Monetizing SaaS: Episode 44 with Kent Keirsey, Founder/CEO of Invoke

Kent, um, I've gotten to know you over the last, I think, 3 years, and I learn so much from you every time I talk to you. But you've had this incredible career. Studied economics at my wife's alma mater, Auburn. You cut your teeth in product and QA for an interesting company that I— you've got a great story with UDA. Then you moved to Atlanta where you did a 3-plus-year spell with Ernst Young, and then you moved into product leadership roles for like many of the best Atlanta tech startups over the last decade. Atomized, OneTrust, Gather, Greenlight Universe. It's like an impressive list. And then a couple years ago, you founded Invoke, and you were the CEO of that company. So I'd love to hear your story in your own words, man.

Yeah, it's a— it's been, it's been a long journey. I think, um, the way that I described that series of events is largely just saying yes to opportunities that present themselves. I don't know that many of those early career moments for me were intentional, where I like set out to go do something. So the first company that you mentioned local Auburn tech company. Just, I mean, 2010, we were still shipping CDs to people with executables on it to install this software and people were buying it. It was like construction software, right? It was like right before SaaS really took off. And so a lot of like, a lot of my time was spent like helping sell the software, but also support like a local installation of this executable that had like a SQL database on device, right? It was like on their computer. And that is just like support nightmare, uh, is like dealing with different computer environments, uh, which is funny because now we're doing the same thing with Invoke and our open source installation, but I digress. That, that was just like a job listing that I found. I wanted a job and I like, I knew about computers and I was a big nerd. And so like I got it and it just kind of set me off on, I work in software, I work in technology. EY was, I went to a career fair. I almost didn't go to that career fair too. Like I remember like, ah, should I put on the suit? It's kind of like a nice day outside. I guess I'll go. And then I went and like, lo and behold, that became the next few years of my life was doing consulting at EY. And then it was just a series of like, hey, I know somebody, there's a cool role, jumped out to do product management. I didn't even know what product management was when I got hired at Atomized. They're like, hey, you know, we're actually hiring for a project manager role. And I came in and I did the interview and the CTO is like, you know, You kind of feel more like a product manager than a project manager. I was like, okay, that sounds cool. I guess I'll do product. And then from then on, it was just, that was what I was doing. I was in product management and, you know, fell down the rabbit hole on product management, started Product 404, now Product ATL, um, as a community and just super passionate about building good products. And that I think is, uh, ultimately how I went around the Atlanta tech scene and, uh, was able to do a lot of different fun things in different industries, uh, and then stumbled stumbled into generative AI right before ChatGPT became a thing. I started contributing to open source, helping like product manage an open source project that was evolving around the Stable Diffusion ecosystem. And a few months later, it was like, okay, well, this is a thing and we've got something. Let's go make it a real business. It's kind of how I stumbled into Invoke as well as an opportunity that I said yes to.

That's awesome. Um, I want to dig into a couple of those companies. Like when you look back at Gather, What was the most important lesson you feel like you learned from that company?

Oh man, um, I think Gather was really interesting. There's, there's a lot of culture kind of questions that can raise when you're running, running a business and kind of like building teams. And I think Gather was one of the most impressive cultures I've ever seen at a company as far as like how people worked together and the type of really powerful relationships that can be built in a company when you, when when you gather people together in a certain way. And I think that company was really led well in that regard. You know, I think it became a lot harder to maintain that culture as it shifted from ownership. Vista became a big part of that conversation. And so there was a little bit of like, as it scaled, running into like, how do we make this work and still also achieve some of these goals that we have? But it really was from a, from its core, a very strong team and one that I still look back at as like maybe one of the best teams that I've ever worked at. Everyone that I know looks back at their time at Gather as like a very, very, um, I would say memorable experience and a shining spot in their career.

Yeah, formative. And every time I meet somebody who worked at Gather, I hear the same thing. It's— you've got like such a strong alumni network from that company.

Yeah, it was great.

What about Greenlight? I mean, that's one of the most successful companies ever in Atlanta. What, what did you take away from that experience?

Well, Greenlight's an interesting, it's an interesting one. Um, there, I think Greenlight had a big run-up alongside the COVID lockdown. Like everything in the world, uh, shut down and everyone thought that everything was going digital. And so it benefited a lot from that kind of hype and everything that kind of went in parallel with that. I think from a, from a lessons perspective, I think strategy and really knowing kind of what your core is and how to maintain that in the long term. I think there are a lot of lessons to be learned in, in Greenlight's journey there as far as like how to maintain direction and set that strategy from, uh, an early point. I think they struggled a lot with that, and that was kind of one of the pieces that I really focused on a lot when I was there, was like, how do we focus our strategy? How do we make sure that we are We're really understanding where our opportunity is, where our competitors are going, and how do we like chart that course to be positioned well for that? Because I think the biggest risk is, especially when you get to that point, is trying too hard to be everything and losing focus on being the thing that you can actually win at. And that's, I think, one of the big lessons for me from—

And they were fairly big when you got there. I mean, they were— they would have been significantly bigger than Gather was, right?

Yeah, I mean, I would say it was right after the Series B, I think, um, and that was when they were starting to hire for different teams and kind of growing out the R&D org pretty significantly. So I was there from that through the run-up into post-Series D. So it was, was a lot of growth and a lot of expansion. The COVID the COVID kind of, uh, situation happened while I was there. And so I got to experience like, how do you transition a company that is actively scaling into a fully remote surprise environment, right? And that was, um, that was tough. I mean, I think honestly it's like, how do you maintain that? What was, I think, crazy for me is we, we actively kicked off a project to rewrite the underlying like application almost from the ground up right before COVID started. Like that, that happened right at the start of COVID And so It was a, a real challenge to get that to the finish line because of the complexity of the project, because of the ambitions of the project. We were not only launching a new application, we were also launching the first investment product that was really targeted to like help kids learn how to invest. And so it was a, it's a behind the scenes massively robust investment product. It's like all the, the legal loops that you need to jump through, you have to do that plus adhere to like child protection and privacy stuff. So it's kind of like more complicated than a normal product. But luckily had a really strong team, one of the most skilled teams I've ever worked at in kind of managing that chaos. And we, we succeeded in doing that through COVID, which I, I think speaks to, you know, the strength of building teams and kind of how you do that. And what, what is completely separated from whether you're in person or whether you're remote, like what does it mean to be a team and how do you build good teams? I think that's was a formative experience on its own of what you can do and how you can do it in a remote environment.

It's a compelling moment in time to just like park on for a minute because of COVID because it's sort of the height of, you know, a 10-year surge in valuation. Capital is probably at its cheapest in a long time. Lots of companies are doing what Greenlight was doing at the time, which is bringing a lot of people into the business. And I'm curious if you feel like you grew as a communicator in that time, given it sounds like you were trying to persuade a business to concentrate in the context of lots of shiny objects, lots of new product potential. How did you grow as a communicator if you felt like you grew as a communicator?

I mean, yeah, that's a good way of putting it. I mean, I think I, that was probably the job in a lot of ways. You know, I was leading a team of product managers communicating across the organization. At one point was effectively like leading the product organization, so communicating as the direct point of contact for marketing and for the C-suite. It was a lot of different perspectives and a lot of different ideas of what should we do or how should we do different things. And I think this is an interesting experience having a storied product leader who's the CEO of the company. When you have somebody who's like that into product, they want to be in the product and kind of part of all the product decision-making. And I think Managing that and kind of helping facilitate so that it's a win-win for everybody, that we actually achieve what we want to do, that we get people's input and we help them understand the context that they may not have for why we're making certain decisions, right? I think that's, that's where a lot of the communication challenges come from, is everyone's working in different contexts at scale. They don't have the whole picture. Very few people have the whole picture. Sometimes even the CEO doesn't have the in-the-weeds picture, and you are trying to kind of like bring that context up package it in a way that's like accessible. So it's not like a 7-page essay that's never going to get read. It's like, this is why this matters. This is the one kind of insight you need to have and really helping everyone feel like, A, their input's valuable and B, they're heard so that we can get to the finish line. And a lot of times that's like compromise, right? It's like that thing that you really care about doesn't ultimately matter. So let's make that button like, fine, the button's blue. We think blue is the perfect color for that button. It speaks to people, whether that's coming from like some person on the marketing team or the C-suite, whatever it is, it's like, okay, these are the things that matter and these aren't the things that matter. So if it doesn't matter, let's just like move on and we can like rapidly make decisions that way. So it definitely was like communication, stakeholder management. How do you navigate an organization that's like scaling that quickly and also still wants to be like tightly coupled in a lot of ways? Very strong lessons, very strong lessons from that experience.

And then you go to Universe and tell me if I've got any of this wrong, but you're, you're effectively looking for consumer software applications. You're buying those, those companies, you're trying to build a platform that you can as a service layer to serve all of those companies and help each company grow independently with their own GM. I think you're running product for Universe, but you also end up finding yourselves in kind of GM roles for some of the portfolio companies.

Yeah.

I mean, important lesson from this one.

Yeah, I mean, I think that was just a really exposure to viewing businesses a little bit more in a portfolio sense, right? So like I've always up until that point been in the business and, you know, working at EY, you definitely like you're serving clients, but you're almost like helping them solve specific problems. And so EY was good because I got to see a lot of different large, large companies and how they operate and the individual problems and the technology organizations that they run into. But viewing businesses almost as like, you know, part of a portfolio, and they're all performing on different dimensions in different ways. But in some sense, you're kind of like, each business is its own special snowflake. It has its own landscape, its own opportunities. But at a portfolio level, you kind of like take a lot of that away, and it's just you're comparing— you're trying to compare apples to apples. Like, how does this business perform relative to that business? What are the right decisions for the portfolio? How do we structure the underlying organization at Universe to make sure that its strengths support its portfolio companies in the same way? Right? Because you, you don't really want to have to build 7 different ways of supporting. You want that to be kind of like unified. That's the whole theory behind building that kind of like foundational platform layer. So I think for me, that was like, that was a good lesson is just like looking at it through that lens. I've always been more of a bottoms up person. I think about the product and I think about that moment of value and like How do I get a person to value? And all of the other stuff for me is just like BS that I have to deal with. My, my passion is like making products that people love, right? Like, that's what I really get excited by. But yeah, you have to make a business around that. You have to like make a profit, you have to price it, you have to package it, you gotta, you gotta at scale deploy that and service it over a long time horizon, right? So you have to understand all of that. And I think for me, Universe was like the top-down look and really connecting all of the dots all the way up into that kind of like investment type perspective. It was really, really formative perspective for me from, from that experience.

Awesome. And so while you're there, you start to, start to tinker with generative AI, right?

Yeah. So, I mean, I think, let's see, the, I'm a bit of a like creative spirit. I do a little bit of everything.

Can't tell from your background.

The long hair, the background, you know, I've always loved tinkering with technology. Like I have been fiddling for a long time. I'm a big like video gamer, right? I used to grow up playing computer games and I would just like fiddle. That's where my technology background comes from. But also really like love the creative technology, right? I've— my parents really kind of like supported that as a kid. It was like, whatever you want to do, go do it. But my dad specifically was in architecture and I grew up around like Photoshop, Corel. I got Macromedia training when I was in middle school to go build websites and stuff like that. It was just like whatever creative thing I could get into, I really liked doing it. And 2022 was interesting because there was kind of like an emerging set of technologies where it wasn't just like hallucinatory images. I had seen stuff since like 2018, 2019 with generative adversarial networks or GANs. They were like creating these like hallucinatory images of frogs and it's like, wow, that's super trippy, but you know, no practical use case. But 2022, you started to see stuff like DALL-E 2 come out. And so I started playing around with DALL-E 2 and it was, it was generating like images. It was like, whoa, this is like actually working. This is great. And all of that stuff was happening on services. And I think DALL-E 2 was charging you almost like I think it was like 15 cents an image or something like that when it first came out. And it's like, wow, that's pretty cheap. And then like later that year, Stable Diffusion comes out and it's an open source model. Everyone has access to it. You can run it locally on your computer.

Right.

And that was the big unlock that, that was where for image generation, people now had access to the tool. People could tinker with it and experiment directly on their own hardware. And that's where you saw this like flourishing of image generation. That, that is where it came from. And so my experience there was, you know, I think it was Hacker News or something like that. I was, I was perusing and saw a post about, hey, you know, Stable Diffusion just released, runs locally on your machine. So I'm like, well, I got to try this because I got a gaming computer. I can run this, right? And I look around, there's like a list of all the repos that you can use to run it. And I click. I don't, I don't know what drew me to this one, but I click on one and it is a command line interface. It's not an application. It's just like a simple script, super easy to install. And the guy who put this together is a man by the name of Dr. Lincoln Stein, who is unbeknownst to me at the time, the head of adaptive oncology at the Ontario Institute of Cancer Research. He does like ML for cancer research. And so he's like running that program, but he's also interested and fascinated by this stuff. And he's like, started that script. So his repo just becomes a flurry of activity of people trying to make it a little bit better and use it and all that kind of stuff. And that's where, I mean, within the span of a month, it turned from a CLI script into like, hey, there's somebody building a full React TypeScript app for this that's going to sit on a web server. And I'm like helping coordinate the community growth efforts. I come up with the name Invoke instead of Lincoln Stein's Stable Diffusion repo because we had enough people. I was like, okay, we need a name. Uh, and so I like, we, we end up becoming this kind of like hodgepodge group of collaborators and contributors in open source that are building this, this vision of creative control for these models. And by December of that year, we had like the world's first canvas. Adobe's got like 50 people following the repo. Microsoft's following the repo. We got like all these people who are looking at what we're doing as kind of like pioneers in the space. And we started having conversations with the AI companies in the space. It was just fascinating how quickly we were plugged into what was happening everywhere.

At this time, is it fair to say this is— this Invoke at the time is just an open source community that has—

Yeah, it is just— it's not a business. It has no revenue. It is just making an open source project. I'm doing this in my free time. If you ask my wife what that was like, it was like we had just had our second son. So he was born in July, Stable Diffusion released in August. And so like, he's down for bed, our kids are down, and I'm like tinkering until midnight, 1 AM. I'm making videos, teaching people how to use it. Like, we have this just like this passionate group of people that are building this. And it wasn't until I think I got back from holiday over Christmas break in January, we had started to get conversations, like people asking to hire our team and like take the repo and go like use it. And my realization was like, man, this is like really taken off. I've always been relatively risk averse when it comes to starting a business. I started a game company when I was in college and it like didn't go anywhere. So that failure was always been like a good lesson of make sure that you have the right people and the like right traction before you really go all in on something.. And I'm sitting there in January with all of these people reaching out with the YouTube video that we put out on the canvas with like 200,000 views. And I was like, I don't know what traction is if this isn't it. Like, I don't know what I'm waiting for to go do this. And that's where, you know, begin to have the conversation with Nick, the CEO at Universe, who's also CEO at Gather, and kind of sharing what I'm seeing and what I see as the opportunity. And that kind of led into, you know, him helping incubate the project out of Universe and starting the company and really like getting it going to the point where we raised our seed in June of that year. So it was a really strange experience for me, but it was kind of like, this feels like it's the right time and the right product and the right moment in history to go build something., and we have a different philosophy and perspective on how to build it. We, we see, whereas most companies I think are selling the dream of automate everything and you don't have to be involved. We have this kind of vision of, okay, well, what does AI look like when it is driven by humans and when you do have creative control and when you can use it as a companion rather than as a replacement for art and creativity. And I think that has resonated quite a bit with the customers that we work with. I mean, we effectively do all of the capabilities that you would expect from like an automation system, but we allow the human to get into the weeds and creatively control that output. And that is ultimately something that I believe is going to be necessary for most businesses. We can get into like copyright if that's an interesting topic, but there's like a whole host of reasons why you want a human in the driver's seat. Like for a lot of this stuff, you, you do not like automate this, especially when you're talking about creative stuff like games and movies and entertainment, taking the AI system and saying, we're going to just print movies and have zero human creative involvement is patently absurd. It's just not going to happen. Those movies will suck. Maybe they'll like fill out the bottom of the bucket in Netflix that people scrape when they've run out of anything else to watch, but it's not going to be like a good movie. It'll be cheap. It just won't be good.

Go ahead. Take us into, into Invoke's product today. You're kind of dancing around it, but what does it do? What is it? Who is using it? I know you've got users from many different kind of places, but what do you want people listening to this to understand about what the product does today and who is getting the most value from it?

So let's take a step back and talk about like the AI model. Right now, there are a lot of services that build an AI model. They build the application layer to deploy that. As well, right? So if you're switching from company to company, you're changing model, you're changing interface, you're changing all of that. The other piece is when models are not open, when they're like behind the closed wall, like Adobe Firefly, for example, if you train that model on your content, if you tweak it, you tune it, you customize it, you don't actually have an asset. You have a customization that's sitting inside of Adobe servers, but you can't take that and go somewhere else because it's all trained on top of Adobe's model, right? And you're never going to get access to the weights for Adobe's model. So what a lot of enterprises have started to recognize with AI is they want to own the weights. They want to own the resulting asset of any training that happens. They want that to be a business asset and something that they can rely on for operational stability. And that's why you're seeing so much happen in open source AI. On the large language model side, we see things like Llama get a lot of attention. It is because you effectively have open weights. You can go use these, customize them, tune them, make them really good at doing a certain job in your business, and you're the one that is dictating how and where those get controlled. In image generation, you have the same capability of customizing it and fine-tuning it and making it your own, but you have to have open model weights and you have to have something that allows you to use those. And so what Invoke is really designing more than anything is the model agnostic application layer. Whether you are using Stable Diffusion or Stable Diffusion XL or FLUX, which is a big model, or now as of yesterday, Stable Diffusion 3.5, those model architectures are changing, but your interface and your way of controlling that shouldn't really have to change. You should have the flexibility to use whatever model you want to customize and fine-tune to do that. And because we're model agnostic and because we, we're not training on our customers' data, we don't have our own model. We're just helping them deploy it, our hosted product is effectively a secure place where they can deploy their own custom models and get all of this kind of like application layer SaaS goodness, like projects, role-based access control, SSO, all the like boring enterprise stuff that nobody really wants to build on their own because it's just like somebody's already built it. That's what we sell to enterprises. Our product, the Core Studio itself, is open source. It's Apache 2 code. You can run this locally on your machine still to this day. And get all the same power and control with the custom model. But what enterprises value is actually all of the stuff that deploys it at scale. It's all the stuff that is like boring SaaS stuff. And that's how we're kind of monetizing and building the business is we handle that piece, but that enables us to make kind of a more open and widely accessible set of creative tools. So ultimately achieving our mission, but also building a super strong business as well.

Super clear. So are you seeing a combination of enterprises using the product for their marketing, studios using the product for their films, and individuals using the product for their, their small solopreneur projects? Is it all of the above? Is it one more than the other?

Yeah, I mean, definitely all of the above. I think the thing that's interesting though is the wide expanse of open source software has— it's exploded in the past 2 years. And so on this kind of SMB side, because it's free, if you download it and run it locally, a lot of SMBs don't really care about SSO or scale or deployment. Like they're willing to take the, like the hit of I have to manually install this like slightly complicated Python application on a computer to get it to run. They're okay with that and they're willing to do that. And I think that was probably one of the lessons that we learned early on when we first started kind of like our hosted product. The idea was like, oh well, we'll do self-service until we get like all of the enterprise stuff built and the self-service will like capture some of that indie SMB market. And that just really doesn't exist, right? Like people who want that type of control are willing to download it and run it locally. Most of the applications that are actually like making money at that time, which is early 2023, they're all kind of consumer-oriented. They're like a website that a consumer can come on and do something really simply and not have a lot of control, but they don't really care. It's kind of a toy and they're making a lot of money on consumers. High churn. It's really like low willingness to pay. And a lot of it is fed upon this kind of like dopamine hit of I've got this new toy, I generated a cool image. Okay, I want to generate another one. I want to generate another one.

Right.

And so people churn through credits or tokens while it's a novelty. But from like a business deployment perspective at that, at that point in time, that self-service product really isn't like super compelling to a lot of people. And it wasn't until we really expanded that feature set and made it more pro-grade and had teams, projects, and all of that kind of stuff that we started to see that's where there's really a need for what we build and what we offer at scale. Like that's where the, the actual willingness to pay is. So I would say right now the majority of our revenue is enterprise revenue. And the way that we've seen AI get deployed across enterprises is Really, I would say very cautiously. It's not like, hey, we're going to go buy this and change the entire organization over to using this new way of doing things. That effectively never happens. Change management is like one of the biggest problems in this space because all of these processes, these are not tools that slightly augment an existing process unless it's like a super simple application. It is fundamental change. It is, we are doing our business differently now. Like, this department works differently. And so for those types of deployments, what you see is an enterprise pilots or buys like a small set of seats to prove out one process, one use case, one team. We can manage that amount of change and then leverage that to show and demonstrate to others to kind of get that internal buy-in to drive change and expand. And so for us, that looks like is land and expand, right? You land, you solve the use case, you do a really, really focused amount of customer onboarding and success, really helping ensure that they get to value and can see what it can do, and then let them be champions for driving and expanding those seats. And so that has worked really well at some of the largest accounts we have. Verticals are pretty broad, like we're a horizontal tool. So we've got some of the world's largest gaming publishers, we have VFX studios, We're starting to see architects come in and use this in increasing numbers. There's, there's a lot of power and potential in this, but it's also like one of the challenges for us is we're horizontal and it's kind of complex. Getting creative control over this stuff requires some amount of like, let's expose all the knobs and dials and levers to you. And that, that's a hard balance to, to really kind of like make is accessible enough that people can get value, but also powerful enough that you get all the creative control that you want.

One of the questions I wanted to ask you as you were describing all of that is like, are you finding yourself wanting to go in and play a role in the education and in the change management, or are you finding yourself wanting to step back from that and serve the customer who's willing to take that change management really on their own, and you're there just to support the implementation of the product and the success of the product?

Yeah, I mean, I think we kind of feel it's, it's almost an existential obligation to be part of the change management and help them see what can be done, primarily because there are very few people best suited to that, like over us. We are really in a unique position because we just live it every day. We see what's happening, we see what's good and what's bad. We're able to kind of give that advice. I'll say that our role shifts depending on how deep they are. So one of our customers that comes to mind, they have like an R&D organization. The R&D organization is maybe as deep as we are in this space, like playing around with stuff. Stuff. And, you know, the buying process for them was like, we're considering building or buying. Like, we just want to figure out what, what we want to do. And I think that's another interesting element here is especially selling into enterprises with AI, build or buy is a real question, especially when you're dealing with open source models. They care about the open source model for a reason. And so you have to have a really compelling value proposition. Part of that for, for AI is like, you own all of the data and we're not trying to train on it. That's just kind of like a fundamental must must have for enterprise. But in this, in this company, the R&D organization is testing everything under the sun, trying things out. They understand the nuances of this. And for that type of customer, you're supplementing that champion. You're figuring out where can I help you succeed? What is it that you need? Maybe you already know like 90% of the use cases that you all have. But the thing that we can help you with is thinking about copyright., right? So a lot of what we have expertise on is not just the technology, but also the legal landscape and how to think about where copyright law is right now, both on the risk of using these models as well as the risk of like using these outputs in a production capacity. Because there's a whole host of copyright questions that still remain unanswered, but we are really, really close on that side of things as well. And that, that's where we kind of just provide a buffet, if you will, of advice and expertise that we can use to support a customer as needed. But I think it's our job to identify they have that gap and make sure that they get supported so that they are successful.

Yeah. This is a really interesting business challenge, right? Like you need to land the customer, you need to succeed in a technical POC. Even if you succeed in the technical POC, you also need to have provided the right type of expertise, advice to make them believe that you're a long-term partner. You're probably working with a champion. I imagine you're also working like through a champion with a financial stakeholder.

Yeah, I mean, I think the interesting thing about AI is that very often the financial stakeholder that we're working with is ultimately the champion too.

Okay, so champion and buyer are the same person often?

Very often. I mean, uh, it's not uncommon that you've got somebody who's trying to convince like upwards, but the way that AI has rolled out at most enterprises is it's top-down thou shalt generative AI, right? In, in 2023, right? We are doing generative AI. Does the C-suite know? Grammatically incorrect, but ultimately more funny as a result. Um, it is, it really is like the, the mandate is we're going to go figure out generative AI. You've got, you know, CEOs that are telling, uh, during earnings calls like AI gets thrown out. Why? Because it's what investors want to hear. They want to hear that we're doing AI, right? We are doing AI. And so I think what that resulted in is a lot of technology organizations, like tech-enabled companies that really understand technology, they're out there tinkering, playing around with, really understanding it inside of the tech or R&D organization, right? They're like, we're going to go figure this out and we're going to determine whether we want to build or buy. Less sophisticated organizations are like window shopping, trying to figure out what generative AI can do for them. And so that's where depending on the organization, you've got a completely different type of buyer and a different kind of like experience of either working with someone who has no technology context, doesn't even remotely understand why it's important to own their AI model. They just want the solution. And then you've got a technology organization that understands very deeply what they need. And that's where I think we've found A, our best buyer because they're early adopters, they understand the technical nuance and they also understand why what we build is valuable versus this other buyer who's like not sophisticated, just wants a packaged solution that does one thing. And there's a ton of vertical, like one-stop shops where it's like you don't have all of these features. You have one drop zone, put an image in, get an image out, and voilà, it's magic, right? Ours is really more of like a broader organizational solution for, for this type of problem. But to kind of go back and answer your question, the buyer in the technology organization is typically that higher up in tech or R&D who has done a project or a set of initiatives to go assess the landscape, to understand the value prop of the tool, to figure out where they can apply it for the organization at the mandate of the C-suite. Their job is to select something and start demonstrating that they've done generative AI, whatever that is, right? The C-suite's not saying this is how we want you to use it. It's more of like, We know there's tools out there, figure out what's valuable and make it work. And so that's, that's where the buyer is kind of like both a champion of, hey, this is the one because it solves these problems, and also the one who's been made responsible from a financial perspective of selecting and utilizing.

Very cool. So really would love to hear about how you initially approached pricing and how pricing may have evolved. What have you learned about the way you're going to price and package as the company has grown.

I mean, I think the, the reality is that any company who's doing this has a very complicated road when it comes to pricing and packaging. On one hand, value, right? And you hear companies try to do like value-based pricing around AI, and they're doing this, I would say, from an ivory tower, unrealistic, foolish, naive perspective of we can charge the val— like the human value that we have created equivalent of that in SaaS pricing or in consumption pricing.

So you see, like, I think what you're saying, tell me if I'm wrong, is I'm an AI, I'm a new AI startup. I think I can go charge you $60, $100, $150 because I'm replacing a bunch of human labor. Therefore, that is my value. Is that what you're saying?

Roughly. So I think more often than not, this comes down to like, what is the unit of value? So consumption-based pricing, metering based on like just general usage is starting to get, I would say, like shifted for AI companies into like value-based pricing, which is per resolution for a customer support system or something like that, right? Like if our AI resolves an incident for you or resolves like a customer request, we're going to charge you per resolution. So every customer that comes in has a problem and our AI solves it. We're going to charge you for that. And what we're going to do is we're going to take the average cost that it would take you to resolve that without AI. We're going to shave off 20% and we'll charge you that, right? That model makes sense to a buyer right now, maybe, right? If I'm doing like a one-to-one. The problem though is that from a fundamental like value creation perspective, the cost to actually deliver a solution that does that is dropping dramatically. And fundamentally, I think that's the biggest challenge right now, is the value add that a company is bringing to this process is almost nothing. It's like, it's minuscule. I think if you get like a Zendesk or one of these other like large companies that has a foothold, what they're leveraging more than anything is the fact they already have all your data and all your processes, and they're— you're kind of locked in. So yeah, you could go build your own AI solution that does this one step, but you don't have access to all of the data and the same stuff that we do, and we're thinking about it in the context of our application. So that they're leveraging that position to create that pricing differential. But underneath that long-term, I think a lot of the cost of that goes down. And so you really have to figure out what exactly is your value add. And if it's not something that's like a unique special sauce that is doing ML on the data in a new way that is never before seen, if it's just an API wrapper for OpenAI, you're not really adding value. You're just kind of throwing a prompt and data at an AI model and saying, give me, give me the answer and structuring it a little bit, right? It's like, it's, it's very, very minuscule value add. So my observation on this is pricing is just going to change dramatically. And we're, we're seeing a dynamic where the cost to deliver solutions is going down and perceived value of the company that's building these solutions, the amount of value they add is relatively low because all of it's transparent. People know that there are AI models. People know that there's Llama 3. People know what you can do with this, at least when you're talking to like large enterprises. They, they're relatively informed about what the value add is. So I just struggle to see that value-based pricing working for enterprise. I think maybe for like a consumer or a prosumer type buyer where it's like the mental math works out and it's cheaper than the other guy that does it, you're maybe selecting that product. But for our pricing journey, what we started with was really like, how do we avoid friction? How do we make it super simple to price this? We don't want to get into like, we're going to charge you credits and tokens and this and that the other because like the buying process becomes super hard and it's already a hard sell, right? Because where people are trying to figure out how to like transition their organization to a new way of doing things. So we want to make it as easy as possible to get in and figure out some of those pricing decisions later because it's all going to evolve anyways. It's all going to change. So our pricing is relatively simple. We started with seat-based pricing and a consumption component that only comes into play if you go past a certain threshold. So basically you've got a seat, it comes bundled with enough usage for the average user. You won't have to worry about consumption. And you only ever hit a consumption threshold if you're using it a lot. So like, we'll worry about that if we get there, and that's affordable as well. And that's kind of how we make that buying process super seamless. I think tying it to the number of users in the organization helps keep the cost reasonable. I've heard some of these AI model companies go into an organization— it's like one true anecdote. There was a company, big SaaS company, big brand. They have a small team of creatives that are creating work inside of that organization. I think maybe it's like 10 people or something like that. And they're just using it to create like mascots and like animation logos, stuff like that, you know, whatever. It's creative work. Because of the brand, because of the size of the company, the AI model company tried to sell them on like a $2 million license or a $3 million license. And they're like, we got 10 people using this for like a small use case. Like, what are you talking about? Right. And That's where like there's just so such a disparate way to get to value and a different perspective on what the value is for a lot of these that like maybe that makes sense if you're a movie company and you're buying a model for 10 people because you're making a movie and it's going to go make $200 million. But then you've got a different use case that's completely different and you don't really have a, uh, any way of selling a $2 million model to that company. And so that's really the hardest piece right now is value is so context dependent, but the underlying cost and the underlying like process to serve that value is pretty like static. It's like the same thing. It's the same process to create an image for the big movie company that's going to go make $500 million as making a mascot. And because of that, I struggle to get, get to a point where I believe that people will actually be able to monetize based on the value that they're creating because the cost is going to be so low unless they have some differentiated product, like differentiated model that people are willing to buy and it is special and secret sauce. There's no moat there. It's all open. It's all transparent.

It's a really fascinating thread. I'd like to have another entire conversation with you about it at some point. Yeah, but maybe the next— to me, the logical next question is how do you determine the value of your seat?

Yeah, it's incredibly hard. Right. I think I would say, if anything, I feel like we're underpriced to what we will eventually need to charge to sustain the business. Part of that though, is because we're anchored at these consumer solutions, right? The pricing that we're competing against is not just other like enterprise apps. There are a number of consumer-ish competitors that enterprise organizations have already been using. Right? And so you, you were kind of anchored at that consumer price because that's how the market emerged. And those are best, like, best-in-class models selling to consumers. And it may not have all of the tooling and all of the, like, sophistication, but if you go into that conversation saying, yeah, you know, it's, it's going to be your own model and it's in this enterprise package, it's also like 20 times more expensive, they're going to be like, well, do we really want that? Like, maybe, maybe we can give up our data. Right? It's just a harder sell. And so for us, the most important thing was making that transition super simple, making it obvious, and then helping them understand the value in the long term. Does that mean that our pricing will need to change at some point? Probably. I mean, it's going to have to evolve anyways because the compute costs are changing and it's just like constantly in a dynamic environment. But the other thing is we can build add-ons, we can build other services, we can sell into that organization. So our strategy right now, and I think this is kind of like maybe is a shift in pricing and packaging strategy perspective for me running this business is, I mean, I've historically been very like unit economics focused and very like fundamentals oriented around what are we building, how do we make revenue today, and how do we make a product that like sells and is profitable today. And we're profitable, like our product on a unit basis is profitable, but the strategy for us is more of We want to build a strong relationship with the biggest companies in the world as their partner solving this problem. And over time, the contract value will increase as we deliver that value, as we prove out that value, and as we expand into that organization. It's a different way of selling than just focusing on like pricing as an end.

Yep. And just to play it back, what I think I hear you saying is early on in this business, our pricing, like the actual price itself of the seat, will be predicated around reducing friction. It will be contextually aware of the anchors given the large consumer companies. It will be designed for adoption. We have an opportunity to lift pricing over time as we earn the right to lift pricing by providing more product value.

That is exactly right. I want to earn every dollar that we make. I want to earn that by building a product that delivers value. And I think that is what our customers start to understand as they work with us is like, Hey, you know, we, we understand it's early. We understand that you guys are probably going to have a significant amount of time where you're just figuring out how to make this work for you. We want to be the partner that makes it low friction to get in, low friction that like actually begin getting value. And as we earn the, the price increase over time, as we demonstrate the value that we're offering as a partner, that is how we increase our prices.

Pricing has a big impact on the customer experience. Pricing is dictated by a lot of systems that you use internally. Can you just talk a little bit about how you've implemented pricing into your application? Like, how does it work from a systems perspective between feature flags, entitlements, billing tools, provisioning? Can you paint a picture of how the systems work together?

Yeah. So I think we've had this conversation before too, but every SaaS company that I've ever worked at has that admin panel that you build, right? It's like one of the earliest things that you built gives you some amount of like management access to set things on the account, give them extra credits or whatever, and do all that kind of stuff. So we've built that to manage both our enterprise and our self-service accounts. Our enterprise accounts are effectively like fully manually managed. They're provisioned independently or manually when a customer signs a contract. We're invoicing directly to them and managing the configuration directly. Our self-service accounts integrated via Stripe. People go through and purchase their seats and it's all kind of automated and gets provisioned dynamically as they hit the site. So that's kind of like the, the two dimensions that we have in, in the tool that designates like different account types. Our enterprise accounts, that's where we go in and actually manage individual features and entitlements based on what the order form and the contract has, has been executed to have, right? So Do they have this many seats? Did we give them a couple of free seats? What, what are the things that we're doing to help them get this started? What features have we turned on? What have we sold as like an add-on? For example, we have a training application and that training application is the first add-on that we've sold where customers get the Studio and do all of the generation there. But if they want to like get a nice custom model generation product where you can go in and train a model. That is an add-on, right? And that's, that's how we sell that. That is managed through that admin panel. We go in and say, okay, this is how many seats to that they have. This is the features on. And that's kind of like how we're managing that right now is all manually. And it's certainly a significant amount of effort to maintain and change that.

The question that I always ask at the end of these conversations is what's been the most formative moment in your career? You've described many already. And so I want to ask a different question and that's. What has been the most important moment of the last 2 years in the Invoke journey? I'd love it if you could try to isolate on a moment or a decision.

Yeah, it is a constant stream of formative moments and decisions, but I'll try to, try to narrow it down to one. You know, I think for me, really the most, most important moment was just making the decision to do it. I thought that really is like, I would say, I remember this conversation with Nick that I was having. I mentioned Nick is the CEO I've worked with a couple times. And, you know, I was, I was really wrestling with this idea, is like, I know I want to do it, but like, I'm super risk-averse, right? And which seems like super counter to like entrepreneur— entrepreneur is supposed to be risk-on. And I've done risky things in my life. I left EY to go join a startup, right? Um, it's not that I like don't take risks, it's that I really only want to take the risks that I feel really confident I won't have downside with, right? I won't lose. I like to take safe bets, right? And I think the conversation that I had with Nick was something to the effect of just like, it's okay if you fail as long as you give it your best and you try, right? That, that's what being an entrepreneur is, is it's like taking the risk and knowing that failure is possibly going to happen. There's a lot of upside in failure too. It's just a matter of like, are you honest with yourself that you gave it your best, you gave it your all, and you did what you could to make it work?. And I think that was like a, just like an unlock for me of like, I'm one of those types that does not like to fail. I hate failing more than I like winning. And I think for me it was just, it's okay if this doesn't work out, if it fa— if like if I go out and it just doesn't work, as long as I gave it my best, I tried, I worked hard, I really did what I could to make it succeed. That, that was like a mental unlock for me. And I think pushed me into a place where like, I, I can do anything now. I have like zero, I, zero fear of failure as long as I try, you know, as long as I go out and I really, really put my effort into it. And that's kind of what I've been doing over the past 2 years. It's a constant, like, it's a constant check-in with myself of, am I, am I putting everything I can into this? Am I doing the right things? What do I need to change to really be honest about that? But it is, it is a really good place, I think, for, for an entrepreneur to accept that.

It's an incredible story. Kent, where can people learn more about Invoke? Reach out to you if they've got questions about it?

Well, you can always reach out to me on LinkedIn. Just put a message that's different than like, I'd like to add you to my professional network because I get too many of those. But that or invoke.com, it's where the product is. Links to our open source tool that you can download if you're particularly tech savvy and want to play around with these Stable Diffusion models or any of the other AI image generation models. Link to our GitHub is on the website. Like I said, open source and free so you can tinker around with it. You're one of those tech people.

Such a fun conversation. Thank you so much.

Yeah, man, thank you for having me.