GTM is a Data Management Problem — How AI (& Better Data) Can Fix It

0:00:03.5 David Chao: Welcome to Data Radicals. I'm your guest host, David Chao, Chief Marketing Officer at Alation. Artificial intelligence isn't just reshaping business, it's reshaping the fabric of society. Every industry, marketing included, is on the brink of transformation. At the heart of this revolution is content. But what happens when you harness AI not only to create content but to predict exactly what your customers need and deliver it when it matters most?

Today we're joined by Paul Yacoubian, the founder and CEO of Copy.ai. Paul and his team are pioneering a new wave of AI-powered tools that empower go-to-market leaders to rapidly generate highly relevant content. But as Paul says, creating content is just the beginning. The real value is being able to get closer to customers. Join us as Paul shares groundbreaking insights on how AI is transforming marketing and sales as we know it, making it easier than ever to engage. You don't want to miss this.

0:01:03.5 Producer: This podcast is brought to you by Alation, a platform that delivers trusted data. AI creators know you can't have trusted AI without trusted data. Today our customers use Alation to build game changing AI solutions that streamline productivity and improve the customer experience. Learn more about Alation at alation.com.

0:01:25.6 David Chao: Welcome everyone. Today on Data Radicals, I'm joined by Paul Yacoubian, Founder and CEO of Copy.ai, a generative AI platform that accelerates sales and marketing with data driven AI workflow automation. Before founding Copy.ai, Paul worked as a CPA that moved into hedge fund investing in SAS companies. He later helped grow and sell a tech startup before transitioning to venture capital until he left in 2020 to launch Copy.ai. Since then Paul has dedicated himself to building and scaling copy AI to become an indispensable platform for over 15 million users worldwide. Paul, welcome to the show.

0:02:02.1 Paul Yacoubian: Hey, great to be here.

0:02:04.0 David Chao: We are going to talk about all things AI, about AI agents and hopefully glean some of those experiences that you've developed over these last few years serving these 15 million users.

0:02:16.2 Paul Yacoubian: Yep, sounds good. Let's do it.

Why high-quality content needs context (and the GTM bloat problem)

0:02:18.1 David Chao: Let's dive in. Maybe first of all, for those listeners that maybe haven't heard of Copy.ai before, can you just provide a brief intro? 

0:02:27.9 Paul Yacoubian: Sure. So we started the company back in 2020 as one of the first GPT-3 applications which was really focused on sales and marketing content generation. So we started the company back in 2020. We were really excited about the possibility of applying generative AI technology (and that at that point was GPT-3 grade models) to solving a lot of these content production bottlenecks. 

And so for a lot of people, both at small companies and large companies, get a lot of writer's block, get a lot of trouble actually producing high quality content and driving sales campaigns and marketing campaigns. And so that's really the core thing that we really care about is how do we grow companies, how we help companies grow faster and streamline their operations. And today the entire product is totally different than it was then. So with the arrivals of more powerful models like GPT-4, in general, what's possible now was not possible when we started the company. And so we continue to solve these content generation problems, but going deeper into the actual workflow, in the end, workflow automation is a big driver of cost efficiency and quality improvements and standardization. So these problems now we work with the largest companies on the planet with the most stringent enterprise requirements for content quality and consistency.

0:03:50.3 Paul Yacoubian: When we started the company, we thought this would be possible in a couple years, but it was not possible then and it is now. So it's incredibly exciting to be able to apply the latest and greatest in terms of technology to solving these problems.

0:04:04.1 David Chao: That's amazing. What's an example of something that wasn't possible before that is possible now? 

0:04:10.1 Paul Yacoubian: There are a few things that I think anyone that's had experience so far with even simple like chat-interface type AI products is that the content quality isn't necessarily very good. Right. So if the model doesn't have context, it'll just spit out AI slop. Right. I think we all feel that problem. So that the way to get the models to be more valuable and deliver better outputs is providing it with relevant data and context about the thing that you're asking it to do. 

So for most companies, especially go-to-market teams, they've suffered from data bloat forever. There's data overload. There's so many different systems to pull information from these huge sales transcripts. You have all of this contact information across hundreds and hundreds of people and you have millions of websites and web pages of information. It is not easy to actually take that and make sense of it, and manage it and deliver it into the models so that they can actually do something productive with it. So we've spent a lot of time building out the data infrastructure needed to bring those processes and workflows together for companies. The output come of that use case is content that's very high quality.

0:05:33.6 Paul Yacoubian: And you know, we're doing a really large number of projects around account planning at an enterprise-grade level. So as you know, sales reps, they come in every year, here's the sales kickoff here, 200 accounts go wild and then they have nothing to work with. So they don't even have, we don't have a point of view, we have no way to help them really prioritize those accounts. And there are a ton of companies trying to solve that as a component of larger problems. The challenge is again, if you just ask an LLM to generate an account plan, it will do it. And most of the time it isn't relevant to the company, it's not relevant to the product set, it's not relevant to the contacts that someone's reaching out to. You have to get fine-tuned and very, very specific around how to execute these use cases.

0:06:24.1 Paul Yacoubian: And more problems are the end users don't know exactly how to set these systems up. And so what companies are looking for is how do we standardize the best practice and ensure that high, high-value, high-quality content is delivered to the teams right where they need it. That's a really hard problem. It's not as simple as asking ChatGPT to go research an account and give you a back an account plan. It doesn't work that way. That's not how we create really high-quality account plans. 

On the content generation side, you have the same thing you have to be very specific about who's this content going to, what are we trying to do, what area of the business are we trying to highlight? And then how do we make that very consistent from piece to piece so that you have a unified message across the entire organization? So when you look at all of these problems, it seems like they're all separate products and separate use cases. And there are companies that are trying to do that just one thing and go very vertical in that one thing. But the challenge is companies have again, they have this data overload problem. There's so much data they need to bring to bear on this. The underlying data sets are very like, they are the same data sets that are needed to power all these use cases. So being a point solution in the go-to-market space right now just does not solve the actual system-wide go-to-market problem that companies have.

0:07:51.0 Paul Yacoubian: And so they've ended up with go-to-market bloat, which is when you have, you have too many tools, they don't communicate together, it silos the data out, it breaks all the workflows. And so you don't have a seamless integration of technology to the business process that the teams need to operate. 

And so when we go really deep and we're saying hey, what is world-class level and what do we need to do? How do we use technology to solve that problem? That to us, that, that actually does work and that's what customers are really latching onto and demanding more of. And so we're seeing just incredible backlogs of use cases that customers want help with. And then the other, the other challenge I think I mentioned a little bit of this is when you're buying all these little point solutions, somebody has to try to go integrate that all together and tie all those systems together. And so you end up creating more operational complexity and not streamlining and simplifying the way that you operate. So from a platform standpoint, you have to be able to execute all of these use cases at a world-class level. That's our fundamental belief.

0:08:58.5 David Chao: Thank you for taking us through that, Paul. There was certainly a lot there that resonated with me, both as a buyer of software, but also as someone that's worked in the integration space and now in the data space as well.

0:09:11.0 Paul Yacoubian: Yeah.

Best practices for leveraging LLM tools

0:09:11.4 David Chao: You know, one thing you mentioned that I wanted to follow up on is you talked about the importance of data and training the models to get to the right output. Can you talk a little bit more about that? And in particular, what are some of the best practices that you've seen from your customers in working with their data? 

0:09:29.4 Paul Yacoubian: That's a great question. It depends on the use case as to the tactics of how do we get this to work. Generally speaking, there are a few ways you can go about doing this. 

One is we can be opinionated about a best practice and have a template for that. The second way is customers say, hey, we know exactly what we want. This is what works well for us. We want to take this and we want to replicate that with an AI workflow that allows us to execute that at scale. We can execute that very easily too. Then the third approach, it's everything else, which is, hey, we want this. We think this workflow is going to be really valuable. We have no idea what great looks like, but we have this belief because of these reasons that this looks great. A lot of times there's been an inability to execute on that particular workflow for whatever reason. And AI workflows allow them to unlock that use case and that workflow. We will invest with our customer to get those workflows up and running as well. A lot of those can be pretty custom and bespoke to their business or their industry.

0:10:39.2 Paul Yacoubian: I've seen a lot of ROI delivered through those and that gives us as a company, the ability to actually go much deeper with customers, which they really appreciate. And then it's a little bit of an iteration process. We have our own post-sales engineering team, professional services. We can help them, guide them on that process, like back and forth until they get the quality level to where they need to get it to.

Other companies where they might have more of an IT overlay into their maybe sales or marketing organization, they have internal kind of power users that can develop and do that iteration on their own. So again, it's a little bit, it depends on maturity of the business and then how deep and broad the team is.

0:11:28.2 David Chao: You talked also about this idea of data bloat and we've absolutely seen that as well in just this proliferation of data sources, proliferation of data. What are some best practices to address that data bloat? 

The path to insight: from data silos to shared context – and better content

0:11:42.0 Paul Yacoubian: Yeah, historically you've wanted to manage the data and have governance around, okay, this is the data, this is how we're going to do it. In the go-to-market world, take a sales rep, right? A sales rep has a job to do. They are their own agents, they're independent agents, right. Trying to understand the market, it's an art form, right. No two conversations are the same, no two messages are the same. And as they learn about the market and work with customers and prospects, they're training their own mental model about how the world works and where opportunities are and exist. 

They don't go query Snowflake like a data warehouse and Snowflake, they're not, there is actually no interaction between the data foundation and those teams, historically. On the marketing side we kind of just dump a lot of stuff in there. And then also historically there will be maybe like data science teams who have jobs to own specific use cases and workflows that could be like churn reduction initiatives or conversion rate optimization and like really high scale things. And so there's been a little bit of a mismatch of okay, we have data, but the data is actually coming through the go-to-market team's browser.

0:13:00.7 Paul Yacoubian: Right. Somebody opens a tab, we go research something, it shoots into the browser and then they close the tab and the data is gone. Right. And we relied on them to take action on that. And there is no back and forth on okay, well where is that data going? What did we as a company learn from that? And the answer is, historically has been we learned, we didn't learn anything. Right. 

And even worse, if that team member is no longer at the company, you've lost all the workflow around that and the data. And then I've seen this even with there are a lot of AI startups, as you know, most of them still don't even attempt to solve that data problem at all. They give you another interface for generating lists of things and that doesn't tie into any system. And so people are still operating in spreadsheets. Fundamentally, that's just not, that doesn't really help the company make progress long term. So as we build our platform out, the data foundation is really key. What are those core data silos from different systems? Once you have those systems in place, you want to be able to sync and bring that data in to actually drive the work.

0:14:13.1 Paul Yacoubian: So for far too long you've had maybe a data system and then the analytics side and maybe it's sending you a lot of insights. And now you've got alerts and notifications. It's still not, that's just not solving these problems in any sophisticated way. So what we're working with is: take those data silos, sync them up, and then make them available to the workflows. 

A lot of teams are using call recording. That's a pretty popular way to bring that information in. But the call recording software has bundled some predefined workflow and a simple use case is meeting notes. Well, if your meeting notes that you would write don't look anything like the summarized meeting notes that come out of that, you have a real mismatch. You know your business and you know what matters and that's what you're extracting out of these systems. The systems that are really point solutions are bolting on separate functionality – don't solve that business problem because they don't allow the customer to be opinionated and customize that enough. 

One of the other things is when you look at: how do we take this huge set of information and deliver what's needed to the right person, all of a sudden you realize, wait a second, the meeting notes that the sales leader might care about or the CMO might care about are totally different than what NAE might care about or the SDR might care about.

0:15:38.2 Paul Yacoubian: So how are we going to go solve that? How do we disseminate the information that matters in the right shape and context and deliver that when it's needed? These problems are, are really, really hard. 

And then the last thing is when you silo things out and you're buying point solutions and you're every team has their own go-to platform, there's a trade-off that's being made. And that trade-off is inter-departmental cohesion and efficiency. And so if you have everybody now is operating on different data sets, they don't really sync well together. Maybe you do have some information that's flowing from one to another, but what you don't get are these like multi-order insights because you cannot run queries on that joint information. You know, maybe some companies, really large ones, will try to join that in a data warehouse. 

But overall it's very hard to even run queries like what are the four most common initiatives in our customer base across this industry? By this role. Right, by this particular role. So what is CMOs in financial services that we've talked to, what are their big four initiatives? And show me the quotes from that.

0:16:53.5 Paul Yacoubian: Like what are they actually talking about? Can't run that today. And so what we do is we say, okay, well that's valuable. But we don't have capacity, we don't have enough product marketers to actually go in and there's no time. You can't even watch all these calls. 

And so what gets missed especially as you think about great content, what's the input information to great content? It's actually like an idea, it's a brief. It's like we think our prospects and our customers would find X, Y and Z valuable. And a lot of times it's how do we rethink or reposition the conversation? You could talk about, oh well, everybody's talking about what are we going to do with AI? What are we going to do with AI for this year? That's actually not the conversation at all for most companies. It's like, hey, the org just changed dramatically last month and we're having to clean it, clean up and figure out what we're going to do. That's a real conversation. Another one could be we're launching this huge bet on a new product line and we want to win this entire segment that requires us to reposition all of our marketing around that segment so we can speak to that market.

0:18:04.9 Paul Yacoubian: Those are real problems. And when you start to dive into this information, it's just sitting there and you're just thinking, man, there was no way to even extract that out of all this data we already have. 

So when we think about it as a data problem, those queries that were like wish list queries you might send a person to go do a deep dive for a couple weeks, you can get that instantly. And then once you get that instantly, now we're moving now we're taking action on it. We're creating content around that, and then we get that content out of the marketplace. Suddenly people are like, wow, okay, they understand my business. That gives us credibility, right? That helps us get demos booked. That gives our sales team credibility when they're having that conversation about their business objectives. So once you start to unlock this, all of a sudden doing business becomes a lot easier, right? You're having the same conversation with the market and instead of the market being over here and your company is kind of totally out of touch, maybe the sales rep’s a little bit in touch because they're actually talking to customers, but the rest of the org is dramatically out of sync from the real market.

0:19:11.0 Paul Yacoubian: I think it's the most important problem to solve, which is how close can you get to customers? How close can you get everyone at the company close to the market and close to customers in every interaction. And that's never been possible before. Content is one way that we take action. The other way we take action is how do we deliver the content. It's not enough to post on the blog and say, job done, we're done.

0:19:36.2 Paul Yacoubian: Distribution matters, right? So if you have a lot of SDR teams, it's like, what am I sending this person? It's not a contact, it's a human! And they have things that are going on in their lives. If we know who we were trying to reach out to now, we can predict and kind of understand what content is going to be hyper relevant to that person. Once you have that production process for content now you can go create the content and distribute it right through your SDR right to that account. So you can do these little tactical campaigns and start to parallel that together and say, we're going to ask our own prospects to write the content that we can then share right back with them and allow that conversation to happen pretty naturally.

0:20:18.9 Paul Yacoubian: And we do that. We pair that with other things that the market finds valuable. Webinars are a big one for us. Even things like just bringing people together over at dinner. It's not a sales dinner, but just bring them together, understand what is actually happening. We find a lot of value in that. We feed that back into the rest of our go-to-market motion. We try to understand where you know, what is actually top of mind for people. And it's again, it's not, most of the time it's not what's on everybody's website. That's not what's on your mind. So if I'm a large language model, I can't predict your brain by looking at your website. It's not enough information. So this is this is really setting the stage and the foundation for new best practices that solve these problems in a much deeper way.

How AI supports go-to-market velocity

0:21:09.0 David Chao: Agreed. And you mentioned there, I think the context and the collaboration around data is so important. Everyone wants access to data, but no one has full understanding and ownership of all of the data that's needed. And I think that's something that particularly Alation and our customers, that's something I hear about time and again, particularly in the context of AI.

0:21:32.7 Paul Yacoubian: Yeah, and one of the kind of early insights we had about data in general was if you just take data and you feed it into an LLM model and then you ask it for information to describe what that data is and how it might be useful, it can do a really good job of that. So what that means is that data contains this hidden metadata that we didn't even realize existed. 

But when you think about it, if I show you some data, you're like, oh, I could use it for this X, Y and Z. It has this, I don't know what to call this, but like a resonance where it drives the use case. You can just have data and say, what can I do with it? In the way that most data gets managed, it gets siloed up, locked away, lock and key. No one has access to it. The go-to-market space is the worst because the IT org and the developers that are managing that, even some of the data scientists, they are not domain experts, they have no idea what to do with it. And so you end up with a really bad mismatch of okay, we have data, we're going to store it, but every single day that it sits there, it's become the half-life is like, like a week so the next week, that day is worth half as much as what it was.

0:22:50.8 Paul Yacoubian: So the goal of pretty much all the companies that started out in warehousing was, okay, great, let's go start to move this data. How we move it through all the places where it needs to go. That's the end state. Take this data, take this information and go process it in real-time and move, move with it. Don't let it sit, don't let it stage just like piece of content about the markets. 

For example, we have a customer, large bank, they create market insights reports. It takes them six weeks to publish it. And if you're looking at six weeks ago, President Trump was not in office; has the world changed in six weeks? It's like, yeah. So all of a sudden just reducing the timeline to publish, the value of that content's going up. All of the action that needs to come downstream from the insight, that timeline, that's a workflow for us. So feed that data in, get the insight, drive the action and execute it and move on. That is working really well. That's increasing the velocity. Right. Go-to-market velocity, which is something that we talk a lot about.

0:23:55.4 David Chao: Yeah, I think velocity is key. And as you were saying, I think it's so interesting so many companies now are talking about metadata. I really think it does provide this incremental context. I think using an analogy of vision, there's what you can see, but there's also so much other data that there's infrared data, there's radioactivity readings of what you're seeing. And all of these additional data points to many companies, is the metadata they're not seeing. And there's so much more richness and so much more fine-grained decision-making they could have if they had that additional context. 

How LLMS can optimize workflows

0:24:32.1 David Chao: In the space over the last few years, I feel like there's been this progression. We've gone from workflow automation to robotic process automation to AI-driven automation and now AI agents. As you think about this continuum, are there certain aspects that you think demarcates one stage from the next, or are they all kind of variants of the same thing? They're all marketing terms.

0:24:58.7 Paul Yacoubian: They're a little bit, yeah, a little bit buzzwordy. I think the big change was you have a general-purpose LLM model that you can stick in any like data flow. Right. And so that's different. You could take RPA, it's doing things in a very structured way and you could build you could build purpose-built models to do specific tasks inside of that workflow. 

With an LLM it's way easier to accomplish more things just by sticking that into the workflow. So all of a sudden the AI workflow could accomplish what you would have needed data scientists to do and a lot of back-and-forth customization so you can get off the ground a lot faster. Which is why I think the adoption's been faster in the AI world.

0:25:46.5 David Chao: Yeah, Copy.ai, I think messages itself, I think as a data-driven AI workflow automation company. What's the data-driven aspect that you think of there? 

0:25:57.9 Paul Yacoubian: Yeah, the data-driven side is you have the automation component, which is the process that it needs to go through. And the LLM makes the process pretty flexible and adaptable to what data is coming through it right? Now, once you take the process out, what you have is input. You have an input to the process, you have the process itself and then you have an output. So to us, data-driven means I don't want necessarily someone manually clicking a button to kick this thing off. And I don't really want someone manually in the middle of this process that bottlenecks it. And I want that whole system to go end to end. And ideally over time, the components of the system are improving. 

And so what you've seen to date is when LLMs get updated, right. There's a new state of the art model. We hot swap that into the workflows, all the workflows performance gets better. This is pretty great. In the RPA world, you got to send a developer out to go update some RPA automation and go through all the testing phases manual like it. The maintenance cost is so high on that.

0:27:10.7 Paul Yacoubian: The data-driven side of it. If it's not data-driven, Copilot's usually a better type of interface for that where, okay, it's a one-off use case. It helps me reduce the amount of time it takes to execute this task. That would be the right tool for that job, for example.

0:27:28.2 David Chao: Building on that, I was reading something you recently tweeted. I'm going to mess up this quote, but you said something along the lines of if you sell software where a user can be replaced by an LLM, your software is going to get ripped out and replaced with an AI-native workflow. Hopefully, I didn't misquote you there.

0:27:46.9 Paul Yacoubian: No, I did say that I was wondering, well, which tweet is it? Because I tweet a lot of crazy stuff. But no, that's. Yeah, I did tweet that. Yes.

0:27:55.2 David Chao: Which is I think both intuitively correct, but also a very… both exciting but perhaps also a little bit of a scary statement to make. Can you say a little bit more about your thinking there? 

0:28:07.7 Paul Yacoubian: So a lot of software is just a workflow. It's a way to manage the process and kind of get insights. So a function of that process was usually taking information, doing something with it, and then hitting like send for review as a button. Those processes where you could take, okay, here's the input, it's already set up for you, and here's like the classification recommendation, the output and I'm going to click the button. That piece of the process is getting ripped out. So you don't need a person to sit there in that process and spend time doing that. If that's all the software platform is doing is queuing that interaction up for that workflow, you don't need that whole thing. So you don't need the interface. You could just run the whole thing, like stick the LLM in to replicate what that person was doing in that process and then whatever was bringing data in, if you can find an API to go get it from somewhere else, then you can execute the whole process and send it out where it needed to go. So for some products you rip the whole thing out because that's all they were doing was essentially just showing you a lot of project management tools kind of fit in this bucket.

0:29:27.7 Paul Yacoubian: They're just showing you how slow your organization is, right? Hey, I want a blog post. Okay, it's going to take three weeks and here's how we map the whole process out. And it takes three weeks, right? So they're showing you how slow you are. They're not really helping you solve the problem. They're actually adding more overhead to communicate and maintain this process flow. 

Other companies will find that AI supercharges the value proposition of their use case. So all of a sudden, if I'm the go-to place for certain workflow now, I've got AI can bolt onto it, I can execute the whole thing and move on. And as it turns out for them, the value prop was maybe somewhere else and this was a necessary part of that larger workflow that they were orchestrating and coordinating. And in that case it's easy. It's like, hey, we're just going to move on. And here's the next thing. 

I think a good example of that is probably like Intercom or support ticketing platforms where the value of the company is having great support and the best support is instant answers and satisfaction of the problem. And all of a sudden you have a big documentation base.

0:30:41.9 Paul Yacoubian: You have a very easy way to assess did this work or not? Can we close the ticket out? And then you can work to chip away at that. And Intercom's done a great job migrating from how many seats do you need to a usage-based consumption model that has some scalability for the customer.

Which roles will be replaced or augmented by AI? From SDRs to content marketers

0:31:00.9 David Chao: So maybe with that as a bridge, as we were talking about this conversation, you had said to me, I'm all about hot takes, the spicier the better. So I think if we were to bridge from the conversation we just had, there are many workflows that I think are ripe to benefit from the automation that you just discussed and If I think about technology more broadly, 50, 60 years ago, we had a specific role called a telephone operator. And someone you would dial into a switchboard and tell them who you wanted to speak to, and they would route that call for you. And as technology and switchboards and telephone exchanges matured, that role steadily went away. So I'm going to put a number of different roles in front of you, and for each, I'd love your take on whether you see them being kind of removed or retired through AI or somehow evolved through AI.

0:31:50.9 Paul Yacoubian: Sounds good.

0:31:51.9 David Chao: All right. SDR going to be replaced by AI? 

0:31:56.4 Paul Yacoubian: No, I think you need the SDR.

0:31:58.3 David Chao:  Why is that?

0:32:00.0 Paul Yacoubian: If, especially if this person is calling another person, I think there is, there's an element of trust there now for some, some companies, some use cases, once the automated version of that, like with voice, gets good enough. I think, you know, that role gets reassigned. I think the role, the highest and best use of that role is talking to as many people in the market as possible. And that's a great training ground to becoming a great sales rep and really understanding the market and really understanding the messaging. I do worry a little bit, just what is, you know, if you remove that role, what's the next generation of sellers? What's the pipeline of that talent?  

That's an issue, but you know, in general, that's the SDR role was a carve out of the sales role. So what do you look at what's on a sales reps plate? What can we carve out and shrink the per hour costs of the set of tasks, right. And hand it to someone else. So at the end of the day, it is still a sales role at its core. And, maybe, yeah, maybe that mix between SDR and rep changes a bit. I think where we've seen, we've seen probably the most consolidation around undifferentiated SDR activities. A lot of times it's like very manual research data entry, where it's almost impossible to accelerate that process. Some companies, when you accelerate the process, it makes that team more efficient, and if they have a large market opportunity, they can actually expand the team out of other companies. When you make that team more efficient, it doesn't expand the opportunities for the company. It just helps them execute on the ones that are there. And they can do that more efficiently with a smaller, more efficient team. So it depends on the company. That's a long way of saying, yeah, probably look what tends to have with roles as technology evolves is you get rid of the title and you just call it something else. There's probably, you know, new roles that get created to fit the, okay, what's the value proposition of this role to have a person performing that particular function.

0:34:09.9 David Chao: Let's try to get through a few more. Maybe closer to home. Let's talk about marketing. What about content marketers? 

0:34:17.8 Paul Yacoubian: I think the content marketing is. I always look at it what's the output of content marketing roles? Great content, right? That matters, that's relevant to customers. So if you just switch it and put it in the customer's lens, the customer probably doesn't care if that's a person or a system generating the content. Internally there's a content production process. And my thesis here is, historically, the process was so broken, bottlenecked, and bloated that if it just takes forever and you can't personalize it, it's not valuable at all. Like, you're creating an asset that most of the time is not valuable.

0:35:03.1 Paul Yacoubian: And some of the time it is because we had to have something to do. But I look at the opportunity for more content that's better. And I think that opportunity is infinite. Like, it's almost infinite. How much content could get created? So then the question is, well, what's the mix of talent and systems and people to go create that content and deliver it? I mean, I'm very, very bullish on that. On the content marketer role. I think the systems approach to that role will go up. Like, how much of a systems thinker will that role need to be? And then the most important thing is still connection with the market and the prospects and the customers and being laser focused on that because the bottleneck of the production flow was so big, I think a lot of teams had the wrong metric to shoot for, which is the output metric. Like, how many blog posts did we write this quarter? It was like five. Okay, great. We wrote five blog posts. Let's pat ourselves on the back. That's not what the customer cares about. They don't care that you wrote five blog posts.

0:36:08.8 Paul Yacoubian: They care that there was something in there is super relevant explained to them how you could add value and be helpful to them and answered questions they had and showcase that trustworthiness and just does this company have credibility or not? Or is it just slop like everybody else? So every single company we work with, we go in, we look at the content and going, what's working? Does this work or not? And it's really hard to get clear answers out of that. We know the process is broken. So I'm still coming back and saying, what does fantastic look like? It might be okay that we're producing this, but what does fantastic look like? What does that world look like? And then how do we apply technology to go get there? And these big bottom lines are slowness and speed. We don't need more people writing that first draft from 0 to 50 and it takes a week. It's just too slow, right? So if you reapply technology to that now, you can get the first drafts 75%, 80%. And they're very bespoke to that audience that we're writing to. All of a sudden now you have the ability to go put that time right back on better content.

0:37:20.0 Paul Yacoubian: And that's what we really focus on is, okay, what content's really going to move the needle? Which one is somebody going to actually sit down, read? And then how do we deliver that? Both on the production side and then most importantly is distribution. Where does this need to go? And putting it on the website is not enough. I mean, it's just really not enough. If you're not distributing that out and getting it to the right person's hands, you are not harnessing nearly enough of the value in that content.

0:37:47.3 David Chao: I think that's right. I think there's... To your point, the ability to get to a first working draft much quicker is something that we've seen enormous value from. And then in addition, the release back of that time that you would have spent getting to that first draft can then be spent on these more meaningful and ultimately, I think more business value added to tasks as well. 

How should CEOs approach AI?

Let me ask you one last question to close is given all of that opportunity from your CEOs, see as you've been speaking to other CEOs, how do they figure out where to get started? Is there a decision framework or a way of thinking where you're like, hey, this is. Think about all these opportunities, stack, rank them this way, and really prioritize in this way.

0:38:29.2 Paul Yacoubian: It is a mix of approaches. So I think from a CEO standpoint, you don't need to go take a deep dive course on how LLMs work as a CEO, you can just set the tone and the pace for the company to move faster. And the company will figure out how to move faster. As you get bigger, you will basically bloat into the size of like a stasis. Like it doesn't move, right. You continue to hire people until everything breaks. 

So with AI, you can now organize information way easier. And so what that unlocks. And this is, this was like a very, Ronald Coase, very famous economist described this, which is like, what size of an organization can technology enable? Before you had computers, it's very hard to organize information. And so even in the retailing business, the biggest store you could have is 10,000 square feet because you had to manually inventory and log everything. Once the computer became available. And at a form factor that fit into a company and into one of these stores, it was IBM. You had three companies get started in this in a 12-month period to commercialize that. One was Kohl's, one was Target, and one was Walmart.

0:39:49.8 Paul Yacoubian: Same 12-month period, all got started because IBM allowed them to electronically manage that data and information. And so did that mean that the store's gonna get smaller because they could do things more efficiently? No, it meant they could get bigger. So now imagine you're a large enterprise that's slow and stuck because information is siloed and your decision-making processes are all manual, right? They're getting blocked every time you need to review something, edit something – it's going to a different team and it's sitting. So the work time is actually a very small amount of time; in almost every business process you have communication overhead, which is about to get eliminated in large part. If the whole process can be automated in the end, why do I need to send emails back and forth trying to review and edit things? Like I don't, because I just codified the review process, I codified the edit process and I brought data in through it and pass it end to end. So what that means is especially for companies the size of JP Morgan, Microsoft as well, I never wanted to work in a big company because I didn't want to... I can't bear slowness. It's painful to me. And I want to drive impact. So with these systems, the future state is big companies are moving really fast, which is very surprising. I think we're already seeing that in the AI space. You're seeing the big winners market cap wise. Being the largest companies.

0:41:17.8 Paul Yacoubian: They already position themselves around the data silos and owning the infrastructure to support that. They're in prime position to take the next level of that, which is what are we doing with large language models? And then above that, they have the largest consumer distribution bases as well. So they're all in really good position to monetize that. 

I always believe in the culture. You can just set the culture and the team either figures it out or they don't. When you don't exactly know how to apply that tech to your business, you have to again, lean really hard on your C-suite that they understand it. And then the C-suite may say, well, I hope somebody else understands that on my team. Right. So there is upskilling that needs to happen and really understand the use cases and like the different approaches there.

0:42:05.0 Paul Yacoubian: And you know, one of the reasons we put on events and dinners and things is so that people can talk to each other. People in similar roles share best practices, what's working, what's not, what's top of mind. Yeah, you build that trust and that's, that's a lot of leverage. That's a lot of leverage that you get there. I've seen CEOs struggle with C-suite hiring to like accommodate what needs to happen from a technology standpoint. Because again, I think as the foundation of the workflow across an organization needs to get re-architected, the people at the top have to get a little bit more hands-on, to really understand, all right, what does that mean? Because if it's a restructure in certain parts of an organization between people and technology, don't expect that team to spearhead that effort. It's just not going to happen. 

The other approach was based in Elon to some extent, both at X and now the federal government. There's also the approach you just cut your headcount and that is a forcing function to figure out what's important, what's not, and then what broke and what didn't, and then rebuild from there. I've not met anyone that enjoys that at all.

0:43:25.0 Paul Yacoubian: And it's not for everybody. I think maybe investors wish that they could just invest in companies where CEOs and leaders could go do that and break that and rebuild it. The key is who's on the team and how do you hire for that. I think a lot of companies will totally break. And if you look at the federal government, there's going to be some stuff that's just. If you think it's slow now and you fire half the headcount, it'll come to a standstill. So you have to be prepared to live through that standstill, deal with the political heat and have a pretty good plan on, on how to rebuild it. If your team doesn't build things quickly, that's an issue in itself. So watch out for that. If you're getting the sense that nothing's getting done, that's probably the problem. An issue, not necessarily an AI issue. Last point here is I've seen so many companies with broken business models asking us how can AI help them, and I have to deliver the bad news. Your business doesn't work. Like I don't think, I don't think technology to accelerate what you're doing is going to help.

0:44:40.7 David Chao: It’s going to accelerate you learning that maybe your business model doesn't work.

0:44:45.4 Paul Yacoubian: No, it's not going to accelerate anything. It's going to everybody's going to get fired is what's going to happen. And so I think there's still a lot to learn around. What is AI going to help us do? What's it not going to help us do? What's the role of people in these processes? What's the role of technology in these processes? Yeah, it's a learning curve. Everybody has to get on the bus and learn.

0:45:06.6 David Chao: It's certainly an exciting time. I think there's a lot, as you said, of learning in real time, a lot of changes in approach based on those learnings and we certainly appreciate Paul, you sharing those learnings with us. Paul, thank you so much for your time today. We've had a really interesting and wide ranging discussion and for me personally took a lot away in terms of some of the insights you shared around velocity, around the need for collaboration and ultimately how data is the key input and success lever for AI. And hopefully some of these learnings will resonate with our listeners as well. Again, thank you so much Paul and look forward to our next conversation.

0:45:43.1 Paul Yacoubian: Likewise, thanks for having me.

0:45:44.4 David Chao: Bye now.

0:45:45.3 David Chao: What a great conversation about AI as a means to better understand and connect with customers. As Paul emphasized, AI isn't only about replacing human insight, it's about enhancing it. AI tools like Copy.ai show us that the future belongs to those who can blend AI's capabilities with deep human understanding, empathy and creativity. But to truly succeed, AI must be fueled by high quality contextualized data. As Paul put it it's about knowing your customers intimately and predicting the exact content they need when they need it. Leaders, it's time to embrace AI boldly setting the example for your teams and unlocking innovation across every layer of your organization. I'm David Chao, CMO of Alation, Data Radicals. Keep learning, stay curious and never stop pushing boundaries. Until next time.

0:46:39.2 Producer: This podcast is brought to you by Alation. Your boss may be AI ready, but is your data. Learn how to prepare your data for a range of AI use cases. This whitepaper will show you how to build an AI success strategy and avoid common pitfalls. Visit alation.com/ai-ready that's alation.com/ai-ready.