S15E4: From Starlink to AI Marketing Agents with Doug Tallmadge, Co-Founder and CEO of Gradial

Marketing is fundamentally trying to win the hearts and minds of the same, you know, 100 million households or 100,000 enterprise customers. And so it's never done. It's like every day, tomorrow I have to wake up and I have to come up with a new way to do marketing.

Welcome back to a new season of The Room Podcast. Hi, Mads.

Hi, Clauds. As you all know, we are opening the door to the room where it happens for today's technology founders. I

mean, I can't believe this is Season 15.

Crazy.

Probably one of our most exciting seasons telling stories up and down the stack.

After 6 years in the room, we're just delighted to be back together in San Francisco more and more, both being based here live from Vercel HQ. I think you have an update as well.

Yes, I'm also now at OpenAI. Wow. Which is fun, pulling us deeper into this revolution we're seeing globally.

Thank you everyone who's been listening to us since the beginning, 6 long years ago. And for those who are new here, please like, follow, subscribe. Subscribe wherever you listen to podcasts, or of course at theroompodcast.com. Today on The Room Podcast, we have an extra special conversation for me with Doug Talmadge, CEO and co-founder of Gradiol. Gradiol is the company that has built the platform content operation engine using AI to automate the entire content supply chain from the authoring to QA to page creation, proving a recent $675 million valuation Even in the unsuccy space of plumbing for marketing, there's a massive market. Let's open the door.

And now a message from one of our sponsors, Rippling. One thing we constantly hear on The Room is that great operators don't just need more data, they need a faster way to turn that data into decisions. As a founder, I know how painful it is to wait days or even weeks for a report while a real business problem is unfolding. That's what makes Rippling AI different. You can ask Rippling AI, is our sales team at risk of missing quota? And Rippling AI instantly pulls real-time data from Rippling and your CRM into presentation-ready dashboard data showing quota attainment, headcount trajectory, and monthly revenue to quota by region. But it doesn't stop at the insight. Rippling AI can act on what it sees, recommending and building an automated recruiting workflow when open roles and slow time to fill are putting quota at risk, all while following your existing permissions and approval processes. Question answered, action taken, problem solved. Ready to rule your business? Head to rippling.ai/theroom and sign up for exclusive access.

And now a word from our new sponsor, Gradiel. Whether you're building a company, launching a product, or growing a brand, you know the idea is just the beginning. And marketing's bottleneck has never been creativity. It was always the work it takes to get an idea out the door. Gradiel handles the operations so that marketers can stay on craft. An agent will learn your process, work across the systems you already use, and delete point solutions you don't need. Designers will ship on-brand pages without counting prompts or tokens. Product marketers will launch across markets in days, not weeks. And digital experience teams test customer journeys, catch what's not working, and fix it all before launch. Radial learns about your brand, business, and process and brings that continuous context to every workflow. Every task gets done your way with brand, quality, and approvals built in. No more token maxing, slop, or runaway AI bills. More time to think bigger and create the impact people remember. Graydeal, the system of work for marketing. Ashurst Perkins Kewey is a leading full-service international law firm known for providing high-value strategic solutions for startups and investors. Lawyers at Ashurst Perkins Kewey support the most innovative entrepreneurs and investors in fast-moving and high-growth sectors across the country while addressing the myriad of legal needs that arise when fundraising and scaling their businesses. The firm's tailored legal services are informed by understanding of each of the client's discrete needs. To learn more about Ashurst Perkins Coie, visit their website, ashurstperkinscoie.com.

Doug, thanks so much for joining us today in The Room.

It's awesome to be here.

We're in San Francisco, which is not your normal home. Where do you live right now?

I'm based in Seattle.

We're going to start more at the beginning and then dive into the story of Fraydl. So to kick us off, I'm curious, Where did you grow up and how has that shaped your view of the world?

I grew up mostly in Cincinnati, uh, for most of my childhood.

Cool.

And, you know, I think, uh, it was a great spot to grow up. I mean, I think one thing was very much a P&G town. And so, you know, in terms of, we'll get into a little bit about how that might lead to Gradio a little bit, but, you know, a place that has a lot of appreciation for consumer goods and marketing and building businesses like that. And, um, definitely something I remember from childhood.

Absolutely. Yeah, it's in like the heart of America and there is a lot of innovation that happens there and has always happened there. But necessarily not a founder hub. So I'm curious, did you yourself always think you were gonna become a founder?

I think at the time my terminology was inventor. And so I, I definitely told my parents that I wanted to be an inventor from an early age. And I think it was because we had this like book report project in school where we were supposed to design a pumpkin that looked like some famous person from history. And so I made mine about Bill Gates at the time, and I was like, I wanna be an inventor. And I think later on that kind of has had, had, you know, changed a little bit into like, what is, what exactly does that look like today? Like, you know, you know, Thomas Edison working in a lab. Okay. What's the equivalent of that in like, you know, 2026, 2027? But yeah, that was the origin for sure.

Okay. So this is making more sense why you went on to study physics at Dartmouth though. That's definitely an inventor major. I feel like, I mean, that's actually where we met so many years ago. We're not going to say how many years ago, but Um, at, in undergrad together. And I'm curious, like the experience of going to Dartmouth, doing the physics thing, and then ultimately you went into like Lockheed Martin and found yourself at SpaceX. Like how did those early career moves and kind of experience at undergrad shape what you did ultimately?

Dartmouth has a lot of pull towards finance. Yeah. And, you know, I was not immune to that. I think, you know, during the sophomore summer recruiting, it was like I applied to all the kind of banks and financial institutions, hedge funds. And, you know, the best and worst thing happened. I actually got an offer at a hedge fund, and I remember calling my parents and, you know, I felt like, okay, this is obviously good. I'm excited about this. Like, you know, I had some relief, but then I was like, you know, a little bit disappointed 'cause I was like, is this what I'm gonna do for the rest of my life? And so I like called my parents and they're like, well, what would you rather do? And I said, well, you know, I maybe would rather work at SpaceX. And they're like—

You said that?

I did. I literally said that cold. And I remember I was sitting in the library and, My dad's like, well, have you applied to SpaceX? And I was like, well, you got me there. And so, you know, I applied that night and I did not hear back for about like 4 or 5 months. And this was like during a time in SpaceX where like, I think Falcon 9 had only launched like a couple times, like a handful of times, maybe 15, 18 times. Now it does that, you know, in like the first half of the year or first quarter of the year. And so, you know, I ended up getting the offer to go and I'd already, actually accepted the hedge fund offer. And so, I remember looking and thinking, okay, well, I've already got Lockheed, I've already got my hedge fund, like, where am I gonna fit this in? Like, how am I gonna do this third internship? And I was able to convince, um, the dean, um, to let me take an extra term off.

Really?

And come up with a plan to still graduate. And so, I was able to fit in that SpaceX internship. And of course, that was much more satisfying in terms of like, you know, you got out there, you saw the energy of the, the rocket factory, and like, you know, every morning you'd have your cup of espresso, like, sitting right there while there's like someone on the, what they called the Octoweb, which is like, like the 8 engines that are in like a kind of octagon type format with the 9th engine in the middle on the Falcon 9. And, you know, the energy was electric. And so that was, that was kind of how things got started for me. And then after college, it was a choice of like, do that or kind of go down this like, you know, seemingly more short-term lucrative route in the finance space. And, you know, I would say the inventor side of me definitely, you know, pulled through on that one. Um, although obviously, like, you know, a lot of engineering turns out to be, uh, more execution than like inspiration. And I think that's something that you learn over time the more you get your hands, um, into actual problem solving. But that was the intro. Yeah.

With startups, you know, building something interesting and iconic is that marriage of execution and invention. So before we kind of get into the founding story of Gradial, you were at SpaceX for about 5 years, um, working on building out Starlink, which is incredible. I took a United flight recently and the internet actually worked.

Yeah.

Sorry.

Imagine that. Imagine that.

It's great to hear.

During your time at SpaceX, did you think Starlink was going to become as big as it ultimately has?

Well, my first day on Starlink, uh, this was before we had any satellites really. And so it was kind of like everything started from the whiteboard. I found out that, um, you know, there'd been a pretty big change in the leadership on Starlink just in like the kind of week before I got there. And, you know, things were, were pretty, pretty chaotic. And so I remember thinking to myself, I called my parents, I was like, did I make a mistake? And, and it was not obvious that the business case was gonna work because you started to do the math and you said, well, we're gonna have to launch an insane amount of satellites just to be able to serve 50,000 users.

Yeah.

Right? And then you say, well, if 50,000 users are paying you about $100 a month, that's about $50 million in revenue. How does that justify even one launch? Right? Like $50,000, you know, $50 million revenue. It's like, okay, we just paid for like one launch. Yeah. Right? So like that became something where I was like, Does this make sense? Like, how does this actually work? And of course, you know, the answer was we had to scale Falcon 9, um, by like an order of magnitude, maybe even more than that. It was like, you know, it used to be like when I was interning, you know, I think it was like 6 launches a year, like maybe, maybe it was 10, maybe it was 12. And it was like with, with Starlink, you actually had to get to, you know, 25, 50, 100 launches a year.

Wow.

In order to get the scale that you needed to build a business that was eventually gonna go do, you know, tens of billions of revenue. Um, and so that was, um, that was interesting. I think the, the cool part was seeing a team work through not one, but what seemed like probably like 10 existential challenges, right? Because another one is like, Starlink is also a consumer hardware business. People think of it as like this satellite business, but actually one of the hardest parts is you have to build a consumer device that can talk to the satellite economically.

Yeah.

And previously people had only really built those, those phased arrays for like, you know, $20,000, $50,000, $100,000 a phased array, right? It was like mostly government applications that were extremely expensive. So it was a question of, can anyone actually build one of these devices at a, at a price point that lets you build a business like Starlink? And of course the answer was yes, but like even internally, like I remember back in 2018 when I first joined, like it was not clear how that was gonna work. There was not 100% agreement around, you know, even the engineers who were experts in the space of like, you know, do we go this way? Do we do that way? Um, you know, some of the competitors to Starlink chose different approaches where they had like physical dishes that like would basically have a motor that would point the dish at the satellite.

Yeah.

But the disadvantage of that is you actually need 2 dishes. So you just doubled your cost and you have a reliability issue as well where like, you know, if the motor goes out, you lose internet. So all those little challenges became like, you know, very, very important to getting Starlink over the line. But I think the thing that they really showed was like, you know, it's actually possible to kind of just fight your way through all of those barriers as long as they're not limited by the like fundamentals of like, you know, either the business incentives or also like physics, right? So that's, That's probably the big lesson from Starlink was like, you know, as long as the fundamentals are on your side, and in the case of Starlink, the fundamental was if I can give you a magic box that can deliver internet anywhere in the world, would you pay $100 a month for it?

Yeah.

And if the answer is yes, then it's like, okay, now I just have to go solve that engineering problem. Right. And I think it's interesting because I didn't really think about this until after starting Gradial, but there's different types of product market fit. In some cases, the product market fit is like a search process where it's like, you know, you hear about this a lot from software companies of like, okay, we're going to go discover product market fit.

Yeah.

Starling didn't have to discover product market fit. It was like, we already know that if we have this magic box, people will buy it. The hard part was discovering how to build the product.

Yeah.

And, you know, I would say Gradle kind of fits a little bit more in the other model, which is like, you know, going and learning about the market, figuring out what product to go build. Um, but there's something to be said for the opposite problem where it's like you've got a very clear objective, you know it's gonna work if you can just achieve it, and you just have to go engineer against it.

I'm kind of struck by the story when we had Sridhar Swamy on many years ago. He was the founder of Neeva. He's now currently the CEO of Snowflake. But when he was building Neva, he had left Google to create the first like paid search experience, more so on the premise that people don't want their data monetized and so are willing to pay a premium on having a private search experience, if you will. And it was right directionally that people are willing to pay for search because now we do with Chat and Claude. But that was wrong that like at that time, that was what the premise would be in the sell.

Mm-hmm.

And kind of seeing how like the consumer experience and expectations also like evolve alongside like the willingness to pay. Is like really interesting, I think, to the premise of what you're saying.

And I think like regardless of whether you're approaching it from like, I'm going out to like pivot my way to finding product market fit, or like, I know what needs to exist in the world, we're going to figure out how to build it. You're distilling everything to a customer promise and a discrete problem that you're ultimately solving. Um, and I think that's a perfect transition to learning a little bit more about, um, the early days of Gradial. Tell us a little bit more about the aha moment. Um, when you realize that, okay, this is going to be a company that you're starting?

The starting moment was, you know, even just the first time I used ChatGPT, realizing that the technology had kind of crossed the chasm to being useful. Um, because I remember even using text-davinci-003, like, or, you know, the GPT-2 model before that, it was like, this is incredible. This is like surprising that this is possible, but it also doesn't quite feel like it can do work yet. And so, you know, like a lot of folks, I think, You know, ChatGPT was the moment where it was obvious that it could do useful things. But I think the first thought that, that, that, you know, we had, you know, my co-founders and I was like, this would be amazing in a work context. And it's like, you know, ChatGPT is great, but I want it connected with my Slack. I want it connected with my Notion. I want it connected with all my data. And I also want it to do stuff. I don't want to just ask questions. I want to have it actually go do things for me. And that was kind of the moment that got us to, you know, quit our jobs and start the company. And we did that, you know, just a week or two after ChatGPT. And it was like, you know, very much in this kind of broad, we're gonna build for the entire enterprise type headspace. Um, after which we, we quickly kind of realized, okay, we actually probably need to be a little bit more focused and, you know, come up with some part of the enterprise to, to actually go build around. Um, and then the discovery from there was like, you know, we were— had to do some reflection and say like, what are the things we believe about enterprise work that we think, um, you could build a company around? And the core was basically this idea that the enterprise runs on tickets. Mm-hmm. And so, you know, as a software engineer, you're doing everything off of like a Jira ticket. Nowadays people are doing, you know, mostly using Linear, but like, you know, there was a time period where it was like everyone was using Jira and it doesn't matter what type of software you're doing, you're cutting a ticket, you're moving the ticket through. And then you start to realize, well, customer support runs on tickets. You know, in some cases at a large enough company, you might even run legal on tickets. CRM kind of resembles a ticket, right? A deal is kind of a little bit like a ticket. And so then we, we realized, well, if you could automate that movement of those tickets through the workflow, that would be something that'd be valuable. And this was in early 2023. We had a whiteboard where we basically wrote out what are all the ticket use cases within a large company. And then we looked and we said, okay, well, which of these are we gonna focus on? Right? And you could do, you know, finance, legal, accounting, HR, you know, people have a Workday ticket, right? You have a new hire, you have a Workday task.

Yeah.

You know, we picked marketing kind of for 2 reasons. One was, you know, partly luck of discovering some early customers in Seattle that we could go and work with. The other was this belief that marketing had, um, the ability to influence top line. And it's like, okay, what's the best I can do with an HR ticket? Best I can do is go from 100% cost to 0% cost. There's a fixed ceiling on how much value you can, you can generate. And I'm sure you can come up with creative ideas and, you know, there's probably companies that would tell me that I'm, I'm not thinking big enough there, but I think To a high level, there are certain back office type things where it's like the best you can do is cut cost. Yeah. Marketing, you actually can build something in the brand that delivers top line impact. And I think that was important for us from a TAM perspective. It's also a much larger part of the budget. And the other thing is marketing's a competitive sport where it's like marketing is fundamentally trying to win the hearts and minds of the same, you know, 100 million households or 100,000, you know, potential Enterprise customers. And so it's never done. It's like every day, tomorrow I have to wake up and I have to come up with a new way to do marketing if I'm a marketer. And so there's an early adopter mentality that comes from that because marketers know that they're competing against, you know, other marketers to basically tell their story and also build a human, you know, connection as well with the, with the, with the brand and the person they're trying to reach.

Walk us through sort of like the lifecycle of the marketing ticket and how that has changed with Graal maybe in the last kind of year?

It really depends on the size of the company, right? But at a large company, marketing is, is many different functions that are trying to coordinate together to tell one story. And so a lot of times that's going to start with something like a campaign brief. And the purpose of the brief is really how do we synthesize what we're trying to accomplish across the marketing that we're doing in a way that the email team and the, the team that's responsible for the customer data platform and the team that's responsible for the website can all coordinate together on. So it starts with that brief. A lot of times that's a Word doc, and in many cases it's something that like goes through a couple rounds of revision, and typically that gets attached at that stage to a ticket, right? And so in some cases that's a Workfront task or, you know, something in Asana or something like that. And then you start to have someone in the marketing ops team go and say, how am I going to break this out into all the different tasks that have to happen? And so those tasks could be something like, you know, we need to go create Photoshop files. that go and create different aspect ratio versions of the product on top of a background that's been approved for the campaign. So it's like, okay, we're, you know, launching the pumpkin spice latte. Like, we need to have an image of a pumpkin spice latte, but we actually need to have different sizes. We need to have different settings. We have to have the latte on a table. Right. And so that becomes part of the kind of bill of materials for the brief. And this is actually something that was very surprising to me coming from a rocket background. Like, you know, in the, in the rocket and satellite space and engineering generally, Yeah. You have the concept of a bill of materials, which is like an Excel spreadsheet that's like, here are all of the things we have to buy to go build this rocket, right? Like a list of parts. Marketing teams have bill of materials as well. And it's like, here's the bill of materials of things that we need in order to run this campaign. Like we need these photos, we need these videos, we need these like crops and edits of the videos, all of that kind of material. Then that has to go through and get translated into all the software systems that marketers use. And this was the thing that we got excited about, is we realized people think that a lot of the work is in the content creation, and the content creation is important and it is work. But a lot of the, I mean, it's the same with something like your podcast here, right? It's like a lot of the work is actually in the editing, it's in the post-production, it's in the distribution. And so in marketing, what that looks like is, okay, we have all these assets. Most of the time you have an agency who's produced those videos or assets. Yeah. And they've uploaded them into a Dropbox. And then you've got someone who has to manually go download each of those things from Dropbox, save them on their computer, rename them, you know, edit them a little bit more, and then start putting them into the email system, the website system, the digital asset management system, or, you know, upload them into LinkedIn or Meta to go actually, you know, leverage in an ad platform. And so you end up with literally like 100 steps in the campaign. And then after you actually launch the campaign, you have to go process the data and figure out, did it work? Did this not work? You know, run an experiment. And so you end up with something that really takes like 6 to 8 weeks. Um, and I think the word campaign, again, it's like, you know, when you, when you launch a rocket, you, you call it a launch campaign, right? It really is. It's, it's disconcertingly similar, um, just to see it up close. It's like even, you know, even the Gantt charts, like people use a Gantt chart where it's like, here's the things that are happening, like each swim lane of the campaign, do the same thing with a rocket. Yeah. Yet people, you know, have this kind of mental bias that they're different, but it's actually surprising how similar.

Nice.

Um,

and to pull the thread on the rocket analogy, no one builds a rocket alone. So let's talk a little bit about your co-founders. You were roommates with your co-founders. Um, tell us a little bit more around sort of the decision to get this group together to go on this multi-year journey.

We had actually always been interested in, you know, the idea of starting a company. And, you know, obviously that evolves over time as you, as you learn more. But like, you know, I met Deep, um, our, our CTO, um, and one of my co-founders, freshman fall as lab partners in physics. And I met Anish and Anoop as well freshman year in various settings. And so, you know, it was something that we'd been talking about, you know, at the table in the library going back, you know, basically a decade before we started the company. And I think we felt like, you know, I remember when crypto was kind of, you know, big in our senior year of college, like feeling like we had missed the moment. Like, oh my gosh, like, you know, how could we have not just dropped out and started a company, right? And then of course crypto crashed and, you know, end of 2017 and we felt a little bit better. We felt a little bit better. So that was kind of our journey. And then, you know, the stars kind of aligned to have us all in Seattle at the same time. And so, you know, Deep and I both moved out for work after college. So we were roommates in Seattle. We were having such a good time that Anish decided to move out and join Microsoft. And then Anoop was like, I can't be left out. I also have to come out there. So during COVID Anoop came out, And so next thing we know, it's like the 4 of us are living together in Seattle. And you know that there's kind of, you get the sense that there's a limit, there's a clock on that. Like, how long are we gonna be living together? You know, at some point life is gonna move on, right? And then to have that kind of precondition when ChatGPT came out was really the kind of stars aligning. And it was something where, you know, I think we all felt like this was the moment and it was not gonna be a better moment.

Yeah.

And that was important 'cause, you know, you needed that kind of impetus to just take the leap and start the company.

It resonates so much. Certainly Claudia and I met 2018, like right out of undergrad too. Yeah. And just had this propensity to want to do something together, but you're not sure the form factor.

Yeah,

right.

And then for us, like COVID was the catalyst.

Yeah, completely. And it was this moment of like we were all stuck at home, literally. And even though we weren't roommates, we were like in our same COVID pod. And like it was like a touchpoint to sanity, and it created the space for a podcast to be born. And I think A lot of people ask, and I'm sure they ask you guys now, like, how did you know it was good to build together? Yeah. Have that conviction. And I think there was so many things that didn't get out of the group chat. Yeah. Like before for us, that feels probably similar for you guys too.

The idea of enterprise ChatGPT was just something that felt so big that we were like willing to just leap without looking and, and it was hope it would work out. And I will say like the first like 6 to 8 weeks were definitely like the trough of disillusionment in the sense of like, you know, we realized that enterprise ChatGPT was actually a little bit too obvious,

right?

'Cause the first week we thought we were like so smart for like this idea. You know, it was like, oh my gosh, we're so right. Like, I've never had more conviction that this is gonna be right. And, and, and like, to a certain extent, that conviction was accurate because, you know, we look around, it's like,

yeah,

everything is enterprise ChatGPT now. Um, but I think we underestimated that it was not one company, but probably hundreds of companies. And also, of course, that ChatGPT itself would become an enterprise product.

Right.

And you'd have Anthropic come in and build another enterprise business that, was, you know, comparable or larger on any given day. That was really the, the kind of core moment was it wasn't even the moment where you start the company. It's actually the moment where you restart the company, where you like decide like, okay, here's how I'm gonna go focus. That was actually, I think in some ways the more magical moment is like, you know, 'cause you could easily get to the end of that 8 weeks and just be like, all right, maybe I was wrong. Like I'm just gonna kind of go back to what I was doing before.

Yeah.

Um, but I think deciding that we were gonna focus down, go narrow and find something to kind of build through was like, I think the more interesting moment in some ways.

Well, fast forwarding to today, this earlier this year, you, you announced your Series C at $600 million valuation. So first of all, huge congratulations for all the success. So great over the last 3 years. Uh, um, have some great investors alongside for that journey. Madrona, Insight, Outsiders Fund, et cetera. We're curious, who was the first person to say yes to investing in you all?

Um, well, I think, uh, you know, it, it, I think there, there's always different yeses along the way. Um, but like, you know, Outsiders Fund was the first check into Gradial and, you know, in that kind of 8-week period. And obviously it was a huge impact for us to get that kind of, you know, first yes that we ended up, you know, going forward with. And then from there, you know, you start to just kind of, you know, learn to kind of put your head down and just keep building. Yeah. And I think that's something that I definitely took away from this was like, you know, we started the company and immediately started trying to talk to investors. If I was gonna do it again, I would probably spend the first like 3 or 4 months just like building. Yeah. And have those conversations just a little bit later. Totally. Uh, so that's like one takeaway, but yeah.

Yeah. I, I mean, obviously like this is kind of a fun, extra fun special episode for me because I got the chance to look at your guys' first pitch deck and you sent it to me. I've known and knew probably the longest of the crew of co-founders. Yeah. You met Mike Rastenbaugh too. And just like getting the, the Pano at the time, that was the name of it, kind of deck. And it was an idea related to your point, like enterprise GPT experiences. And I remember thinking like, holy shit, like this is how AI is gonna to evolve. And these guys are thinking like light years into the future. And I, I was an investor at the time, and I was just so impressed with your guys' like conviction where you knew the technology was going to go.

In some ways that conversation, so you were the first investor we spoke with, and I almost wish that you had said like a stronger like, no, you know, like— No, I, I knew it was brilliant. Uh, no, no, but, but like, I, I think like we probably got a little too, too, too over our skis on that, you know, like, Uh, and so, but yeah, in terms of thinking that like we had a really good plan and, you know, that we're just gonna go raise money with a deck. And, and I think, you know, in, in general, like sometimes that works for some companies, but like it probably isn't the right way to plan, you know? And I assume that's gonna work.

Well, it, it's always an ebb and a flow. I know Claudia had like her own experience, of course, like raising as well. And I was grateful to be investor in that company as well as in— we're both investors in Gradyall. But I think, you know, I was like reflecting on this as well because of course having the Madrona team come alongside you guys at your seed, um, is near and dear to my heart being close to that team as well. And I I remember we met for coffee in the lobby of the Four Seasons.

Yes.

When I was in town that March of 2023, and upstairs was the Madrona annual meeting. And it kind of, to me, just feels like a bit of our vision and mission for the podcast of like, you guys weren't yet in the room.

It's true.

And you were literally metaphorically downstairs, like not in the room. And now you guys are a flagship company backed by Madrona.

I mean, it's, uh, yeah, I, I, I think like a lot of things, you know, it's, it's, it's kind of on a 6-month timescale, things can move a lot faster than you expect. And on a 1-day timescale, things can move a lot slower than you expect. You know?

One thing we constantly hear on The Room is that grid operators don't just need more data, they need a faster way to turn that data into decisions. As a founder, I know how painful it is to wait days or even weeks for a report while a real business problem is unfolding. That's what makes Rippling AI different. You can ask Rippling AI, is our sales team at risk of missing quota, and Rippling AI instantly pulls real-time data from Rippling and your CRM into presentation-ready dashboard data showing quota attainment, headcount trajectory, and monthly revenue to quota by region. But it doesn't stop at the insight. Rippling AI can act on what it sees, recommending and building an automated recruiting workflow when open roles and slow time to fill are putting quota at risk, all while following your existing permissions and approval processes. Question answered, action taken, problem solved. Ready to rule your business? Head to rippling.ai/theroom and sign up for exclusive access.

And now a word from our new sponsor, Gradiel. Whether you're building a company, launching a product, or growing a brand, you know the idea is just the beginning. And marketing's bottleneck has never been creativity. It was always the work it takes to get an idea out the door. Gradiel handles the operations so that marketers can stay on craft. An agent will learn your process, work across the systems you already use, and delete point solutions you don't need. Designers will ship on-brand pages without counting prompts or tokens. Product marketers will launch across markets in days, not weeks. And digital experience teams test customer journeys, catch what's not working, and fix it all before launch. Radial learns that your brand, business, and process and brings that continuous context to every workflow. Every task gets done your way with brand, quality, and approvals built in. No more token maxing, slop, or runaway AI bills. More time to think bigger and create the impact people remember. Graydeal, the system of work for marketing. Ashurst Perkins Kewey supports the most innovative entrepreneurs and investors in fast-moving and high-growth sectors, addressing their myriad of legal needs. But the firm doesn't just provide end-to-end legal and business counseling to its startup clients. It also facilitates introductions to key advisors and sources of capital. Ashurst Perkins Coie's interactive website Startup Percolator offers access to programs, resources, and rich dynamic content designed to assist entrepreneurs in their startup journey. To learn more, go to startuppercolator.com.

Well, obviously by all metrics you have succeeded incredibly as a founder and entrepreneur, and it's been a journey, right, of this AI wave and Learning how to tell the story of what you're building with real traction amidst many hype cycles. Do you have any advice for a founder who might be in that seed stage or that Series A stage where they're like, gosh, like I just wanna keep going? Like what would you give for this current climate?

Everything that I've learned, you know, we're, we're still, you know, obviously very early in our, in our journey and there's probably a lot more to learn, but everything I've learned is really that you want to just try to emotionally dampen the like, kind of sinusoid that you feel. Because it's like, you know, otherwise you get very— Oh yeah. Like, oh yeah, it's very hard early on because on the bad days you just can't tell. I like to use this metaphor: am I drowning or am I swimming? And so it's like if you were to look through a telescope far away, you see someone just kind of, you know, splashing around in the water. Are they swimming or are they drowning, right? It's like certain types of swimming, it's like they're doing butterfly, it's like it looks like they're splashing a lot, right? Yeah. But you can't really tell and you have to just kind of be like, you know what, I'm gonna put 6 months in and I'm gonna see where I end up. And I think every day the question is like, what does the next 6 months look like? Can I underwrite 6 more months, right? You're not asking yourself, can I underwrite another decade? It's like, can I, like, you know what, if I don't feel good today, let me just give it 6 months. Yeah. And typically you feel so different even 6 weeks later that you almost feel like it was a silly question. But I think there are a lot of moments where you're like, I don't know what, tomorrow's gonna bring. Like, I don't know how there's gonna be a path forward, especially early. And, you know, certainly I did not know how we were gonna get our first customers, right? There's about 6 months where it was like, we're trying to get our first real customer. And there was a lot of false starts. Like at one point, you know, I was talking to like, you know, some people at the Navy, you know, I was trying to get through the US Navy, right? And, you know, build agents for, you know, automating operations before we kind of focused on marketing. And it's like those sorts of things, you know, For, for a while you feel like, am I just striking out at every single thing I'm trying?

Yeah.

Um, and you have to have the belief that like 1 out of 100 things is gonna work. Um, and I think that that belief keeps paying off for me. Um, but it really comes down to just like avoiding the emotion of it and just being like, I can, I can give it 6 months.

Trying to dampen the emotions, trying to plan maybe 6 weeks at a time. We know that founder journeys are not linear and sometimes those plans don't go according to plan. Um, so maybe tell us about a time in which that was the case. the case and kind of how you reacted and came back from that?

Well, I mean, I think the fundraise did not go to plan in our, in our pre-seed. Um, and so, you know, I, I think that's kind of one example. We kind of touched on that. I think, um, you know, this is, this is a, I guess maybe a little bit tangential, but, you know, we, we basically set out to initially build this kind of copilot-like capability as a service for all software companies. So kind of our path of narrowing was we're gonna do all of enterprise ChatGPT. They go, that's too broad. We're gonna do Copilot as a service. We're gonna basically build Copilot so that you can embed it into your software. So we went to all these software products and we were like, you know, you need agentic capabilities. And before people were even really using the word agent. Yeah. Um, and we, we did some incredible demos, like, I mean, really cool stuff. Um, and we built stuff around like, you know, all sorts of random companies, right? We had, we had this customer, um, that actually built their business on top of Salesforce. So we'd built a Copilot for Salesforce. And it could do things that like you could not do in Salesforce. There was no AI features at the time, right? And so we'd done this before there was even function calling, right? Like OpenAI didn't yet have function calling in the API. Wow. So we're generating these giant JSONs to configure software. At that point in time, you know, we start to get these customer conversations where, you know, after 6, 8 weeks of trying to sell something, you know, we had a customer who, you know, multi-billion dollar software company and they're like, You know, we can't really pay you $3,000 a month. We could do $500, $500 a month. And I think we took the deal. It was like, we have to have a non-zero amount of revenue. Yeah. And so that, I would say that was not going as planned. And I think the moment of pivot was actually, you know, when we did get kind of our first real marquee, like, use case in the marketing side, going and saying, you know what, we're actually just gonna cut all of the other stuff that we're doing. And these are things that we'd worked for like 6 months to get those, those deals.

Right.

And, you know, we said, you know, here's the software, we're gonna just give you the code that we built here because we know that in the next 12 months that code's not gonna be that worth, like, it's gonna eventually, this is gonna be easy, right? So we're gonna give you what we built for now. We don't wanna leave you in a lurch. Um, but we need to go build a different business. Yeah. Um,

speaking of that kind of like evolution even of like the product itself to be an agent, I mean, that is not, to your point, been vernacular that we're using very long, even though it feels a little silly. It's been like a year of us using that word almost every day. I'm curious because the Gradio product has evolved so much, like how you thought about architecting that agent and kind of like that, you know, for someone who's listening who might not actually totally understand like what even your core user would use, like how have you guys thought about, uh, kind of empowering a non-technical user, um, with a very technical solution?

Initially, we did not even have a UI for Gradio. And so, 'cause, because when you have like a 4 or 5 person company and you're trying to serve Fortune 500 customers, you have to find some way to narrow yourself down and get really focused.

Right.

And so the first thing we did was we actually just leveraged the ticketing system like a Jira as the UI. And so you would tag Gradial, um, you should be like @Gradial. And it's actually in some ways very similar to what you see now with like, you know, the, the like bots within Slack, right? From like the different providers. But like that was before that was really a thing, right? So you would trigger Gradial from like a comment in Jira. And then we would come back with a link that was like a link to the webpage that was done.

Cool.

And so, you know, that's kind of where we started. And then very quickly we realized, okay, if we wanna really build a durable business, we actually have to have an application where you do more than that because there's things that you just can't do. Like for example, you can't review a preview of the content or a preview of the image that you're trying to like, you know, to manipulate. And so then we built kind of, Gradle v1.0, 'cause the first thing we actually, we had a different name at that point, but we, we had to change the name. Um, and so, you know, the, the kind of Gradle v1 was this web application that would sync up these tickets, but give you a preview of the web app.

Yeah.

And then, you know, as the models got a little bit better, you know, we realized that the, the way that we had architected the software to make all of this automation work with kind of the less intelligent, like GPT-4, Um, was actually standing in the way of the more intelligent models that were coming up. Right. And so there was this, you know, and this is a lesson that we kind of had learned over and over again, but have, you know, really taken into, into heart is like, you know, almost all of the infrastructure you put around your harness becomes eventually tech debt. And so, you know, it's all about simplicity. And so, you know, that was kind of the next big transition was going from like what was in some ways a, semi-deterministic series of decisions where the, you were using the models to make the decisions, but you had kind of, you know, trajectories of decisions that were supported going from that architecture to something where it was like, actually the model makes every decision.

Right.

There's no, there's no defined trajectory and the model can make a series of decisions. And then, you know, then there was kind of Gradial, you know, like the 3rd kind of iteration of this. I think we called it Gradial V2, but it was really kind of the Gradial V3, which was like, leaning into like truly agentic capabilities. Um, and this was about the time that Claude Code first came out. Um, so if you think about that, that was end of '24, I think going into early '25. Um, that was when it first launched. Yeah. And it didn't really work at that point in time.

Uh-huh.

'Cause like you think about when did Claude Code have reliable 30-minute sessions? Sure. That really kind of came with like 4.5, in my opinion, kind of like, you know, November, December last year. Um, but the concept of what Claude Code represented, which was like extremely agentic behavior, basically was something that, um, you know, we kind of arrived at around the same time. And, you know, obviously we're, you know, partly inspired by what we were seeing in the coding space, but also like, you know, in some ways it was easier and harder in marketing. Easier in the sense that like, it's not as open-ended as coding, but harder because there were a lot of capabilities that you have to bring to the model that they're just not good at out of the box.

Yeah.

they're still not very good at, you know, for example, uh, manipulating and designing a Photoshop file.

As like this week, I mean, Astra has launched to the public. Um, People-5 obviously, um, is incredible. How have those sort of like new models pushed your product kind of in the last couple of months?

Well, I think we spent a lot of time thinking about what's the most useful thing we can do for marketers, um, in 2027. And I, you know, I'm gonna have to update this now and say 2028 very soon. Um, but it's always like, what, what are the things that would be amazing if we could do them? that you just can't do today. And, you know, where I think that the, you know, the direction I see is spending a lot more time thinking about long-duration marketing tasks, right? And this is like end-to-end optimization of a campaign over weeks or months.

Wow.

Right? Because, you know, if I'm running a lot of campaigns, like for example, like, you know, an anniversary sale campaign, right? That's not something that's just a one-day thing. I'm constantly tuning it. I'm saying, you know what, what email should go out today? to help drive conversion? What are the things I should do on my website? What are the things I should do, you know, with my SEO and paid, right? And so I think what we've been able to start to unlock is to say, you know, there's an aperture of marketing tasks that are becoming possible, and we like to focus on the things that are barely possible or the things that are still only 80% reliable with one of these foundation models.

Yeah.

In the most idealized harness, right? And so then it's a question of how do we eke out performance using our marketing harness? which is at the core of basically what our product is, is like, you know, it's a marketing harness in the way that like, you know, something like Claude Code is a coding harness, right? Sure. And that's the push, right? And I think that's been a continuous push from all the models. I think where I see Astra and Fable 5.1, like they have really unlocked a lot of kind of computer use and asset manipulation type workflows that I think expand some of the aperture. And I'll give an example, like, you know, we had a, we did an experiment back actually in December of last year, which was building product videos, not using an LLM to generate the video, but using an LLM to edit the video. And the idea was like, you know, there's a lot of video production that, especially for B2B, is actually more text and animation based than it is like image based. And so you could actually have the agent go construct frame by frame, animations using HTML. So those sorts of use cases have gotten much better. And what that lets you do is then you can bring that as a capability into a larger campaign, because if you're trying to run an end-to-end campaign, you're basically limited by your rate limiting step.

Yeah.

And a lot of times there's still many of these rate limiting steps like, oh, we need a motion designer to come in, we need someone. So in the podcast example, it's like, okay, great, we have all of this, you know, content that we've edited, but we actually need the animation for the intro frame, right? Or the

logo files for our sponsors.

Your logo files for the sponsors, and we need them to animate across the screen or do some sort of transition. Um, so we've been able to build the right tooling into the harness to let these models start to do some of those use cases because they're getting to a point where, um, they don't just crash out.

Wow.

Um, where they used to.

I mean, it's so complex actually to peel back the layers of what needs to be in place for this to even work. And then of course, it also assumes that you've hooked into the systems of record inside of the enterprises through which You're operating Gradio, operating your agent layer, and you have some incredible public enterprise customers like T-Mobile and AWS that you've talked about how you're working and helping empowering their marketing teams. But that is a complicated enterprise sale. And so kind of speaking to the heart and this idea of like secure agents as well, how have you been able to navigate the procurement cycles and actually like bringing Gradio to the enterprise?

To me, that's actually one of the really fun parts, honestly, because it is, it is quite an interesting challenge. Um, you know, one thing we did early on is we built Gradio to be omni-cloud, omni-model, like very much an open ecosystem. And we want to connect with every marketing piece of technology that someone has in their enterprise. And that was very purposeful. Um, and so, you know, it lets us be relatively flexible in the same way because it's a harness, it's not a model. We're not selling you a model. We're not selling you just like, you know, a very narrow fraction of the AI market. Sure. We're saying like, here is a method to do marketing that can work with all of the systems that you have, right? And then the security story, it can be actually, you know, basically at that harness level, which is like, you know, here's where the harness lives. Here's the information that's part of the harness. It's all in a cloud that you trust, you know, one of, you know, a few of those hyperscaler clouds that you can trust. And we actually keep it in a, you know, an isolated environment. We're not training on it. We're not taking that data and using it with, you know, to go train an LLM or something like that.

Sure.

So those are really important, I think, for getting enterprise procurement through. But the other is like, you know, having like very clear security boundaries of this is how we interface with your other systems.

How do you feel that you are able to balance the relationship between like the creative buyer of a marketer with the execution orientation of technology's advantage?

We all want to scale the amount of human connection we can make, right? If I'm a brand, I wanna tell more of my story.

Yeah,

not less. And I wanna tell it better and I wanna tell it more effectively. And I, and I think most marketers at large companies have dealt with the fact that their ambitions have to be par— like pared back by the capacity that they have available, right? Like it's a, it's a resource constrained, uh, you know, team. And, and so, you know, what Grego's let people do is say like, I'm going to put more of my time towards the part of marketing that I actually wanna do, right? Cool. Everyone is like, everyone is choosing to do more. They're not choosing to do less. Why would you choose to do less? Because again, it's a competitive sport. It's like, If my competitors are buying Gradial and then they're 10x-ing the amount of campaigns that they're able to produce and they're able to be more ambitious as a marketer, um, then I need to do the same thing, right?

How do you think about pricing as sort of outcomes evolve? And, you know, there's more and more subagents that are getting embedded into workflows. Like, are you thinking about sort of new pricing primitives to make sure that your end customers are paying for the value that they're getting, especially as like margins change? Yeah.

Pricing is gonna be a very interesting and moving target over the next couple years.

Yeah.

So like, you know, there's been some companies, like take Harvey for example, at some points or another they've had a seat-based model, right? I think it's very clear that in the future, the eventual final output is going to be a, like a consumption-based model for every AI company. Now the question is what type of consumption-based model? And is it by the hour? Is it by the token? I'm pretty sure it won't be tokens, uh, in the limit. I think that the more likely thing is you're going to have some other discrete, um, measurement of the agent, whether it's hours of agent capability or number of parallel agents. Yeah. Um, especially in you, as you get into a world now where like, this is kind of my hot take, right? But, you know, agents are increasingly getting shipped with a computer. So if you're using the Gradle harness, you're not just getting a model, you're getting a model with a computer with marketing infrastructure around it. And so you actually want to sell that not by the token, but as a packaged good. And then you say, well, how many agents do I need? Like, it actually is something that like matters, right? Because if I have 1,000 agents working, I actually have to have 1,000, you know, effectively virtual CPUs. Yeah. Each of those needs a web browser. Those web browsers, you know, I mean, if you use Google Chrome, it's like sometimes your laptop crashes, right? Because it's like 4 or 5 gigs of RAM. So there's real cost associated with the compute side, not just the tokens. And this is true for OpenAI and Anthropic as well. I mean, there's obviously companies that have been built in that kind of sandbox space. Yeah. But the point for pricing is that you're gonna end up in a world where eventually I suspect you're gonna be paying for those agents in something that probably resembles by the hour. Yeah. Um, that's not our pricing model, but that's like where I see the industry going. And so, so then there's different ways that you could take that underlying hourly cost and package it. And that's, I think, kind of the question that companies have to ask right now. And so you could do it by the token, in which case you say, I'm going to bundle that compute cost into the token cost, or you could do it by the output, right? So if you're like in the support space, you got the outcome-based pricing. I think in marketing, um, you, you have to think about what is the outcome that you're trying to achieve. Um, and we've tried a couple different things, but I think what we're excited about now is coming up with a model that is more like, you know, more limitless marketing. Yeah. Um, where it's like, you're not trying to basically, like, I think the AI companies should be responsible for figuring out the tokens on their side. Yeah. Right? Like marketers shouldn't have to think about tokens. And especially as tokens get cheaper, that optimization of the efficiency of tokens should be one of the incentives that a vendor should be going and doing for you. Because token maxing, I mean, like we see this with our internal use cases, right? It costs a ton if you just have people using tokens.

Just like to add one anecdote, I work now in marketing at Vercel, and I was on a call with another customer potential, like in a call with other marketers talking about this, and And one of them was so like upset on the call that their Contentful credits were running out.

Yes.

And obviously Contentful recently got acquired by Salesforce, and I know we're here for Dreamforce week, so like it's an interesting time to be thinking about it. But it was like almost to your exact point, like this adverse reaction of like, wait, like you're— you shouldn't actually care. Like, that's like sad that you're worried and stressed about your Contentful credits running out. And like, why would an end product like want to create the dynamic in which their end user isn't actually being able to use the product? So that's an evolution for sure.

Well, I think it comes back to mis-sized expectations. Like the credit model is fine if you always use exactly 100% of the credits you expected to use, right? When you sign the contract, that's almost impossible, right? And so then the question is, well, how do we design a contract where I can plan my budget 12, 24 months in advance? And that's very tough because the AI costs are moving so fast and things are changing fast. But I think that's one of the things that a company like Radial should be doing for you, right? That is a customer service that we can offer is we can say, we are going to give you this amount of marketing for sure, guaranteed at this cost, and you can plan your budget around it and you can bet on this as like a, you know, take it to your boss and then it's our problem to solve. And I think that's fundamentally like, we've seen that be something that has been important and like helpful for, for our customers.

Doug, you've walked us through sort of the early days at Dartmouth. Yasu. building Wi-Fi to agents and like really how people work in the future. So thank you for this amazing journey today. What's next for Gradiel?

Well, there's, there's so many parts of marketing that, you know, are really interesting use cases that we've got to continue to figure out how to serve. I'll take an extreme example, which is, you know, the importance of events in marketing is actually growing a lot, as I'm sure you both know from, you know, the events that you guys put on. And I think that, you know, there's a ton of manual work in that. And I'm not just talking about like, you know, going and putting up like the banners at the booth at Dreamforce or something, but like, you know, booking the restaurant, you know, contacting sponsors, talking to partners, right? The logistics of it. And so I think there's a lot of use cases in marketing that we are really excited to unlock. We want to continue to build, you know, a harness that kind of is the enterprise marketing brain.

Yeah.

For our customers where it's like, You know, you can, you can give Gradio the prompt. Like, I think about what I call hero prompts, which are like, what are the different ways that you could prompt an agent and what would you want it to do? And that's how I think about product in 2026 is like, you know, what are the prompts I want to enable in the harness that don't work today? So one of those is like, go plan Dreamforce for me. Like, I would love to be able to say that to Gradio for, for Gradio, because we're, you know, we're obviously going to be a sponsor this week. And, you know, there's a lot of pieces that we already do. Um, like we can do web, we can do email, we can do social, we can do paid media, right? Um, but I'd love to be able to extend that into even more parts of that kind of planning.

We would love that too. Seriously, sign us up.

And, uh, what's next on the personal front?

Um, well, I'm getting, I'm getting married this year. Um, and so that's exciting. And, you know, um, yeah, uh, good, good to juggle with the startup.

Yeah.

Yeah. Always fun.

Yeah. That's incredible. Congrats. We're so excited for you. Um, and thank you for being with us today. As we come up on time, we'd love to ask our hero question, and that question is, who is a woman who has had a profound impact on your life and career?

Well, you know, I touched a little bit on how impactful the SpaceX job was for me. Yeah. And the story there was not just that I applied and, you know, got in. It was, um, actually that, uh, freshman year I approached, um, my physics professor, Professor Millen, Robin Millen at Dartmouth, um, who was incidentally the same professor that Deep and I were taking the lab with when I met Deep. And I asked Professor Millen to sponsor doing this fusion project, um, which, you know, it really was kind of way out on a limb. Like, you know, typically most of the undergrads are joining a professor's lab and doing something on their project, right? And I didn't ask to join her project. I asked, can you sponsor this like random side quest project that I want to do? And, um, and, you know, she, she signed up to do that and that was incredible. And then I was able to, work on that my freshman summer in, you know, under her supervision at Dartmouth, um, and then actually raise money for it, um, the next year. And so, you know, she sponsored that project for 4 years. And, and I mean, it really was, um, something that was, was quintessential towards, um, you know, ultimately getting that job at SpaceX. Wow. Um, like, because for sure that was the thing on the resume that, you know, caught the radiation team's attention. And, you know, I've just, I'm really grateful to Professor Millen and, you know, She was an amazing advisor all through that process.

It's incredible to see how those early relationships snowball and how that impact has really like transformed your career. Uh, well, Doug, thank you for joining us in The Room today. It was an honor to have you here.

Thanks for having me. It was a super fun conversation.

Thank you so much for joining us at The Room Podcast. If you want more from The Room every week, subscribe to our newsletter at theroompodcast.com. We'll be back next week with a new episode and inspirational guest. Tuesday, 10:00 AM Eastern, 7:00 AM Pacific. See you in the room.

And now a message from one of our sponsors, Rippling. One thing we constantly hear on The Room is that grid operators don't just need more data, they need a faster way to turn that data into decisions. As a founder, I know how painful it is to wait days or even weeks for a report while a real business problem is unfolding. That's what makes Rippling AI different. You can ask Rippling AI, is our sales team at risk of burnout? Missing quota, and Rippling AI instantly pulls real-time data from Rippling and your CRM into presentation-ready dashboard data showing quota attainment, headcount trajectory, and monthly revenue to quota by region. But it doesn't stop at the insight. Rippling AI can act on what it sees, recommending and building an automated recruiting workflow when open roles and slow time to fill are putting quota at risk, all while following your existing permissions and approval processes. Question answered, action taken, problem solved. Ready to rule your business? Head to rippling.ai/theroom and sign up for exclusive access.

And now a word from our new sponsor, Gradiel. Whether you're building a company, launching a product, or growing a brand, you know the idea is just the beginning. And marketing's bottleneck has never been creativity. It was always the work it takes to get an idea out the door. Gradiel handles the operations so that marketers can stay on craft. An agent will learn your process, work across the systems you already use, and delete point solutions you don't need. Designers will ship on-brand pages without counting prompts or tokens. Product marketers will launch across markets in days, not weeks. And digital experience teams test customer journeys, catch what's not working, and fix it all before launch. Radial learns about your brand, business, and process and brings that continuous context to every workflow. Every task gets done your way with brand, quality, and approvals built in. No more token maxing, slop, or runaway AI bills. More time to think bigger and create the impact people remember. Great deal. The system of work for marketing. Ashurst Perkins Coie is a leading full-service international law firm known for providing high-value strategic solutions for startups and investors. Lawyers at Ashurst Perkins Coie support the most innovative entrepreneurs and investors in fast-moving and high-growth sectors across the country while addressing the myriad of legal needs that arise when fundraising and scaling their businesses. The firm's tailored legal services are informed by understanding of each of the client's discrete needs. To learn more about Ashurst Perkins Coie, visit their website, ashurstperkinscoie.com.