E67: Your Buyer's AI Is Pitching YOUR Product (Badly) | Adrian Rosenkranz, CRO @ Webflow

Adrian Rosenkranz is the CRO at Webflow. Before that, he ran global sales at Tableau, a multibillion-dollar org spanning enterprise, SMB, and self-serve. Now he's rebuilding go-to-market for a world where AI agents are evaluating your product right alongside human buyers.

What we're seeing is that the web and web traffic is fundamentally shifting to a world where you're going to have to solve for both human agents to get to your prospective buyers.

Adrian breaks down a major shift that's already upon us. Nearly 20% of website traffic is now AI crawlers. These agents are having extended conversations with your prospects and answering questions on your behalf, completely bypassing your traditional marketing collateral.

The pressure it puts on revenue leaders is that people are going to now know way more about you. The question is, are we going to step up and shape it or are we going to be shaped by it?

We also get into the efficiency trap and why giving your reps the freedom to build their own AI workflows might actually be cannibalizing your revenue.

I was talking to a sales director on our team and they were telling me about all this cool AI coaching they're building. And I said, great, are you on calls more? If you're doing stuff with AI, than my ops team can do, you're not adding value.

Adrian shares the exact internal architecture he's building to feed his own agents and reveals the absolute most important data you need to be piping into your LLMs if you want to win against your competitors.

I always say the two most useful pieces of information and data you possibly can have to give LLMs context in any go-to-market function is your shit.

Welcome to the Revenue Leadership Podcast. Momentum is a powerful tool for turning sales and customer conversations into go-to-market intelligence. Using GenAI, it extracts, analyzes, and automates customer intelligence across your GTM org. The best part, it's not a new platform for your team to adopt. It integrates seamlessly with your existing stack, or it can straight up replace your conversational intelligence tool. At Owner, we ripped out Gong and went all in on Momentum. It writes back to our Salesforce directly, captures forecast and churn risk, autofills CRM fields, shares product signals, and tracks sentiment. Companies like Cursor, Zscaler, Ramp, and 11Labs use it every day. Check it out with a free trial at momentum.io. Today's guest is Adrian Rosenkranz, who is the CRO at Webflow. And before Webflow, he ran global sales at Tableau, which was a multi-billion-dollar org spanning enterprise, SMB, self-serve. And now he is rebuilding GTM for a world where AI agents are crawling, discovering, and evaluating your products alongside human buyers. And at Webflow, they've seen a 125% jump in LLM bot crawls across their 300,000 customer base, uh, which is really interesting because it means our pipeline is changing shape in a pretty significant way, whether or not people are paying attention. And so we're going to talk about what that means for modern go-to-market and talk about what they are doing to adopt AI in their GTM org. And, uh, Adrian has some cool, interesting things to share. So looking forward to this one and, uh, appreciate you joining.

Yeah, happy to be here.

Let me dive right into that stat. So you guys are seeing this like massive increase in bot crawls across your customers. What, what, what does that mean for me as a CRO or marketers? And how does that change how we're, we're building our journey? GTM orgs?

It's amazing how fast it's changing. When I started at Webflow about 16 months ago, if you went and looked at what percent of our customer base, um, had traffic from AI, it was less than 1%. And it, that was mostly just AI doing some model training, but really different then. Since then, you know, a lot of things have changed. Not only have there been model updates, but since then models can now surf the web. They changed the way they did citations. And so a lot has changed. And so you fast forward to today and all of a sudden when we look at it, it's nearly 20% of traffic going to Webflow customer sites are actually AI bot crawlers. Now what's most interesting to me is almost half, it's like 43%, is actually AI going to a site to, to get information to then go back and answer a user the question they asked. And so that little change is pretty material. And so what we're seeing is that the web and web traffic is fundamentally shifting to a world where you're gonna have to solve for both human and agents to get to your prospective buyers. Uh, and so that shift and that change is really going to materially change how you think about marketing, how you think about acquisition, how much knowledge your customers have of you, And the thing that's pretty remarkable, unlike anything before, is the speed, uh, just because it's really changing quickly. And so I think we're all gonna have to start to really adapt very fast. I mean, if you go back 7 or 8 months, we were really looking at actually just referrals at the time, AI referrals. And then now you fast forward and now we are looking at citations, we're looking at other metrics because a lot of people are getting their answers all within the LLM.

And so can you just explain the difference between a referral and a citation and, and like what that means for us?

Yeah, sure. So when you think about being in an LLM, you ask it a question, um, you know, it could be what's the best agentic web marketing platform? And it could say Webflow. If it says Webflow, that would be a citation, right? It's referencing Webflow. If I then see a citation with a link and I click on it, it go in sources, you might click it and it takes me to webflow.com. That's referring me. What happens more and more often now is not the referral, but actually you get an answer. And what we tend to see is a lot of times users actually just go up to the URL and type in webflow.com after that experience. So they're not necessarily getting directed, but they've gone back and forth and, and got answers to the questions. And so those nuances and how that's changing and the conversion rates are all a different way of thinking. And so for example, we saw, we see typically when someone does get referred, it's about a 6x improvement in conversion. So they're much, much farther down the funnel.

Which makes sense, 'cause they've probably had an extended conversation and this is the like recommended thing for them. And so you mentioned that the way citations work has changed. So like, what, what, what did you mean by that?

A good example is if you think about 6 months ago when web started becoming part of the way that LLMs answered web search, uh, and let's take ChatGPT as an example because the majority of the bot calls we see actually are from ChatGPT right now. Google's growing a lot and Anthropic as well, but ChatGPT is the largest amount. There was a period of time where when you asked the question and it answered it and it referenced a website, it would actually have a link right there in the line. And then it shifted to actually aggregate it all and say sources. And so you'd have to click, see the aggregate sources to see where it got the information from. And so just the way that with each model change, it's choosing to, to change the way it's showing you information is changing the behavior of your prospects and your customers. It's changing so quickly that you constantly have to be on top of it. But we do see material changes. So when that change happened from like last August, I think it was when it went from like In The Lime to like Source, we basically saw a pretty significant 30, 40% drop in our own referrals, but it just showed up in a different place, which was people typing in lovelo.com as an example.

And it's just gonna keep changing like every, every couple weeks.

It's gonna keep changing. But I think the thing we can all agree is going to be the norm is that we have to build experiences in our marketing, in our sales that get agents answers to questions, cuz they're gonna be answering on our behalf. And though they'll be answering on our behalf without brand control, without our judgment, it's actually going to be its own judgment based off of the user that it has, uh, context on. We're going to have to solve for making it easy for agents to answer questions and humans, and it's very different experiences. I mean, they just don't care about beautiful animations and engagement that lets you visually understand the information. They care about markdown files. And so being able to build an experience that solves for an agent and a human is just going to have to be the norm if you want to win.

Should companies have dedicated pages on their domain that are optimized for agent crawl? And it's just like a big markdown file of all of the FAQs with answers that are, I won't say prompt injection, but it's like, like you could, you could stuff a lot of information in there that, that specifically like would, would impact the context window of that. Users search?

It's a great question. I think I get it probably every week from either a CMO or a VP of marketing. Do I need two websites is the bigger question. Like, should I have a different experience for an agent than a human? Based on the data that we see now, that would actually be a disadvantage. So what we know is with AI and AEO or GEO, we know that what you do in SEO, you get an AEO for free, but there's some other things you want to add. Uh, and so what we're seeing is that if you decide to create something separate, you have to go build your brand, you have to build your authority, all that stuff.

I know.

You then have to go and do, because those things matter in both SEO and AEO. So we actually see the biggest advantage is doing it right in one experience. It's just different things that are needed. And so I'll give you an example. You should have schema markup. And so you can go to schema.org. It's a pretty general standard. Most websites have pages where that doesn't exist. And good news is with agents, it's a lot easier to be able to keep it updated everywhere. But that gives guidance. Or another example is to your question on Markdown, most pages should have Markdown for agents where the page itself has a version available of it that's a Markdown so that an agent can go get what it needs as fast as possible.

Is that a part of the page that just has a, like, a bunch of markdown information on it, or like you like link out, it's like agents click here. What does, what does that actually mean?

So the way we solved it is we just give people a feature that they can just toggle on and it automatically creates markdown for agents of every page. So that way, like it can find it. This is where a lot of people use different techniques to make it available. But the, the whole point for those of you who are not as deep into this part of the technical setup of your website. The whole point is that there is certain types of code that the agents don't pay attention to. And so the Markdown version of the page, which is oftentimes you'll see examples, metadata is available, like you have metadata tags, et cetera. That might be something you're more familiar with. In our case, you just toggle on Markdown for agents and it makes, it makes your pages available in very clean code so that it can understand what's going on.

So it's not content on the page. It's basically having a bunch of the information that you want to communicate in the metadata as Markdown files or Markdown.

And so the way that, the way I like to think about this is actually this. When you look at, think about a beautiful website. So like I'd like to use Landon Norris's, cuz Landon Norris's website's built on Webflow. But if you go to Landon Norris, he's got, it's the agency that worked with them, Offbrand, built something amazing. It's a very beautiful website. So check it out. But if you go to Lando Norris and you look at it, all the animation and everything, the agents don't care at all. And so it's really for us, for engagement, for to, to bring us in. But then if you go look at the code, if you switch that entire page and you went to like inspect it, just like you were wanting to look at the code, that is what the agent looks like, looks at. So when you think about it, there's what we experience and then there's the actual code and the code is what the agent's consuming and I think the real opportunity right now is how do you make sure that for a long time we would optimize it for SEO. Now we need to make sure there's added AEO as well as the human. But the reason I, I'm highlighting is just because I think we didn't really have to think, it was academic, you know, 12 months ago it was an academic experience of like, what if we needed to solve for agents? But now it's, Nearly 20%. So it's a, it's material and, and growing.

And a bunch of that is, is, is sort of a live crawl from a user search, not just like the mass indexing, right?

And we're seeing it's close to 43% is a user live question.

How does that change how we need to operate like more holistically? So. Okay, make information available on your page. That's like a one piece of it. What else should revenue leaders be thinking about with rank, like AEO and sort of LLM rank alongside their normal like SEO and efforts?

You have to think about the fact that with SEO, you're competing for a click. And you control the content that they see, cuz you're eventually clicking onto content that you're likely controlling. Now there's content that maybe third parties create, they maybe it's PR, maybe et cetera. But with LLMs, there's an added judgment, which is what the user's context has given the LLM. And so it's constantly changing. And I think what it's going to demand is it's going to demand us to be able to answer more and more questions that people have. The thing that's unique about an LLM is that it's gonna give you an answer to pretty much any question you ask. So the long tail is just so much longer. And so things you never even thought of asking or solving to make sure people get an answer to, they will ask an LLM and will get an answer. And so the question becomes, how do you become that knowledge base that an LLM will use to answer questions about your industry, about your product? The pressure it puts on revenue leaders is that people are gonna now know way more about you. I mean, I was in France with a customer of ours and I was on the, we were in a room together and I knew that they were actually just asking questions of an LLM based off what I was saying in real time. It was pretty funny. And that's the type of experience. So the knowledge and information they have access to is completely different. And so the way that I think revenue leaders need to think is, are you going to help shape the information that goes to LLMs or are you going to be shaped by it? Because at the end of the day, your customers, your prospects are going to be asking you questions like, is this pricing fair? What are our other alternatives? Like, how are you, how should I, what questions should I be asking? And I think that's going to be the question is, are we going to step up and shape it or are we going to be shaped by it?

What I'm experiencing right now is like our product team has increased their velocity of shipping so much that it is almost to the point where like you can't enable people to know all of the things that are coming out as fast as it changes, which is like the best problem to have, 'cause it means you're getting a whole bunch of new product and, and it's all very exciting. But at the same time, it puts real pressure on Like, what is the capacity of a human rep that is ripping like multiple demos and follow-ups every single day to engage with and learn and remember all of these new things and features and, and, uh, data points. And I would imagine your world is changing just as fast because you are, you are so, uh, deeply affected by the changes on the model side. Have, have you guys thought at all about like How do you enable a rep when so much is changing so fast? Like, is it about having Cluely for sales and like a little copilot there giving them answers? Or we're talking to, uh, OneMind is a sort of avatar product we use on the website. Like, can I use that in a call so that it's like an on-call AI solutions engineer that you just invoke when you like get stumped or you're like, oh yeah, like we have this new product and it, I can't like totally remember what the feature set is and you just have that thing on, on call. Have you guys thought about how to prepare reps in this like rapidly advancing world where more is happening faster than ever?

The first part is how do you create the illusion that not that much is changing while pumping in a lot of change? And so I'll give you an example. If we go to anyone in sales and you tell 'em, everyone drop what you're doing, we're doing a call blitz. Everyone will roll their eyes and be like, man, the leadership team, they're just like reactive. Like they're pulling us off what we know we need to do. If everyone knows that I'm gonna do a call blitz every Monday and we are like, oh, it's Monday, we're doing the call blitz. Same action, same activity. No one's gonna like think it's out of the ordinary. And they're like, no, this is part of the run of the business. And you might be able to slot in different messages, et cetera, within that. Right? So my analogy of the call blitz is actually of a narrative. If you have a really strong narrative and positioning, then you should be able to slot in different features, functionality into the narrative. So the thing that the rep actually knows really well is what's the narrative, what's happening in the market, how is it changing, what's our point of view on it, why does it matter to the customer? And then when new features come out that support that narrative, if we've done our job, then it becomes more about, hey, it's the narrative, you know, now connect the dots. And I'll give you a very specific example with our narrative. I think one of the things that's— and we just talked a lot about it— one of the things that is changing rapidly on the web is the way you get traffic, um, and what kind of traffic you get. And so when you think about it, most of our customers are trying to drive traffic and convert it, uh, the majority. We do have some customers that have slightly different use cases, but most of the time You're trying to drive traffic to your site, you're trying to convert it. And so if you are trying to figure out how to help them, the reps sell that and understand that it's understanding how, you know, we've gone from less than 1% to 18% of traffic is LLM crawlers. It's understanding how half of our customers, more than half our customers are trying to figure out how do I improve conversion with AI? How do I create dynamic experiences that are changing in real time? So they're trying to figure these things out. And so as you're trying to think about how to enable all that, if they know that narrative and then all of a sudden agent Markdown for Agents comes out, since we just use that as an example, I can say, hey, great news everyone. Like it's not, it's a one button turn on, it's toggle on and off. Good news everyone. This whole story about how do you make sure you show up for agents, you got really easy because we just enabled, um, this new feature that helps it be even easier because you just turn it on and it'll create markdown files for your pages. Now it becomes less about the feature and more about the story. And then the augmentation there is, to your point, like, how do we make sure that when they get us further details of it, they can get answers quickly, either via an LLM answering a knowledge base in Slack, or whether it's a solution engineer. But I do believe the first part of the problem is make sure you have a really good narrative that can that can funnel everything through. The second piece, I think, from an AI perspective is if you have really good docs, then you should be able to provide people the answers in real time. And I think it puts a new pressure on making sure external documentation— for us, we work with agencies really deeply. Agency documentation, people need to be able to access it in real time. And I think that's a fundamental shift. And they, we need to be able to, uh, access it from where we work. So for, in our case, a lot of the reps are working out of Slack and they use Slackbot all the time as an example. And so you need to be able to just ask the question and it gives you the answer. So these are just like two of the things that like come to mind. I'd say the, the third thing that is probably most interesting is product should just release when it's ready. And it's our job as go-to-market folks to figure out how to not only enable our reps, our customers, and our partners, but it's our job to figure out how to make it as effective as possible for people to learn and use that. And I think it means more pressure on us to know with certain features and functionality, like which part and segmentation of our ICP does it fit.

And do you think it means that you need to niche down your team more potentially? Like if there's, um, if there's certain features or products that fit a particular market, then, you know, certain people can focus on learning like fewer things. Or is there anything structurally that maybe you would, you would evaluate changing if there is just so much more getting shipped and changed?

I'd probably get to that second. I actually a week ago was rattling around with this problem, and so I built an agent that basically goes and looks at all the Slack channels. So we have Slack channels that basically we call like Okay to go to market, and which basically is like the sign-off between the PMs and the PMMs, for example, that we're releasing the feature. And there's different tiering, of course, but basically I realized because the volume's so high, I actually built an agent that said, okay, based off on our ICP, based off our messaging and positioning, I want you to look at these channels daily and I want you to flag any features that really squarely fit into this narrative. And then I want you to go look at all our open deals and match which of these features and functionality based on the customer transcripts. Might be worth mentioning. Because I, the way that I break it down in my head, and maybe this is overly simplistic, is I'm like, is this features and functionality that gives us a reason to call about a problem that they have and it's a new thing? Or is this features and functionality that actually are gonna just like strengthen our position in deals? And it's not necessarily like compelling enough on its own to call someone, but it's a good add. And the question is like, how do you insert it? And so, Right now my approach is how do I proactively scan deals or like prospects and figure out how to use the, uh, feature. But I find that because of the speed, it's actually a really good use case for the agent to constantly be scanning those channels and saying, oh, this feature, this is the feature, here's the PRD based off that, who should we be focused on?

I, I actually really like that as a mental model. Like, is it a feature that drives a reason to talk or drives the deal forward? Or is this a feature that you need? Ambient knowledge of, and maybe that's where you take a more just-in-time approach of just like giving people, we use Guru, so giving people access to like, you know, at their fingertips, asking a question, whether it's in Slack or in the app experience. Okay. I want to talk about what you're, how you're using AI internally. So that, that's a really cool use case. It's before we started recording, it sounded like you're building and, and doing stuff of your own, which is great. Clearly you, you get it, but you've said that there's this efficiency trap. You've warned about teams following into this efficiency trap where you're optimizing certain inputs with AI without, with, while also losing sight of the outcomes. And so like, just explain what, what you mean by this and, and then we can talk about how to avoid it.

The efficiency trap, I think the easy mental model to use is anyone who's really good at responding to emails just ends up getting more emails. And that doesn't necessarily mean it's good or bad. It just means they get more emails 'cause they send out more responses, which get more responses. And so the efficiency trap can— is when you're being more productive at the thing. So in this case, emails, but it may not be driving the outcomes. And so as I think about AI, especially in go-to-market context, It's something that I really think about a lot because you feel this magical experience when the, when an LLM does its thing. It's quite remarkable. And does it get us what we need to get out of, uh, the goal of like the job that we've hired the person to do or to make our customer successful? Like, is it do, making it better or is it just doing something? And like, I'll give you a good example and I don't know, I'm sure a lot of people deal with this, is that Every single person that faces customers wants to use their call data for some kind of use case. They want to integrate it with other tools. And the reality is I don't really need, you know, over 100 people doing their own unique version of it. Where it's most useful is when I have my RevOps team, which I've actually collapsed everyone from marketing ops and RevOps, sales ops, uh, post-sale ops to one team with go-to-market engineers. And they think about like, what's the connective tissue? And instead of being focused on like, oh, I, I'm connecting the dots between, you know, a Gong call, um, and being able to like automatically log it, it's actually the people who face customers using that to get pushed information more effectively and, or being able to coach their reps as an example. And so what I'm really always thinking about with, um, my RevOps team and specifically my RevOps leader, Tara, is how do we create not an efficiency trap, but everyone just, they can use AI, so they start like spending 20 minutes on sending an email instead of just spending 5 minutes on writing it. You know? And so like, how do we make sure that the systems are connected and we're actually empowering outcomes? And how do we always focus on the outcome? And that's really what I mean when I say the efficiency trap of we need to make sure we're getting some outcome, not just that we can do more of it.

It can feel exciting and it can feel like you're really moving the ball forward to play with an LLM. But there's this Stuart Butterfield quote where he talks about hyper-realistic work-like activities, which is basically like, you know, what are the, what are the actions in work that you can take that don't actually move the ball forward, but feel like you're doing something. And reps are the most susceptible to that. Because it's like, you know, what do you, what's the work you want to avoid? Well, what's the hardest thing? Picking up the phone, making a cold call, reaching out and chasing, chasing your pipeline. And you can get stuck in these loops of, oh, I'm going to, you know, build my own little call transcript follow-up automator thing. And that's one of the reasons why we have much the same as you, a very centralized approach to how we deploy AI. And it is our applied AI team and RevOps and BizOps and data. That are doing this stuff and just like delivering outcomes so it's trackable.

And I couldn't agree more. I mean, I can tell you that I was talking to a sales director, uh, on our team and they were telling me about all this cool AI coaching they're building. And I said, great, are you on calls more? He answered, uh, what? And I was like, yeah, yeah. Like, this is nice, but like, are you on more customer calls? Because for me, if you're a customer-facing role, what I need you to do more of is be with customers. And so if you're doing stuff with AI that my ops team can do, who understand systems better, um, and specifically the, the folks who are the applied AI folks, they understand the systems, they are working with prompts more, their, their job is this, then like you're not adding value. Or we're, or we feel like we're doing something in forward momentum, but really I could have someone else do it where I can't have someone else talk to customers like you can.

Very much agree. I, this, I wrote a Substack post, which was basically arguing for a centralized approach to AI transformation in GTM. And it was like by far my most popular Substack by like 3x. And I think it struck a, struck a chord with people because they're feeling this, they're feeling like, oh, your sales manager spent like 4 hours building a thing that like works, but still sort of sucks because. You know, it's just the little thing that you built on the side versus, you know, the, the, what YG, who's our applied AI guy for sales builds is like, not just like, you know, 50% better, it's like 10x better and, and scalable and can be widely deployed.

So, and, and to that point, the hardest part is the management of it. Like I get more fearful, not of the build project that someone did, I get more worried about how. They're gonna maintain it, how they're gonna make sure it works. Like all those hours spent where if your role is to be with customers, I want you to be with customers.

Tell me more about collapsing the RevOps and go-to-market, go-to-market teams like GTME. What caused you to do that and how have you, what have you done?

Well, when you think about, we created all this specialization, I think in our industry around certain roles, like, you know, marketing automation. Or Salesforce automation, right? Like I came from Salesforce, so Salesforce admins, but we created all these specializations kind of because we had to, we were forced to by the technology to create specialization. Now the good news was it's still better than anything we had before, but we were forced into that world. But when you think about the complicated part is that those worlds and those teams always had to create handoffs with each other. Whoever was in charge of marketing ops eventually had to get the lead to sales, and then once it was in sales, you had to make sure you eventually got this to post-sales or customer support. And so there was just a reality of we had to specialize, but the workflows were never isolated within a department. The workflows were always passing from one to the other. And so as we started to get quickly into this world where Part of what we needed to do is create an infrastructure where our systems talk to each other with AI so we can do whatever workflows we wanted. It became very clear that the workflows needed to live across these teams, and I didn't want there to be an unnecessary barrier, which was the fact that they lived in different orgs. Because of that, we collapse it all into one place. And if you think about what they're trying to do, event, like there's a couple pieces. They want to create the system so they can all talk to each other. Then you need to have the context files or any kind of context, which the systems provide. But there's also sometimes augmentation there where whenever you have an agent running, you need to understand all the context of the workflow. And then you need to be able to evaluate whether it's working or not and enhancing it. And those 3 steps didn't just live in one place. You know, marketing is judged by pipeline, which lives in sales. That's not. Salespeople create the pipeline, at least for now. That's why it was very clear to me that the only way we're gonna make it successful is, is break down an artificial construct, which is an org chart, and basically say, hey, go run with this. I mean, I always say like the most, two most useful pieces of information and data you possibly can have to give LLMs context in any go-to-market function is your usage data. If you're in software, right? How they use the product in any way, shape, or form. And then the second thing is your customer calls. Those are the two most powerful things cuz they have the most information. If you can't use those effectively, but everyone wants to use 'em. Marketing wants to use them, sales wants to use them, post-sales support, everyone. That's how and why we decided to bring it all together.

So was marketing ops living under marketing before and now it lives under go-to-market ops or, or like what is the shape of the team?

So if you rewind before, when I got here, I guess, you would have sales ops living in sales. You'd have marketing ops living in marketing, um, support ops in support. So you'd basically have the ops teams in each function. In most companies that I've talked to, once you get to a certain size, they start to specialize it just because having to be the Marketo expert is very different than having to be the Salesforce expert. Or HubSpot or whatever else you're using. And so, and now with AI, the, the, the stacks are just massive. And I think that's one of the challenges that we're all dealing with is that, you know, at least in our space with Webflow, a lot of companies and marketers have created these Frankenstacks that they're trying to like stitch together to create these digital experiments, experiences, and experiments to get people to convert. And it's really hard. And so having one view of everything. And one team that's job is to solve the workflow with AI, not solve a department's problem.

It's interesting 'cause almost everything is becoming a data problem now in our experience. Like what you used to call something that might live in marketing ops, it really is more like a traditional data science problem or data engineering problem. Like we're, we're Webflow customers, for example. And so like, how do you get all of that information into the right places so that you could have this full funnel understanding of your customer? And those problems do cut across a lot of different places. We still, so our RevOps org has always been centralized, but we have specialists for that are embedded with each team, but there we can certainly see the, the times when a problem needs that cross-cutting, like one, one view of the customer journey. And so trying to think through how to solve that is, is certainly top of mind.

I think what's gonna be interesting is like the, the specialized skill sets I think are just gonna change. I think where we were was you needed to know Marketo better than anyone. And now I think it's gonna be something like who are gonna be the people on that team that's job is to create eval frameworks for when agents are doing a good job versus a bad job. Who are the people on the team that are going to be managing the knowledge bases that are constantly evolving and learning that are used in a bunch of different other workflows? And then there's gonna be, who are the people on the team that are gonna be the, the builders who, who talk to the, the business owner, the marketer, the salesperson, et cetera, and translate that into agents to help them do their jobs. Those are different skill sets. Now they might be all the same person, they could be different people, but I think That's gonna be something that evolves pretty quickly. I mean, the conversation I'm having with my analyst right now is I want that— I actually tell 'em to do analysis in the old school way and also ask an LM to do it too. And then have a grade, have the LLM grade itself against what they did because I basically want them to like, for now, keep learning how we do analysis on the business.

Yeah, it's, it's a lot more like a product org. Where you have sort of PMs embedded with their users, the, the business unit. And then you've got the builders, as you said, well, just those are engineers. And then maybe there's specialists, there's a backend team, you know, backend engineering, which is the people building evals and maybe the database structure. It, it really feels like the, the roadmap for this is a product org. Getting from what we do now to that vision, it's gonna be messy, I think. But it feels like the right mental model.

It does. And the users are just people who happen to work at your company instead of not.

Where else have you seen the most transformation within Webflow? Like, where are you guys getting the most value with AI? And, uh, I'd love to just like hear about some specific use cases.

I'll share two simple ones and something more complex. I'll show you three. The first, which is a very simple one. We typically have the need where, whether it's a salesperson, a post-salesperson, they want me to send a note out to a customer or a prospect. The old way of doing it was we'd have a workflow in Slack, that Slack workflow, they would then put in a draft email. I would then go look at it, rewrite it, put it into Gmail, send it. That was how we used to do it. I'd since created a, an agent that is so much better because it takes whatever the rep or customer success manager wants me to send. It then goes, looks at our Slack channel for context. It go looks at our call conversations with the customer for the last 60 days. It then checks my calendar, checks my Gmail and says, does Adrian know this person already? Have they met with them? What did they talk about? And then using all that context, it rewrites the email against my, uh, my voice, which is a different skill I have in Claude. And then once that's all done, it actually puts it as a draft in my Gmail and sends me a Slack. It's automating a bunch of steps with using skills that are designed to make it easier. And then I have it actually grade itself. So every, every time it runs, it goes and looks at the ones that it's actually drafted for me and grades itself with 100% being, I didn't touch, didn't touch the email at all. And I got a response. If I don't get a response. Then it grades it less. If I have edited the email, it grades it less. And right now it's actually gotten close to 90% of the time I don't have to edit it.

And what's the stack you've built it, uh, with?

Yeah, so I built it in, um, Cloud Code. So Cowork is probably the easiest way to do it. I MCP to a combination of Slack. I MCP to our call conversations, to my Gmail, to my Google Calendar, and then I just gave it instructions. So with us, we use Slack very, very, uh, well in with all our customer conversations. And so I explain like in the Slack channel, you'll get the internal context of what's happening in the account. When you go look at the calls, like understand if the person I'm emailing is some, is in the calls or if it's an executive who hasn't been involved. And then I made my own skill, which I highly recommend everyone to do. Just go and build a skill that's your own voice. The best thing I did was I just had to train off emails I already sent. So I was just like, go look at all the emails I already sent. And that's my email voice.

And I'll, I have a prompt for this that people can use. It's like a mimic my tone of voice prompt and you can just dump anything into it. My, so after this, after we record, I'm going to put the transcript into my, into my podcast post-production Flow, and it's gonna send you an email that has pre-written LinkedIn posts in your voice based on your tone of voice that it's built from this recording and scraped whatever I can find on you across the internet. Because I, like, I don't know why people don't have this already. Like when I explain to people I've got like a writing skill that imitates me with like freakish accuracy, like, oh, how do I do that? So this is like my, one of my post recording, uh, surprise and delight things where I'm gonna get, I want to give everybody, every guest their writing skill or their writing prompt snippet, however they want to use it.

That's great. Say you have that, then the only part you're really missing in this workflow, if you want to build it for yourself, is you just have to give it instructions on how it builds context. So I, when I say go, you get the account, now go look at the Slack. Here's how you find the Slack channel. I give it instructions on how to do that. Here's how you build that context of what's been talked about internally in terms of our, our, our plan with the account. Go look at the customer conversations, build pattern match that. So that's how I like took it step by step. And then eventually the last step is now run it through my voice and put it into the email.

Yeah, I was, I was just trying to think of like, how can I make it so dead simple for people to share the podcast? With their networks and like, because I'm, I do a bunch of other people's podcasts and I'm like, and so I am much more slow to like write a LinkedIn post to, to do it. If I have to like put the transcript in my own, I just dump it in my same workflow. I'm like, hey, I was a guest on this thing. Like write me a newsletter and some LinkedIn posts and it knows how to like rip through it. And I was like, oh, I should just do that for every guest. And, and we'll see, you're, you'll, you'll be, the second guest to get that output. So you'll have to tell me.

I'll use it. I will use it for sure.

Tell me how good it is in terms of actual, like writing quality. 'Cause I want to, I want to iterate it because I love, you were saying like every time it runs, it grades itself. I was like, ooh, I need to do that.

Yeah. So basically what you do is at the end, you basically say, uh, oh, actually I put it at the beginning of the instructions. It's not the end. 'Cause once you've done it once, I go, before you run, the first step you're gonna do is go look at previous and basically grade yourself. Here's how you should grade yourself and then make enhancements to the prompt before you run it again with your goal being 100%.

Okay. Well, I'm just gonna drop this part of the transcript in my Open Claw Telegram and be like, do this for me.

Let me know how it goes.

And so this is Claude Code. So, so it's, you're invoking that when when you want it to run, or you have a cron job through Cloud?

Cron job.

That's okay.

Yeah, I have it scheduled.

So that's with Cloud Co— that's with Cloud Routines then?

Basic, yes. And the thing though is that I will say that some days people really urgently want me to send something, and so I'll get a Slack message and I'll just go click run and I'll just run it. But it's still better, like clicking one button versus having to do everything is still way faster. If someone pings me and says, hey, I need you to do this, then it ends up being, you know, less than 5 minutes from start to finish. And I'm doing other things at the same time. The other thing I'll say is that part of this was I actually, we have a workflow in Slack that goes to a channel. So I, I should have mentioned that the way that it starts is the first thing it does, it looks at that channel and identifies when people want me to send something.

Yeah. The, the thing, like I built so much stuff in Open Claw. Because you, because it could be virtually hosted and I could have it on cron jobs. And now Claude has so much of that stuff. I was like, oh, it'd probably just be easier if I had it all in Claude, but I'm like pot committed to Open Claude because I've got all these things built already. And so I'm, I'm, uh, but then now Nate Fullen, who runs RevOps at Perplexity, was telling me about the stuff that they have. And I was like, oh, maybe I should look at Perplexity. I have the worst AI FOMO. I like want to use them all and, and be able to have it all, but.

Well, and then every week something's new come and you want to try that.

Yeah. It's like cloud managed agents. I'm like, oh, that, that's like one of the last missing pieces I have with my OpenClaw instance, but I'm so attached to it. I'm so attached to all the skills and the like.

I know my, my biggest thing that I've kind of gone on one in the last week and I was actually working with our RevOps team because I think. They're uniquely, uh, how I wanna build it is, um, basically there's a whole set of like agents I wanna build, but I was realizing that I was missing something so obvious that we would do as humans, but I didn't do it into my agents, is explaining to the agents who to delegate to and who they should receive information from. For some reason, I never had thought of that, but being very like very specific of, hey, you will delegate to this skill. Cause so that, that way you can actually control different parts of the workflows. Cause before, when you don't do that, you, you, it's dealer's choice what it decides to do. That's been the latest thing that I've been really pushing the team on is making sure we're very clear on architecting and thinking through when we delegate, when an agent delegates into which one and to what type of skill versus, and where should it expect to receive information from?

Yeah. This is the thing that I think OpenClaw figured out early is this idea of that tools file. And so you can have this like a good understanding of, of, of all these things, but it is still like, I'll sometimes I'll get an output and I'm like, this doesn't sound like me. And it's like, oh, I forgot to use the writing skill. And I was like, okay, well now like, okay, then I'll like use WhisperFlow. I'm like, all right, make sure that you update this skill to always use the writing skill, cuz I don't wanna put the writing skill in every skill that it uses.

So, and I think what you're getting at is exactly what we were talking about earlier about the importance of management, like managing the agents. It is not, uh, set it and forget it. You know, I think one of the things that we're seeing pretty clearly with Webflow, but also just in terms of how we're using AI is the day one is never the problem. Making sure things are like working on day 20, 50, 100, that's the hard part.

And I find it's a new way of working like that. That's the thing that took me the most amount of time to get used to is like always being in the loop of if I see an output that I'm like, oh, like that's not a great email or like, oh, that's not good. Instead of just like not using it or, or telling it to do it again, like always going back to the skill and having it update the skill or update. It's like, okay, like, why was that not good? Oh, well, it doesn't really have the right context. So I need to make sure that it's accessing like this MCP or it goes back to these context files. And so the, the habit of every single time you do something, you like do the action. And maybe use that as an opportunity to give feedback to that skill, to, to like update that skill.

My favorite way to do that is I, I, it's when you feel that moment where you're like frustrated, you're like, this was harder than it should have been. You know, that moment that you get, okay, this is way harder than it needs to be. Time for a mini retrospective. Your job is to look at this entire conversation and figure out why it was so hard. You, you can give me feedback, but then you also need to take feedback that you're gonna implement. And then I have it, and then I do a mini retro and it has been a lifesaver because it basically does one, it tells me when I'm taking it the wrong path and what I could do better, but it actually just rewrites how it's approaching it going forward. So I highly recommend that to anyone who will, who like, when you get to that point of frustration, just tell it this is harder than it should be. Run a mini retro, give me feedback, but you, I need you to take feedback that's actionable and it's gonna make this, I think I like to use 10x better or 100x better for the next time.

All right. I cut you off and we went, uh, down a rabbit hole. You said you have 3 interesting use cases you want to talk to, and we've got to use case 1, which is—

Yes.

People getting you to something else.

Number 2, and you and I were talking about this earlier, I think one of the things that we're all gonna be searching for is how do you create knowledge bases that are constantly evolving and learning 'cause the business is changing. And so this whole idea where, you know, at least where I started was, oh, you get a file and you're like, yes, I got the file that explains my business and I'm gonna put it in there. And then a week goes by and you're like, that is old. One of the things we're all gonna be doing is maintaining knowledge bases that are like very key, like canonical pieces of information that are gonna be used. So sometimes it'll be like a glossary of like terms to understand. Metrics definitions of a business. The one that I was really obsessed with was our ICP and specifically how do I use data from our Close One, from our customer calls, from what we see in our Slack channels happening. How do I use this data to basically create a ICP and every week have a drift where I'm like, okay, how much do we drift from what our core ICP was last week to this week based on what's closing and, and, and then be able to look back a month and say like, how has that changed? And so with that, I went down this rabbit hole of trying to be able to create something so my teams could all use it. What I actually found was, uh, a couple things. So one is use, getting our messaging and positioning doc to help frame it was really helpful. So like, here's our messaging and positioning. I also connected it to all our conversations. So I looked at all of our historical conversations. And then also actually, you know, borrowing from like best practices. So, you know, Lenny's podcast has a bunch of stuff on ICP. Like I was in marketing at Salesforce, so I kind of combined things that I've heard that I like to borrow from other people, stuff that I liked of like how to think about building an ICP. And then basically I told it like the output is going to be a Markdown file that is going to be used by other. Agents to go do work. And your job is to make sure that you run this every week. And so it's on a cron job. And so every week it just makes a new file and you can see the drift from the previous one. Um, and it's got everything in there from anti-ICP patterns to ICP patterns to customer quotes to validate certain things, certain segmentations and, uh, certain trends we're seeing. Jobs to be done framework so that you could see what people are using Webflow for, what they're not. What they're choosing for and what they're not. And then, uh, from that once, and the, the, my whole vision for it was to actually be able to make it so that whether it's marketing or RevOps and they wanna do territories differently, they can use this at any given point. But it actually goes much further because I actually think it's, it's creating the basis to be able to ask something super specific. Like I'm competing against WordPress, what are the top, here's my deal, analyze what I've talked about in my deal, compare it to the ICP, of when we win and tell me what I'm missing and what questions I should ask. And so that's where I'm going with it. But my biggest unlock was how do you create that ever-evolving knowledge base? I think in go-to-market it's harder because, you know, not everyone's in GitHub, not everyone's actually creating stores of accessible data that constantly are changing. Um, and for us, a lot of these documents used to just end up in a presentation. Um, and so how do you just constantly update buried in Slack? Exactly.

In a Google Drive. The answer I think is GitHub though, like that, that's sort of what we're consolidating on. Like, how do you think about a central repository that people can push and pull from? And it's in a format that works across many surfaces. We have, my VP of RevOps built something called the Gardener and it looks at business context across a whole bunch of different Slack channels and Notion documents and updates basically like business context. Like what if you are building a thing? 'Cause he's building, you know, like business workflows and process and, and stuff in, in our tools. So what do you need to know about all of those things to do it really well? Like I have a personal one that's all of the things that are important to go-to-market and to me, and it runs every week and it pulls in all of my call, all of my internal call recordings, all of the Notion documents. From, from specific workspaces, things like I've written my to-do list to update this master, like what's important to me file so that any of my agents can, can be most effective. Because if you ask, you know, back to this conversation about centralized versus decentralized, if a rep just asks a generic chat window, it's like, how do I win this deal against XYZ competitor, it's gonna like research a bunch of stuff online and give you a bunch of slop. But if you as the CRO can manage like, this is our competitive positioning, this is how we do these things, these are the new feature releases, then all of those things actually become 10 times more powerful.

And, and here's the most interesting part. I think you just nailed it. Me as a CRO, like there wasn't a world where I could get that type of scalability. That's where I think the unique opportunity is because me shaping how we, how that ICP works, how that document ends up being used is actually shaping a bunch of other workflows. And so I think that that's gonna be how the role evolves is yes, I think everyone is a builder now. And I think for those of us who just love to be curious, it's really fun. But I also think that that's, this is in this lives. The context unlock? How do I use my context to help every rep be more successful, help every customer success manager, help every marketer? And I have a different viewpoint. And so how do I help replicate that? Because I think the hardest part right now is if the world's changing as fast as it is, and especially if you're in tech, how you even, how I have a conversation today is different from a month ago, um, about our product, about the world. And so one thing that it's just too lagging, you almost have to tell it where you're going.

Like, I always jokingly say it's like The Matrix. How do you implant into a first-line, like first-time first-line manager, all of my knowledge about how to solve people problems and orchestrate a team? Like they can get, they can get a lot of that.

Yeah. And you, you, you kind of hint at it there and I think it's spot on. I had this conversation with a manager, a frontline sales manager actually, about this topic, and I was telling her what I need her to do is something I can't, which is she knows her rep, she knows her rep's motivation, she understands the deal cycles, she's been in calls and heard them. It's not just the transcripts, it's actually the tonations, the delivery, all of the parts that make up how that works. I go, if I give you access to something like this ICP doc that you can run against the transcripts of these, You know what questions to ask. Like you can ask different questions so you can coach the rep. You can, you know how that rep responds. And I think augment, basically taking my context but applying hers on top of it is I think the unlock. It's how do I build a, I always am like obsessing with our rev ops leader. Like how do we build the system that everyone can use AI on top of?

Okay. What's your third best use case right now?

Yeah, so the third best use case I think is really around our outbound process. I think right now being able to push to every single sales dev rep a, here's the calls you should be having, here's the script. So like they basically say like, I'm gonna call this company. And right when they say that in Slack, they get a message back that says, great, here's your call script. What it does is it goes and analyzes all the data we have about them, does enrichment, Goes and looks at what's happening in the market, applies the Webflow positioning and some provocative challenger sales into it and says, here's a script that can be useful for them. And so it's like a just-in-time script for each time they want to call someone. I think it's a, it's a really powerful process because our goal is just to get people on the phone. And so I think that one is really a great use case. Sometimes simple is the most powerful, but again, my metric oftentimes is How do I get people on the phone with more customers?

And this, this just compresses the amount of time for them to be ready to pick up a call.

And this one is very similar to when I say like customer facing. I'll give you one more example in this category. One of the things that took us a long time was when we wanted to do a business review with a customer, our customer success manager, if they were, if they were managing the account, had to go look at the data, had to build the slide deck. It was a process. It would take hours and hours and hours. Hours. Now we just start, uh, in Cloud Desktop, we just say, hey, we wanna run a, a business review with this, this is the account. It goes, looks up the information, looks at the history of calls, then it actually builds a Webflow page that's password protected that then lets the customer view all of this. And it happens in like 15 minutes. And so we basically took something that took hours and manual prep and slides, and now it's a dynamic webpage with a password that we can give to a customer so they can review it and have it on their own. And we can make sure we're super aligned on what success looks like.

Have you measured how much time that's saving people?

So the CSMs are telling me that they took what took something that took normally 5 to 8 hours depending, and it's now 15 minutes.

Wow. And then they like the outputs.

Yes. The, what's changed is they're actually, we can just do more of 'em. You know, we can do it monthly. You can increase the volume because it doesn't take time to like prepare it. The call, the problem was never the call. It was always the preparation took a while.

Where do you see revenue leaders investing that you don't think is gonna pay off? Like where are the areas that, that are hypey and, and overrated in your mind?

One of the things that I'm constantly asking myself is that we are enabling so many tools to try to like get the edge. I think we all feel this way. It's like, what's gonna get us the edge? And then you go talk to someone and good salesperson, they're like, this is your edge. The more the strategy is just buying into your edge, I think the more you're going to lose your edge. Everyone's talking about, here's your tech stack, right? It's like, what stack are you using? Because we're all trying to learn together. But I think the best answers I've heard are those that have this mixture of stuff that they're curating themselves. Like you, you gave a great example of like your OpenClaw example with, here is a set of of tools that I'm using as well. And it's a mix. And I think where it gets overhyped is that you can just go buy a bunch of tools that everyone can have access to and that gives you your edge. I think the edge is gonna come from this mix of like what you're building on your own to create the right context, evaluation frameworks, um, integrations to the tools. And then in addition to that, specific AI tools that can help you accomplish certain goals. So I think it's gonna be a mix, but I think I hear all the time like, oh, buy this thing or get this stack. If you have this in your stack, you're gonna be, you'll be leading edge. And I think there's just a lot to what you have to own also next to it.

What's, what's your framework for build versus buy? Like what, what do you think about needing to build in-house versus, you know, use an external tool?

You need to buy systems that you can build with AI on top of it. So you should be able to vibe on top of it and it should give you what you need. So, Things that typically need to exist is, you know, you need governance, roles, permissions, just in anything we do. Like, so like, I don't really want the team to have to go build roles, permissions. I want them to use AI tools on top of roles and permissions, whatever they may be. As an example, um, I don't really want to go and recreate something that we're gonna need for 50, 100 days, 200 days, 1,000 days. And have to manage it over and over again. So I also was like, what's the longevity of this workflow? If something's a one-time, like we need this for this. So like a great example, Artifacts and Cloud are so lightweight. So like, it's great for like a one-time, here's an analysis, like, great, good to go. But once we start needing it over and over again, now you want to actually think about building an application that has like the proper view and it's working. And so my model for the team is, uh, is we should absolutely be building, but I want to be building on top of systems, not that when it's something that we're going to need in the long term. So if it's something that we only need for short term, yeah, go at it, build, build, build a temporary technology, temporary app that is throwaway. No problem. That's kind of how I think about, about it. It really is connected to the longevity and what the purpose of it is. And ultimately I do also ask one other question, which is help me understand why it's an advantage for you to build this versus buy it. Because at the end of the day, it just cut— if you look at it from a dollars and cents perspective, and I actually like using Duolingo as the example, cuz I think it's a great product because, so Duolingo is really fun and easy to use. Um, it's great. If I told you you could have it for free or you can go lower in ChatGPT, which would you pick? You'd probably pick Duolingo, right? Because it's a better, it's better if I give it to you for free. You're like, oh, the value is just, it's a better experience. If I charged you $1,000, you'd probably be like, nah, I'm just gonna go over to ChatGPT or Claude or whatever. So there is a reality of a price to value. And so that's the other part of the framework. And that's why I like to use the Duolingo example because I like it a lot. It's a great product and if the price is right, you're willing to pay for it so you don't have to manage it. You can just like go through the, go through it. You, it has already been pre-designed for you. There's a lot you get. And so to me, with the build versus buy, it's the same thing. It's like, is it worth it to build it or not?

One of the things that I've been thinking about is, is I don't really want to, I don't really want to use tools that trap my intelligence and I want to own the intelligence. And I think that'll probably resonate with you considering our conversation about context capture and knowledge graph. Where I start to get nervous is like, oh, you know, we'll look at all this information and tell you what to do next. Well, you're, you're telling my rep what to do next. Like, why? Like, what's under there? What's, what's the, what's the logic? Like, I, I don't really want black box intelligence. I want to be able to look at like Markdown files and context and understand like what is the prompting behind it. And you know, it's the difference between Gong or, or the sales engagement platforms that are like, well, we'll tell you the next best action, but you can't, like, you can't figure out why necessarily versus like, there's a tool that we use called, uh, Trigg, which it's similar. Like you can build these workflows to tell especially like CSMs what to do next, but you get to, like build the architecture and the intelligence of like, why is that happening? And feed it specific things. Like, no, I don't want to do that in these situations. I want to do these other things. And so I find like the context isn't, or the intelligence isn't held hostage or behind in a black box. Like I want to be able to take my intelligence out of these platforms and be able to actually like influence it in a very specific Man, that, that, that begs a question because obviously the reason the intelligence in some of those platforms are locked in is because that is their moat.

And so you take that away. Yeah. The thing that's interesting is I just don't think that that's a moat that's going to be worthwhile anymore. I mean, I think like our philosophy is, you know, you should be able to work with Webflow via agents headlessly, or you should be able to go into the experience in whichever one you want. Um, and it's, you're gonna choose that based off of whatever has the conduct you need to make the best outcome. So if you want to go into Webflow and click a 3-column thing and say, hey, turn this section into 2 columns, it knows where you're looking and you're not doing that whole experience, which we all had where you're like, no, no, no, this part of the page, I want you to change this part. Cause you don't have to do that cause it has it. But on the other side, like I the other day ran an entire messaging and positioning audit of our website and I did that all from Claude as an example. So. You're going to go to different places, but I think that if you host, if you hold it hostage, there's just not a moat that I think you're going to be able to sustain because agents will just pick other tools that can then, humans or agents will just pick other tools where the agents can get the information in and out to do what it needs to do. To me, it just becomes what gives you the best outcomes and outputs. And so if you're locking it all in, why? I mean, I work really hard sometimes. I mean, it's pretty funny how hard I work to make sure that my Gong calls are not in Gong.

I know this, this was one of the original reasons I turned off Gong is because they want to do the walled garden thing and I couldn't get pipe the information into Salesforce in the way I wanted it to. And I think a lot of people, you know, had that same reaction versus like we use Momentum for it where I could put, push that context anywhere and write my own prompts that can manipulate things in the way I want it. It's, it's, I agree. Like, I don't think that people will allow. That to be a, a lock-in with their own data or intelligence moving forward.

If you're able to provide in your software more context to make an agent more successful. So if you have a set of different pieces of data that it can drive better outputs of a prompt or an agent, then I think that's gonna be the lock-in cuz you're gonna choose it cuz it's better. Cuz you're gonna be like, oh, that it does what I want it to do. So it has all that information, it has the data, it can do better. So I'm gonna pick them. And I think that's gonna be more of a lock-in than like a black box walled garden.

Clearly you're, you're actually very competent with the tools, building stuff yourself. What advice do you give to CROs that want to learn it? Like, where do they go? Where do you point them to?

Well, I would go to your podcast. I'd go to different podcast materials that I think I've heard. And also just talk to your network, ask questions, see what people are doing. It's changing so much that like your best, best skillset is going to be curiosity and willingness to learn. I'm someone who does, gets ideas from like learning from completely unrelated topics. So for example, one of the ways that I build my AI chief of staff is I take notes on every book I read and I love psychology, uh, specifically how human to make decisions. And, and I took all my notes and then from like 12 years, 'cause I had 'em all on my phone and I basically took all my notes and gave it to my AI chief of staff that I was building and said, okay, here's how I want you to look through all my notes. Here's what I, here are the notes I took on all these books. What I want you to do is build a profile of how Avidyan thinks, how Avidyan works, and I want you to figure out The best way you can support him in it, learn all this by understanding the notes he took compared to the books he was reading. And so I used that cause I was like, that's deeply, it's kind of like your Substack example of like The Voice. I was like, it's deeply personal to me. It was my takes on the books. The books are actually documented. So it's almost like an eval. It's like eval Adrian, like his take versus the actual book. Um, and so I would say like. Use all. And so the reason I share that story is I think use— if you want to learn, part of it is use what makes you unique and good, um, at what you do to be able to— and you make AI work for you in that way. Um, that was just my idea, but there's lots of things I've heard that people do.

And where'd you go to learn to use the tools? Is that YouTube and Twitter? You mentioned Lenny's podcast. Like, where, where—

yeah, I would say that part of it is I just like I don't use them until I understand them. I'll look at LinkedIn posts, Twitter posts. I think there's a lot of hype out there, so it's really easy to get distracted. So yeah, I like to go to actual operators. Not everyone posts a lot, but when they do, I try to pay attention. And then the unpopular reality is like non-digital, like in person talking to people who, what are you doing? How are you doing it? Setting up calls where, you know, my RevOps team is talking to another company's RevOps team and sharing back and forth and, and then hearing what the output was. So part of that is like learning on my own. Part of that is really trying to get an understanding from other people, having a coffee with a CRO and asking how they're approaching certain problems.

All right, we'll move to quickfire 'cause we got 4 minutes here. What do you think separates a good CRO from a truly great one?

I think elite CROs are extraordinarily curious, know that they don't have all the answers, but are willing to get in and model the way themselves. Especially in the environment we're in today. I just think that you can't— this job is not about— yes, it's about leadership and management, but it's just as much about being into the details and not instead of, but with your team.

What's the best piece of advice you give to younger leaders?

It's never about knowing everything. It's about being insanely curious, uh, what other people know more than you.

Mm, I like that one.

And a very tactical example is your AEs or your reps or your customers. Anyone who's talking to a customer knows more about that customer than you do.

What's the hardest lesson you've had to learn in your career?

Uh, the hardest lesson is sometimes being right is actually not the goal.

Last question, what's the best thing you've read in the last like 12 months?

I was really interested of how LLMs and this new world probably changed our attention spans. And so I was like in search of a book to help, to that has a bunch of research of basically how to write in a world where people are busy. And so there's a book that came out by two Harvard professors. They did a bunch of research. It's called Writing for Busy Readers. And basically I took the whole book, did my notes as always, and that is now part of my system instructions. But it's not only on like how it writes to me, but how I wanted to write comms out. If you don't wanna read it, good news. Just say, edit this, and then whatever you're gonna post or Slack or whatever, whatever you wanna do, just say, edit this in the style of writing for busy readers. And it most likely will give you an edit that gives you a sense of how it works.

I get knowledge from a book that I don't actually wanna read cuz I think it might just be like regurgitated stuff from another book I've read. I just go to ChatGPT and I'm like, Hey, gimme a full summary of this whole thing. Break down every chapter with the key takeaways and then I'll scan it. I'm like, tell me more about this one. And I'm like, yeah, no, like you, this is just a regurgitation of Kahneman or Cialdini or like whoever the source materials are.

And those are two of the most sourced ones, right?

Constantly. Yeah. Like, uh, now everybody rehashing Atomic Habits, everybody rehashing like, you know, a couple core books.

Well, the good news is it sounds like you hadn't heard of writing for busy readers. So I gave you a new one.

I wrote it down and I'm like, uh, I, I, that's in my follow-up list. I'm going to play around with it in my writing prompt. This was great, man. I'm really glad we could do this. I, uh, think people are really going to take a lot of practical AI adoption insight from it. Uh, and so I appreciate you doing it and, uh, hope to, hope to see you soon.

Yeah, hope to see you soon too. We should get a coffee sometime.

Thank you for listening to the Revenue Leadership Podcast. If you enjoyed it, don't forget to subscribe and you can find a link in the show notes and be sure to leave a 5-star review, share it with your network, and please join me next Wednesday for another great conversation.