CHURN FM
CHURN FM is the podcast for subscription economy pros. Every Wednesday we hear how the world’s fastest growing companies are tackling churn and using retention & engagement to fuel their growth.
CHURN FM
E279 | The Evolution of Copy.AI: From AI Copywriting to Enterprise Workflow Automation with Paul Yacoubian
Today on the show we have Paul Yacoubian, founder and CEO of Copy.AI.
In this episode, we explore how Copy.AI evolved from an AI copywriting tool into a full-scale enterprise workflow automation platform.
Paul shares how LLMs (Large Language Models) are reshaping go-to-market strategies, the challenges enterprises face in AI adoption, and why data management is the biggest bottleneck for AI-driven automation.
We also dive into Paul’s approach to angel investing, why he prioritizes founders over funding stages, and how companies should think about customer activation, churn, and retention in an AI-powered world.
Mentioned Resources
Churn FM is sponsored by Vitally, the all-in-one Customer Success Platform.
[00:00:00] Paul Yacoubian: And it comes back a lot of times to the founder. Yes, it's a founder that has proven themselves capable of doing really ambitious things. I don't really, like I'm an angel at different stages too. I don't really care what stage the company's at. When you have a founder with a very big vision and they're able to take capital and get really big returns on it, you can invest in a super late stage company and get oftentimes a much bigger return than you would have gotten in a lot of very early stage startups.
[00:00:39] Andrew Michael: This is Churn.FM, the podcast for subscription economy pros. Each week we hear how the world's fastest growing companies are tackling churn and using retention to fuel their growth.
[00:00:53] VO: How do you build a habit-forming product? We crossed over that magic threshold to negative churn. You need to invest in customer success. It always comes down to retention and engagement. Completely bootstrapped, profitable and growing.
[00:01:05] Andrew Michael: Strategies, tactics and ideas brought together to help your business thrive in the subscription economy. I'm your host, Andrew Michael, and here's today's episode.
[00:01:16] Andrew Michael: Hey, Paul, welcome to the show.
[00:01:17] Paul Yacoubian: Hey, great to be here.
[00:01:19] Andrew Michael: It's great to have you. For the listeners, Paul is the founder and CEO of Copy.AI, a generative AI platform used by 16 million users to help increase the velocity and scale of their go-to-market teams with generative AI native workflow automation.
[00:01:33] Paul Yacoubian: That's right.
[00:01:34] Andrew Michael: Paul is also an angel invest in companies like Anduril, Replit, Scale AI, and Mercury. And prior to Copy.AI, Paul was the GP at Employee Stock Option Fund. So my first question for you today is, what motivated you to make the switch from investing in finance to startup founder?
[00:01:52] Paul Yacoubian: Yeah, I'd start, that's a good question. I grew up in a family of small business owners. And so I had that on my mind from a very early age. They taught me a lot about hard work, perseverance, long hours, take care of customers. And so those things were always on my mind. And from a career standpoint, I was trying to figure out, you know, what is it that has a lot of untapped value where I could actually go build a big business. And sometimes it takes a minute for you to figure out where that is.
[00:02:21] Paul Yacoubian: And 2020, we saw this opportunity to build software to help companies adopt AI and bring applications, make it easy for them to get value out of AI. So that's what we, when it did and we built five MVPs in 90 days and the fifth one was Copy.AI and yeah, worked like use the lean startup approach to go do that.
[00:02:41] Andrew Michael: Nice, nice. Yeah. And so what made you go into investing though in the beginning if you came from a family of small business owners and was it seeing those long hours and that they drove you in a different direction or?
[00:02:53] Paul Yacoubian: Yeah, I was also, I day traded stocks with my dad when I was a kid in middle school in the late 90s. And so that caught my attention. A lot of those companies were doing the buildouts of the infrastructure that support the internet. So a lot of computing companies, networking companies. And so that was my first experience on the investing side.
[00:03:15] Paul Yacoubian: And when you make a lot of money really fast, then you lose a lot of money really fast. That's kind of addictive. And so I determined I want to figure out what's actually going on underneath the stock price, what's happening at the company itself. And so I want to kind of understand that at a much deeper level, which I was able to do.
[00:03:32] Andrew Michael: Very nice. And then you've obviously like had an incredible angel investing career up to date as well, like just looking at mentioning some of the names earlier. Like what is the secret sauce? How are you getting into these companies?
[00:03:45] Paul Yacoubian: Yeah, I'd say over just overall, you just try to be helpful and picking them, it's kind of hard for a lot of people. Like, how do you know if a company is going to be successful or not or are they taking, you know, and it comes back a lot of times the founder, yes, it's a founder that has proven themselves capable of doing really ambitious things. I don't really like, I'm an angel at different stages too. I don't really care what stage the company's at. When you have a founder with a very big vision and they're able to take capital and get really big returns on it, you can invest in a super late stage company and get oftentimes a much bigger return than you would have gotten in a lot of very early stage startups.
[00:04:28] Paul Yacoubian: So it really comes down to caliber of the founder, the ambition of their vision, and then the team that they recruit. And most of the time you fall short on the team side. Like that's usually like the hardest part to get right.
[00:04:45] Andrew Michael: Very nice. Very nice. Well, talking about the team then maybe you want to talk us through a little bit about Copy.AI, like where you're at now as a business, obviously 16 million users, but where you're at today?
[00:04:56] Paul Yacoubian: We're at 50 people on the team. I won't get into revenue numbers at the moment. But what we're looking really to try to understand is what is the core platform of primitives that are gonna go create the next-gen AI applications that are very easy for large companies and mid-sized companies to adopt and get a tremendous amount of value out of the data that they already have. So how do you take that data, pair it with other data, and then make sense of all that to drive better business outcomes for them?
[00:05:25] Paul Yacoubian: You have to have a pretty ambitious vision to go try to tackle this whole thing. But the problem is for a lot of companies, they're not taking a systems approach to how they adopt AI. They're taking little piecemeal segments and trying to find different solutions. A lot of companies are trying to build some of these things in-house. Some are buying off-the-shelf point solutions and want to get quick wins. But ultimately, the biggest returns are when you start to compound the value of all this data and compound it across the workflow.
[00:05:52] Paul Yacoubian: And the big insight is really LLMs get better the more data and the better the context you provide it. So the outputs get better, the more data you provide that creates a natural consolidating function where it's a data management problem. And we're going to have to bring all that data together in one place and then make sense of it and take a platform approach to how the use cases are executed downstream.
[00:06:14] Paul Yacoubian: So if you take a use cases approach, but do it in a disjointed way, you still end up with more data silos and you're not getting that compounding effect that gives you that holistic, organic, go-to-market motion that you want. You want the motions, all the motions to be in sync. You want the customer experiences to be in sync. And you want to make sure that all of your value props are personalized for every specific company in your target market. A lot of that data doesn't exist inside companies today.
[00:06:42] Paul Yacoubian: And companies... It's not really like an easy landing spot to put that in existing systems. So the incumbents have architected systems, largely built up structured data. And a lot of most of this data is not structured, right? It's unstructured data, but there are these gems in there and companies cannot figure out the data side on how to unlock the value of those gems that are sitting in their data sets.
[00:07:07] Andrew Michael: Yeah. And there's so much now, I think that can be done through automations with AI with agents to be able to pull that out and structure it and bring meaning to it, but as you say, like it's, it's not easy for companies to really grasp on this. And I think because two things, like one, things are moving really, really fast. Things are changing almost every day. And it's like almost hard to keep up with rate of innovation. And I think that's something similar that probably happened for you at Copy.AI.
[00:07:32] Andrew Michael: Because that was one of the things that grasped my attention was when I first reached out to you, I went to go check out the sites and my like memory of Copy.AI was basically to help you write better copy. And then I saw like a whole new total positioning and so I was really keen to dive into that maybe a little bit today and understand like, what was the thought process behind that? What drove this repositioning of a company?
[00:07:54] Andrew Michael: Because I think you already had built a significant business on the generative AI space, just purely on the copy side. And this feels like it's quite a big pivot away from that. And I can make some assumptions why, and it makes a bit of sense, but keen to hear how did that decision come to be?
[00:08:12] Paul Yacoubian: Sure. So if you look at copy itself, copywriting specifically, copywriting is used to power messaging and marketing, you know, marketing processes and sales processes. So when you're going to market every word that's coming out of your organization, that could be printed word, written word, spoken word, videos, all of that information has to be relevant to the recipient of that message, right? So when you just look at what the LLMs can do out of the box, you know, they can create messages pretty easily.
[00:08:44] Paul Yacoubian: So that copy is generic. And so when people run into that, they're like, oh, well, this sounds super generic and it's not relevant to me. So when you look at, well, what makes great copy, what makes great content that's compelling that actually does help buyers make progress through the buying journey, it's research, great research, deep research, a lot of the data about, well, who is this person? What's their company? What are their initiatives and objectives?
[00:09:11] Paul Yacoubian: You have to go bring all that to bear on the on that generation of that content and that copy. So the first generation is, well, what can the LMS do out of the box? That's a pretty simple value prop that a lot of people needed help with. And so that's super valuable for, millions of people to use LMS out of the box to go get that value.
[00:09:34] Paul Yacoubian: The next thing is that an enterprise grade that won't pass that test, it's not enterprise grade. So now you have to start to overlay guidelines, guardrails and actually configure that in a way that's super, you know, it's enterprise grade ready. It's reliable. You can depend on it. You can trust it.
[00:09:54] Paul Yacoubian: And then the second piece of that is all the data and all the research, like how are we actually physically going to go get the right data? How are we physically going to pass it through these LLMs so that the output is truly world-class and not just one time. Because one time, if you only use it one time, it's not going to be worth the effort of configuring all that, you just do it manually.
[00:10:15] Paul Yacoubian: But if I have to do this job a thousand times, 5,000 times, 20,000 times, every quarter, every year, you're going to need a pretty robust system to go build against that. And then you want that system to be highly configurable because every business, it's very different. So if you're a global fortune 500 company, you know exactly how you drive value and you just want to scale it. You want to replicate it and scale it.
[00:10:42] Paul Yacoubian: Well, workflow can be very, very different between companies, even though the overarching use case is the same. And so when you look at well, why the repositioning? Well, when you're going up market into that mid market enterprise size company, they have huge requirements, huge, huge product requirements. We've known about these requirements from the GPT-3 days, but the technology was not there from a performance standpoint to actually execute that.
[00:11:11] Paul Yacoubian: So as the LLMs get better, that's unlocking more use cases, these use cases are even more valuable than the last generation of use cases. So when we have a platform that has a tailwind because it's adding more value to the breadth of use cases that can be executed on the platform. And then for our customers, they don't have to think about, well, we have GPT-5 now. Like, does that mean I need to rethink my whole thing? Like, no. It's like, let's solve the business problem.
[00:11:41] Paul Yacoubian: All the improvements will help improve the value of existing use cases. And then you can just focus your team's time on structurally understanding, well, which of the workflows are gonna add the most value to us? How does the value equation work internally to the company? That's super important to help you prioritize the use cases that you're kind of shipping out on the AI side.
[00:12:04] Andrew Michael: Yeah. No, I think definitely, I think one of the things that I came as well to the realization a few months back, it feels like the value now when it comes to the businesses we build is really in that context layer and being able to pull that out and it feels like you've come a long way from where you just started generating copy to now really thinking about what does that end to end system need to look like and how can you help facilitate and pull things out.
[00:12:27] Andrew Michael: And I think to your point as well, I think it was Sam Altman at some point mentioned it that like you should build assuming that AGI is coming. And I think that's also like a shortfall. A lot of people say, oh, this doesn't work today. So I'm not building for it. And like it can't do these use cases, but I think like the way you're speaking now, it's like you're speaking to the future of like what's coming and making those assumptions, but they, based on like, I think good data and the trajectory that we're on as well, which is very interesting.
[00:12:53] Paul Yacoubian: Yeah, that's not exactly, that's not going to change the data management side is the bottleneck to a lot of these enterprise applications that companies, companies are imagining, you know, hey, we want to be able to do X, Y, Z. If the data is not there either, they have to spend years going and trying to collect it or they can use a platform like ours that connects to lots of different data sources and bring that in in real time to power those use cases. So again, we just want to make it reduce the friction in that process. This isn't... It's a pretty basic business model. Yeah. So I kind of look at it like it's pretty obvious. I don't know.
[00:13:30] Andrew Michael: Yeah. Can you talk through one of the specific use cases? Like what would be a typical use case your customers will be using? What sources would they be plugging in and what would be the output?
[00:13:40] Paul Yacoubian: Yeah. A good one would be like, well, there are a lot, but from a, and the sales kind of use case zone, most of the well-hanging fruit is on target, your target market. So we haven't had a sales conversation. We don't really know anything other than information we can find, how can we align what we do to what this target account is doing and who are the people that are relevant to that?
[00:14:04] Paul Yacoubian: So a lot of these use cases, you have different components of a solution. You have data providers where you have to go collect the right data information. A lot of those data providers are sketchy where they only have some of the data, not all the data you need. So you need to start to bring a lot of those together. When you start to piece through all of the data that needs to be aggregated to be, serve as the input to the use case, all of a sudden you've introduced more vendors than a company can even take through procurement.
[00:14:33] Paul Yacoubian: And so for the market that we're servicing, this is a giant problem. And so we just help, we help bundle all that, make sense of all that. And then what they can work with are these are kind of larger building blocks, more modular building blocks. So, okay, well, what do we want the output to be? It could be like a blog post where we write up very big detail like this, these are the problems that we're solving. This is why it's applicable to this industry or maybe even this company. And the data sources could come from anywhere.
[00:15:05] Paul Yacoubian: So today people are, are browsing the internet. They're finding, you know, they're finding different data sources and they're taking that and using that as context to write something. The models work the same way. So you need to connect to those data sources, figure out, okay, well, what do we want to do with that information? What kind of transformation needs to happen? That's just through prompting. And then what do we want the output to be?
[00:15:26] Paul Yacoubian: So right there between those three steps, what's the data? What do we want it to do? What do we want the output to be? That covers so many use cases across an organization and the content production processes that they have. So that is a huge use case. It could be blog content, email content. You know, companies are using us to translate their market, existing marketing content and other languages.
[00:15:50] Paul Yacoubian: In those cases, you're trying to replicate the essence of the marketing message, not just machine translate each word. And so it's very different set of use cases where you can pinpoint exactly, well, what is the message that the company wants? And these are huge companies that want to be global. Those types of workflows are so slow and manual and expensive today. It's costing millions of dollars. It keeps them from getting to new markets by months. So it delays your market entry by months.
[00:16:18] Paul Yacoubian: And then the messages you're translating are old, they're still on that update. And so if the market is happening today, like today is the market and people are looking for solutions now, I need my message to happen in real time. I don't need to wait three months to go through all the reviews and approvals. So what you end up doing is you codify all the best practices, you codify the review steps, the approval steps. And you send all that out to the language models and then you get the answers back. Now we can move really fast. We can iterate very fast. We can get to market faster.
[00:16:53] Paul Yacoubian: That's where the ROI is really being driven from is just how fast can we get to market and how fast do we feel as a company that we're moving. Large companies are super slow. They just move very, very slowly. They have all the resources at their bank, like at their fingertips. They can spend huge amounts of money to solve these problems, they have the biggest teams yet they move the slowest. So they have the most to gain by getting sped up and becoming way more competitive in the markets that they've entered into.
[00:17:26] Andrew Michael: So would you say this like shift in the business then is a pivot or more of an evolution of the original strategy?
[00:17:34] Paul Yacoubian: Yeah. The original strategy was maximize the amount of value companies get out of AI. That was kind of it. And so the first iteration, you can't execute enterprise grade workflows without GPT-4. So from 2020 to 2023, there's nothing to do. You're just gonna generate and it'd be a brainstorming tool which is super valuable for the go-to-market teams. GPT-4 comes along and unlocks these use cases. So we already had the platform ready to do that. We had a workflow automation platform pre-built for that moment.
[00:18:07] Paul Yacoubian: So now as we think about, well, what what's going to happen with agents? Well, as the agents get better, more accurate and all of these things are API centric that plugs right into our system. We're already using agents to execute different tasks inside of the workflows. In a way where you're confining them to where they do a good job. But you're not just going wild, executing random shit, because this is an enterprise. Like you cannot fuck up. So there is a maturity curve that you want to be building towards. And some of these use cases are ready for prime time. A lot of them are not ready for prime time yet. And when they are, you just bolt it in and bundle it.
[00:18:49] Andrew Michael: So it sounds like you're saying like it's obviously a natural evolution of where the company was always meant to head. But from a customer perspective, I think like the positioning of the company has changed quite a bit. And how did that sort of process workout internally, how did you end up rolling it out to customers and introducing it to those that knew Copy.AI for its original form factor and for now what it's evolved into today?
[00:19:16] Paul Yacoubian: Yeah, I mean, it was our enterprise customers and mid market customers that pulled us into this direction. They said, hey, we have, thank you for solving this problem, but we have this huge problem of how does all this work get done? It's not just one step that we're generating, but the whole process, which is composed of many different steps, we want to configure that whole thing.
[00:19:39] Paul Yacoubian: So it's, we're being pulled into that direction. This was a direction we thought we would be pulled into anyway, because nothing's so simple that you have a one step process and now you're delivering all this value. So you can look at even chat based products, they help with this one task. They reduce the time to execute that one task, which is nice, but they do not solve that entire end in chain of work that needs to get done. So you have a massive TAM opportunity to go solve more of that process and expand across that entire workflow.
[00:20:17] Paul Yacoubian: So once you once you can scale across the entire workflow and end, and this is not, it's not a trivial amount of engineering work to do, especially when you start, when you're pretty broad across the use cases on, you know, intentionally. That's a lot of engineering to make sure that you can do all the things that need to be done across a large set of use cases so that you actually do provide that platform effect and provide the platform value.
[00:20:43] Andrew Michael: Yeah, very nice.
[00:20:44] Paul Yacoubian: So it's like, you know, like you've heard, do we go vertical or we go horizontal, right? Well, if you go horizontal, you still have to make trade offs, like which use cases do we want to support and then what things need to get built into the platform to support the horizontal mess of the solution, right?
[00:20:59] Andrew Michael: And so for, from your perspective, like you've focused horizontal, like, is there something where you're going to be staying or like, are you seeing sort of any more pull in the market in specific directions as well? Like it may lead you, because obviously what you meant, you said now is that like your mid-market enterprise customers were telling you they needed these more complex workflows, they needed these automations, like this drove the repositioning then of the company now. And do you see like that eventually like evolving further into a niche or you're going to try to maintain this horizontal?
[00:21:31] Paul Yacoubian: Well, you have to, you have to maintain the platform and you have to maintain that. That's the core value. So for companies, they get more value, the more use cases they adopt and they cannot buy separate tooling for each use case. So they get more value, the more that, you know, the wider that platform gets. That's why the large companies, they end up consolidated around like ServiceNow, Workday, Salesforce, and Adobe. Those are the core ones.
[00:22:01] Paul Yacoubian: So you need to get to the level of that where you can provide the breadth of use cases and bring a lot of that data together. So what we found is we get a really good return on investment investing into the data infrastructure. We get a good return on investment thinking about like interfaces like a compound startup would, like a rippling or a ramp where they're building off one or a couple of, you know, prime sets of data. So that's the approach we're taking from a product standpoint.
[00:22:33] Paul Yacoubian: Again, it's, we're not innovating on product strategy there. That is not the place where we're innovating. We're innovating more around, how do you connect all of these things together that have historically not been connected in the go-to-market world to get go-to-market velocity, which is, can I respond in real time to new market demands poll? How fast can we move and how can we do that with the most efficient resourcing into that organization as possible.
[00:23:05] Andrew Michael: Yeah. And then, so taking the horizontal approach then, what is your approach into activation and onboarding? Like how are you thinking about, cause you mentioned a few different use cases. And I think when it comes to sort of activation onboarding being like core and vital to churn and retention, how are you thinking about activation into like onboard new customers to the platform?
[00:23:26] Paul Yacoubian: Yeah, it always comes back to the ROI. Where's the value? What's the thesis of value for that customer? And then as we build solutions on the platform and apps and products on the platform, it gets faster and faster to onboard and to the point where you just start to connect to those core existing systems, the data flows in, that sets up the use case for execution. So faster and faster.
[00:23:53] Paul Yacoubian: When we first started, we didn't even have an interface. It was just code. So as you can imagine, it's very slow to onboard. She had to code everything up. And then over time, the, you know, the, the long-term is just, is generative interfaces. So interfaces morphing depending on whatever you want the use case to be, that's almost ready for prime time, the general interfaces. And so it's going to be interesting to see how do people interact with all this information and data and at what level of fidelity people are still, we're still like slowly dragging people, you know, towards that goal, but you can't be too far out of the market.
[00:24:36] Paul Yacoubian: This is a lesson we've always felt is very important. Do not go too futuristic. And the nice thing is we know exactly which customers are the early adopters and like pushing us. And then that's where we do most of the product planning work is with them saying, okay, you guys get it. How do you anticipate this transforming your business? They walk you through it. And then now you can build that in your roadmap sequencing. Like, okay, that makes sense. So now we're taking use cases that they were custom building like 18, 24 months ago. The market is almost ready to buy those.
[00:25:15] Paul Yacoubian: And so you have to be out at least a few years ahead just product wise to get to the point where like, oh, okay, I guess we do need a solution. I guess that's now, you know, maybe a category, product category, or, hey, we're just waiting on the case studies. We want to know what our competitors are doing. We'll just buy that.
[00:25:35] Andrew Michael: Yeah. I think it's like this point of like not being too far ahead where the world is. And I think like, it's quite easy to do sometimes though, when we're stuck in our bubble in the software space and thinking that everybody's on the same level and the same mindset in terms of like early adoption and picking up new tools. It definitely feels like this is a totally different time now that we're in where like, I would say people are much more willing to try new things at this stage, so you can afford to be a little bit more risky on that end, but on the other side, like you said, like getting people to change behaviors is incredibly difficult.
[00:26:08] Andrew Michael: And if you're starting to change behavior within your product too much, that can also bite you from another perspective. So it's like, it feels like this is a tricky tightrope. You need to be walking at this stage in time, which wasn't the case before, which was like, to a point before you could just say, okay, like, we don't need to innovate too much. Like this is where the market is. Like, let's make sure,
[00:26:26] Paul Yacoubian: Where are you resourcing? You don't want to be innovating too much from a data science and ML standpoint ahead of what the models are going to be able to execute. So where we invest around applications and domains specific data science, it tends to be areas where the models are just unlikely to get a lot better with that specific problem. And so we get probably the most bang for the bucks on the prompting side. If you get really good at prompting, you can build prompts in a way that are super scalable across models over time.
[00:27:05] Paul Yacoubian: So whatever the state of the art model is that IP still valuable in the next series of models. You just don't, yeah, you don't want to be overly acute about the ML work, there's billions of dollars going into the ML side. So we just, you have to kind of stay in your land as an app platform and say, okay, whatever's coming in, bolt in. And we'll figure out how to apply that in a valuable way to customers. And that's our job.
[00:27:33] Andrew Michael: That's it. Yeah. I think like, there's a couple of things happening at the moment. Obviously there's this idea like there's huge opportunities now in vertical source when it comes to leveraging LLMs and not really innovating, but just taking advantage of what's off the shelf. It does feel though, that like, this is quite familiar to like back in the day with Facebook and Facebook apps or with like other platforms where everybody started building on things and then slowly but surely like Facebook would add feature by feature and sort of start killing off business by business and restricting them.
[00:28:08] Andrew Michael: In some ways it feels like this is happening with like the Claudes and ChatGPTs where their consumer products are getting so much better and if we think that the models are going to continuously improve at some point, do they sort of like make most of software obsolete? And this is a debate happening or discussion happening. I'm keen to hear your thoughts, like the way you see the world and from your perspective.
[00:29:29] Paul Yacoubian: Yeah, I don't know. I think they're mostly a liner on consumer use cases. Still seems like. So from what's the consumer product, who's going to pay us maybe getting into even in the enterprise is hard, right? You got co-pilots, you got Microsoft bundling this in. You have every single IT leader saying we want to be LLM agnostic. We want to be on, you know, open source models where we control the whole thing end to end. So I don't see that changing a whole lot.
[00:29:01] Paul Yacoubian: Yeah, I mean, at the end of the day, you can build, they can build whatever they want. So they think that the on the consumer side, like I saw the Open AI release of tasks, like you can assign tasks and it can set up a calendar reminder. It's nice.
[00:29:17] Andrew Michael: [inaudible] based.
[00:29:17] Paul Yacoubian: Yeah. Great. Yeah. Luckily for us, we didn't, we don't have a consumer reminder app that connects to that interface. Like, is that the right interface? I don't know. So I think if it's the end of software, you definitely don't want to be investing anywhere in the whole industry at all. The alternative is not the end of software. It's a data problem and companies realize that they have to control their data and how it gets processed. And they're not going to open up all of their data to these, like an AI company that is built off training on data that they don't have access and rights to.
[00:29:59] Andrew Michael: Yeah. It feels like probably now is a good time. And maybe this is a bit too technical, but like for federated learning to come out where enterprises can share data in a safe way without exposing it.
[00:30:09] Paul Yacoubian: They are no way near that. No, no.
[00:30:13] Andrew Michael: You think so?
[00:30:14] Paul Yacoubian: Cause that's not the bottleneck. The bottleneck is taking the LLM, getting the right data, passing it through, getting the value of the output. And companies until they solve those problems, it's not a data we have. We don't have that data yet. We don't run into use cases where the data doesn't exist either in the company or on third party data platforms. The data is there. The logic's there. It just has to get installed and configured. So that's the last mile problem.
[00:30:50] Andrew Michael: What's one thing that you know today about churn and retention that you wish you knew when you got started with your career?
[00:30:56] Paul Yacoubian: I'd say it's a little bit less about those two, but it's more about word of mouth growth, which is the better you make your product, the more people will tell everybody else about it. And it's on an exponential curve. And so I saw some stats. If your NPS score, it's like 7 out of 10, not the score, but just the number is 7 out of 10. If you move it to an 8 out of 10, it will double the amount of word mouth. If you move it from an 8 to 9, it'll double it again. If you have 9 to 10 and doubles it again.
[00:31:30] Paul Yacoubian: And so there's a method of product improvement where you just run that survey and ask people, what was the most frustrating part of your experience? And you can put it in any component, any aspect of the buying journey or the customer journey. And they'll tell you exactly what's broken and then you go fix it. And that powers everything. It powers organic growth. It powers brand affinity. It powers literally everything. So that's something, you know, a lot of times it's not on the roadmap, but like they're asking you to make the quality better. Or reliability is an issue, right? Solve reliability. It's not on the product roadmap.
[00:32:11] Paul Yacoubian: So you actually have to dedicate resources to these things that don't show up as features when you're doing product development, but they're totally core to that customer experience in their massive drivers of growth for a business. So I don't know if that makes sense, but that's what,
[00:32:31] Andrew Michael: It does. I mean, it's obviously word of mouth, but it has a huge impact on channel retention on the other end.
[00:32:35] Paul Yacoubian: That's what drives like bottlenecks to activation, right? Which is how you drive retention. And then a lot of times those same things are the leading causes of your churn. You ask people, why'd you leave? Same thing. I didn't get value out of it. Well, why? Well, you didn't have good activation. Well, what happened? Well, an onboarding flow, I still didn't know what you do. Okay. Well, fix it. You know? And so they seem very discreet, but they're completely tied into the entire journey.
[00:33:05] Paul Yacoubian: And even on the front end with marketing, like what are the promises that marketing is making? Can we deliver on that through that entire end-to-end flow. And until you get that right, you're just not gonna grow as fast as you should and your customers are not gonna love your product as much as they could.
[00:33:24] Andrew Michael: Yeah, I think there are a lot of startups that end up dying like by death of a thousand paper cuts where it's like these small little micro interactions with the product that are a little bit buggy, a little bit off or not like misunderstood, and as you say, they never get prioritized because it's always the next feature that needs to be built, but at some point, there's a frank and sound of a product and nobody knows how to use it. And if you just, like as you mentioned, start it out, it's like understanding each stage, understanding the frustrations and fixing them one by one can have a huge impact.
[00:33:51] Paul Yacoubian: Yeah. I think you mentioned cohesion, but product cohesion is really important. A lot of design leaders will come in with that as a big area of target improvement. It's like, how do we make all this experience flow together. It can be challenging for products with multiple user personas because they're literally different types of people with different types of problems. And how do you do that? And then design and then the single interface.
[00:34:19] Paul Yacoubian: These are all I think most people don't particularly care for their enterprise products. Because they're built for everybody. But the consumer products built for one use case like a ChatGPT. It's not trying to do everything. It's doing chat. And then you can, you can see them on the left. Like what else you want?
[00:34:38] Andrew Michael: But I nailed sort of the word of mouth as well from that perspective. Really, really simple. It does something very, very well and,
[00:34:45] Paul Yacoubian: They cannot, it does everything fine. It's not the best at any specific use case, but it's a go-to, right? And so how do you, why would you turn off that from what I can tell people will move to Claude for domain specific things. So like engineers might find the coding better, code quality better. Medical people might find the medical knowledge to be superior.
[00:35:08] Paul Yacoubian: So it gets into like, it's like wine tasting, right? It's like a, it's not a 90, you know, and maybe it's an 88 and smells like a 94. Well, over time, people will find it and they'll recommend it as well. And the AI space, when we launched, we were tracking all this stuff and you could build, like we had products that had really high NPS scores and then the next model would come out and state of the art moved and the NPS scores fell off a cliff.
[00:35:41] Paul Yacoubian: Because what people expected just massively changed. So when you're building consumer products, you think you have product market fit. And if you're in a highly dynamic environment, you could be way off and you wouldn't even realize it. And if you think about like, ChatGPT, how fast did it grow and how fast people find it? If there's a 10x better product that comes out, those people will fly over to it fast, like very, very fast. So even these markets that form really quick, it is not cemented in any way.
[00:36:15] Andrew Michael: In any way, yeah. That for sure, like I see it and I watch it and I think it's also like Sam again, like comments on this quite a bit is like, he says, he like gets blown away by some of the reactions from the different models as they're released and how the perceptions change of the models and it's-
[00:36:31] Paul Yacoubian: Sam's great, Sam's seen it all. He has seen it all and he has worked with hundreds of startups, you know, hundreds. And to his credit, people always are asking, you know, asking for a lot of advice. And he's like, hey, look, we actually violated almost all of the startup advice that we'd given out for the years. Right? It's true. Like if you want, if it's going to be new and unique, it's a unicorn and like by default is going to be very different. And so you would violate a ton of advice along the way.
[00:37:03] Andrew Michael: The more like founders I speak to and like success stories I hear, like nobody really follows the best practices. Like there's always some unique twist or somebody did something different. Like it's like,
[00:37:14] Paul Yacoubian: Well, I look, I look for where I can violate the advice. Cause that's alpha. I know that alpha because I know it's harder for people to mentally get there. It's easier for founders too, because you're not thinking about job security, right? You're just like handing the best thing. But for other people, people you bring on your team, if they're gonna violate some well-known piece of advice and it doesn't work, it just makes them look really bad from like an optic standpoint, right? But as it turns out, we've grown the most in areas that completely violate all the traditional advice.
[00:37:52] Andrew Michael: Traditional advice, yeah.
[00:37:53] Paul Yacoubian: Yeah. The key is understanding like, well, what's different? There's usually some different thing that creates that opportunity. Sometimes you know it going into it and sometimes you're like, what's the worst that could happen? You just go do it and see what happens. And they're like, oh, okay, this advice is in this one frame of all this context. There's a frame around it. But this is why it's different today. Right?
[00:38:22] Andrew Michael: David [inaudible], when we used to work with Hotshot, like that was one of these things, [inaudible] like basically screw best practices. Like as you said, there's no alpha in best practices. If everybody's already doing it's like, you need to be figuring out like where the puck is moving and to be able to innovate and find and cover our journey.
[00:38:38] Paul Yacoubian: I don't necessarily, I don't agree with that either. I think the best, there's like a lot of beta maybe in best practices, like go get that value. That's the floor. Just go get all of that. It's low risk. You know, our, the way we have built the team and company, we look a lot at success, like what worked for other companies, and you kind of mix and match and bring that in. And that's an easy source of like, you know, figuring out, well, what's worked and let's bring it in. That's a very natural motion.
[00:39:09] Paul Yacoubian: Then there are other things like, well, what was the lesson? What was the principle of that thing that worked crazy well that nobody ever expected? There's usually some first prints where you can go figure out and attach to what was that how do I get that? Cause that's the extra valuable stuff.
[00:39:27] Andrew Michael: I think there's also like danger though, and looking else at what's worked and then expecting that to work for you as well in some instances. And I think like, there's definitely like many stories in cases where people have left companies and join new startups and tried to apply exactly what worked at the previous company and totally failed.
[00:39:45] Paul Yacoubian: There are a million ways to fail. I try not to, I mean, there are a million ways to fail, especially new companies. A lot of people I talked to that the thing they did didn't work out. They're a little bit unclear as to what, why it didn't work, which I find interesting. So like, I can't really articulate why it didn't work. I find this also to be the case with, you know, some people I'm interviewing for roles where the company is doing a layoff, something's not working. I'll ask them like, hey, what's, why is it not working? They don't quite know.
[00:40:20] Paul Yacoubian: You know, it's a little kind of surface level, but I feel like that's kind of the same problem. Like, do you understand your business? Does everybody at the company understand the business and have that context in mind? Those are... to me, it's wild because I've always put the owner hat on. Everything I've ever done is like, I'm the owner and I'll optimize against that. And that's a great way to learn cross-functional, like gain cross-functional domain or cross-functional experience, domain experience.
[00:40:51] Paul Yacoubian: You know, put a sales hat on when no one asks you to. Nobody ever complained about me doing that, but I learned a lot about sales and the process and sold. So like those are things that I feel like there's a spirit. You've heard the like the founder mode, right? There is another mode for people that aren't the founders of companies, which is founding mode. But it is basically a mindset of I will, yeah, I want to understand everything and I want to drive the most impact wherever that is, and I'm going to take on that mindset, even though I didn't found the company.
[00:41:29] Andrew Michael: Yeah. I was speaking to Yuriy Timen recently on the show and like he had a line. It was an excellent line. He said like, I never ever let my role dictate the limit, the scope. I never let my title limit the scope of my role. And I think like, that's always been my mindset.
[00:41:47] Paul Yacoubian: Well, his title has been like growth.
[00:41:48] Andrew Michael: Yeah, it's growth.
[00:41:50] Paul Yacoubian: Growth. Okay, well, that's everything. Yeah.
[00:41:55] Andrew Michael: So, maybe we're...
[00:41:56] Paul Yacoubian: I learned so much from Yuriy that I think I've watched hours of different conversations with him on YouTube. And then I had a Zoom call with him. And I realized a lot of the questions I wanted to ask, he had already answered in all the YouTube that I've watched. And so I was like, hey, man, just thank you for sharing these insights. He's made us millions of dollars, like just that input directly into how we thought about growth.
[00:42:25] Paul Yacoubian: And the big, there's a big quote that he has, I don't know how well known it is, but when it comes to growth and you have the different channels he could go after, you know, his big quote was, did you invest enough into that channel to get it to work or not? So a lot of times people say, well, we tried ads and it didn't work and we tried to influence your marketing. It didn't work.
[00:42:47] Paul Yacoubian: His whole insight was you can get pretty much anything to work if you do it with enough focus and enough effort behind it and intensity behind it. Like how does it work? And that I use that, like I said, that's like, yeah. So now I'm like, oh, okay. If you're going to do growth and you have people focus them on one channel at a time. Give them a high level outcome to hit, give them a budget and let them go. Just do that and ignore all distractions and they will figure it out.
[00:43:25] Andrew Michael: Yeah. 100%. Nice. So last question then for today, obviously you speak to a lot of founders and I'm guessing you had a lot of questions, but what's one question that you don't get, but you wish more founders would ask you for advice.
[00:43:43] Paul Yacoubian: Oh, great question. Yeah, that's the question. Yeah, the question, the right one is it's really around value of what we're building and how that sequences into the product or not. So I will pull out a spreadsheet and I'll say, what is your product and what is the value equation, dollars and cents, that your target market is going to get out of your product? Walk me through the numbers. Okay, well, it's productivity thing is going to save people, you know, three hours a week.
[00:44:20] Paul Yacoubian: In the spreadsheet, I put three hours. Like, well, how much does that person cost the company? Let's say it's 100 bucks an hour. I'm like, great, 300 bucks a week. All right. How many people have this problem? And we'll go look at everybody that works in their target market and actually size it up and go, okay, here's the number. Now as a customer, if it saves them three hours a week, it's 300 bucks.
[00:44:44] Paul Yacoubian: I can't fire someone. So I'm not actually going to get a dollar cost savings on it. Right. Cause my P&L is not changing. So what do you recommend? At that point, it starts to shape how they think about the value of what they're building and put it into a spreadsheet because that's how customers buy a lot of times. So if it's 300 bucks an hour and it's a productivity gain, there are rules of thumb about how much they're willing to pay for that.
[00:45:12] Paul Yacoubian: So it could be like and if it's super dialed in, maybe like 10% of the value, they could charge for that. So for that one, it's 30 bucks a week. You're talking about 120 bucks a month. $99 a month sounds pretty good. Like test that and then move your way up. Maybe you can start low. Will you pay $20 a month? 50, 100, 200 and price it out that way. But what you know, once you get the first thing working, that's valuable.
[00:45:41] Paul Yacoubian: If somebody will pay you for something, this is like a big milestone check mark. Second person, big milestone check mark. Now we have something to work with. And they're going to tell you... they're going to pull you into the right direction. But you still need to frame it in that spreadsheet math because that's super important to understand your market, the TAM of that. And so what I've seen way too many times, companies will sell to like product managers at B2B SaaS companies.
[00:46:09] Paul Yacoubian: There are only like 2000 of them and they're like really easy to get a hold of. So it's a great community to sell into, but it's 2000 people. All right. Well, how big is that market ever going to be? Like, well, you have tens of millions of dollars for whatever the use case is, but they can grow really fast. So I see so many companies pivoting their when they're at like the hundreds of millions of dollars range. They're going, well, we've ran run out of budget.
[00:46:39] Paul Yacoubian: It's like, no, you didn't. You failed to create more value that's dollarizable for those people. They don't have a reason to pay you more money. They don't understand. You're not driving more value to them. So I've seen it over and over again. A lot of these companies, too, because you get these different patterns where like my investor hat interacts with my angel hat and my operator hat. And the big one is if you're selling to PMs, by default, the logos are going to look great. because only great logo companies have this role.
[00:47:14] Paul Yacoubian: So you're like, oh, we sell the Canva and Dropbox and Miro, it's like, dude, they're the only ones that have the problem, of course. Of course you're gonna be selling to them, like, all right, the question is not can you land them. The question is this a hundred million dollar market, like total, or is it 50 million? That's where it rubber hits the road, it's really hard to know that. And so, and then for a lot of younger founders, they don't actually understand the economy, like how many companies are in these different segments and roughly what is the market sizing of all that.
[00:47:48] Paul Yacoubian: And so they'll just kind of pivot and iterate until they get paying customers and then they'll follow the traction off on that, but without kind of a direction of vision. So try to avoid that if you can, where you're building without passion, without a vision. You can end up in a hole where you didn't intend on it to be. And then the other thing that happens in venture is you get these big finding rounds when you grow really fast. But when you have a very small market, that's where you can grow the fastest because there are only so many people that have the problem. You can talk to them and get them to come in.
[00:48:25] Paul Yacoubian: And then you get the big funding around, you put it to work and then you run out of market runway. That could happen at any round. It could happen at A, B, C, D, E, F. And then you go, holy shit, what I got to sell it. What's it worth? And then the answer is like two times revenue, you know, and I fundraise at 20. So you're under water. So that, that don't ever,
[00:48:46] Andrew Michael: Businesses get destroyed that way.
[00:48:49] Paul Yacoubian: Yeah. You can get destroyed that way. So don't, you know, don't depend on other people, people that invested in you or employees or even customers to tell, to question your TAM. A lot of times, none of them are incentivized to do that, or would have like thoughts about it, but it's something that you have to think about all the time.
[00:49:06] Andrew Michael: I think it's like, again, going back to a quote from David, working timely, like one of the things he said as well as like, you got to see, where are you on the budget list? And when times get tough, are you the first to go out or are you the last to go out? And you ideally want to be the last one to go out in the business. And that means obviously you're delivering the most ROI. And,
[00:49:25] Paul Yacoubian: All the time, always be delivering.
[00:49:28] Andrew Michael: Always be delivering. You need to figure that out and actually quantifying it. Cause I think that's not like an exercise [inaudible].
[00:49:36] Paul Yacoubian: That's the key. That's key. You need to have a very strong belief about that, because your product development and engineering has to tie back to the value equation. And you need to be building the inputs that drive the TAM higher over time. And you need to be building the inputs that drive the specific customer value props higher over time as well. A lot of companies just cannot figure that out. They cannot tie their engineering work to revenue generation. And so they'll build random shit like, oh, strategically, this is going to be really important. Nobody uses it. And they just fire everybody. It's incredible waste of talent, time, money.
[00:50:17] Andrew Michael: I would say that was one of the biggest shifts at Hotshot was when we really started like bringing a focus into business cases, what's going to be ROI of the specific feature, how much revenue is tied back to it. It's like, what do we expect the impact to be like? Prior to that, it was just like,
[00:50:31] Paul Yacoubian: How are we going to distribute it? So like we were the five customers that will buy this. These become like process bloat a lot of cases at really huge companies, but there's a balance like, dude, we built something that nobody cares about. Like why, what are we building? How's this going to impact the number? And like, what do we need from product this year? That's the big trade off you're making. And it's about customers. Like what, what's the customer really have a big pain point around?
[00:51:02] Andrew Michael: Yeah. Cause as you scale the mistakes become one more expensive as well. So you need to have a way,
[00:51:07] Paul Yacoubian: Dropbox is a good example. They've laid off of a lot of people that were building new products and they're building stuff that I use because I just use sync. I just need the files synced. That's what I need. And so you build an interface with all this other stuff. Like it's questionable.
[00:51:27] Andrew Michael: They need to keep growing that. They're on the- They will die.
[00:51:29] Paul Yacoubian: They need to keep growing. Exactly. As it turns out, that's like a second thing. So the core thing needs to grow, but you're not actually growing that. You're growing the consolidated top line number, which is different than growing the core thing. The core thing always has to grow. If the core thing stops growing, you're dead.
[00:51:49] Andrew Michael: Yeah. It's the irony of these businesses as well, but, Paul, it's been an absolute pleasure chatting today. I'm sure we can continue going for another hour, but in respect of your time. So thank you so much for joining. Is there any final thoughts you want to leave the listeners with before we wrap up today?
[00:52:05] Paul Yacoubian: Anybody, I always, you know, I'm an open connector. I like helping people out. So find me on LinkedIn, connect. I'll accept the connection requests and just let me know you listened to the pod and if you found anything valuable. And if not, let me know that so I can never talk about these topics ever again on another podcast.
[00:52:21] Andrew Michael: Nice. Yeah. I hope you get some good feedback, but yeah, I wish you best of luck now going forward. And thanks again for joining for the listeners. Anything we discussed today, we'll make sure to leave in the show notes so you can check that out and Copy.AI as well. But thanks for joining, Paul. Have a great day.
[00:52:36] Paul Yacoubian: Thank you.
[00:52:43] Andrew Michael: And that's a wrap for the show today with me Andrew Michael. I really hope you enjoyed it and you were able to pull out something valuable for your business. To keep up to date with Churn.fm and be notified about new episodes, blog posts and more, subscribe to our mailing list by visiting churn.fm. Also don't forget to subscribe to our show on iTunes, Google Play or wherever you listen to your podcasts. If you have any feedback, good or bad, I would love to hear from you and you can provide your blunt, direct feedback by sending it to andrew@churn.fm.
[00:53:20] Andrew Michael: Lastly, but most importantly, if you enjoyed this episode, please share it and leave a review as it really helps get the word out and grow the community. Thanks again for listening, see you again next week.
