lending company, Chris Holson at Drive Capital, and Ryan Neese at Next Legacy. If you don't want to miss any of these, subscribe to our newsletter, the split linked in the description to get each episode plus a transcript emailed directly to your inbox every week.

Thanks again for listening. See you next time.