Human Signals in the AI Era with Andrew Au of Intercept

AI is reshaping research, marketing, and professional services, but real human insight remains a critical part of the equation. In this episode of the Greenbook Podcast, Lenny Murphy sits down with Andrew Au, Co-CEO of Intercept, to explore why organic human signals remain essential—even as synthetic data and AI-powered personas become more sophisticated.

Andrew shares how Intercept uses its Watchtower platform to analyze unprompted buyer conversations and combines those insights with synthetic audiences for message and creative testing. The conversation also explores agentic AI, hyper-personalized B2B marketing, and Intercept’s evolution from a traditional agency into an AI-native organization. Andrew and Lenny ultimately examine how human judgment, creativity, and innovation can become even more valuable as AI makes execution faster and more scalable.

Key Discussion Points

• Organic vs. synthetic data: Why Andrew sees real human conversation as the foundation for understanding buyers, while synthetic audiences can add value for hypothesis, narrative, and creative testing.
• Listening instead of asking: How Intercept’s Watchtower uses public, unprompted conversations to uncover buyer concerns and signals that structured research can sometimes miss.
• Turning insights into action: How AI and agentic workflows can connect fragmented data, personalize B2B experiences, and move insights closer to sales and marketing execution.
• Reinventing the agency model: Why AI is pushing agencies and consultancies away from labor- and hours-based models toward technology-enabled delivery, experimentation, judgment, and value creation.
• Why human expertise still matters: As average execution becomes easier to automate, Andrew argues that judgment, category expertise, creativity, and the ability to generate genuinely new ideas become increasingly important.

Resources & Links

• Intercept

You can reach out to Andrew Au on LinkedIn.

Many thanks to Andrew Au for being our guests. Thanks also to our production team and our editor at Big Bad Audio.

Transcript

[Lenny] Hello, everybody. It's Lenny Murphy with another edition of the Greenbook Podcast. Thank you for taking time out of your day to spend it with me and my guest. And today I'm joined by Andrew Au. Quick little backstory here. Andrew has a kick-ass PR team because they reached out with the pitch. And normally I ignore those things. And I was like, nope, that's actually a pretty cool pitch. So before we even get into all of that, Andrew, welcome. Appreciate you taking time to be here.

[Andrew] Thank you, Lenny. I appreciate you having me on the show.

[Lenny] So why don't you give the listeners, and me, because this is our first time talking, your origin story.

[Andrew] Yeah, sure. So just maybe as a level set. So Intercept, we're a B2B marketing agency. We specialize in the global tech sector. We started actually 20 years ago as of this year, 2026. So we started in 2006 as a second-year uni student. And so I co-founded this agency with Shaheen, who, we were studying at the same university at the time. And she had this really great idea that, you know, we worked really well together. And so she's like, well, what if we took all the same courses and you did half, I did half? We would technically free up half of our week. And so that was really the whole start of this.

[Lenny] All right. And it worked. We had half of our week freed up. And so we're like, well, what do we do? She came from an entrepreneurial family. I had an entrepreneurial bug. And we started this out of university. And so, like most marketing folks, you start in the consumer game because it's like what you know, right? It feels familiar. And, you know, we grew. We started working with bigger brands. And we actually raised with a private equity partner that specialized in agencies. And so we were year three, I think. We were a year or two from graduation. And I remember Shaheen and I were like, do we graduate? Because we've made commitments to people now. But our parents would have killed us. So, graduated. And we kind of earned our MBA in both entrepreneurship and marketing. And found our way into the B2B world, which is where it was kind of the whole blue ocean strategy, right? Like every corner there's a marketing shop. There are fewer agencies that specialize in B2B. And we're like, God, there's so much opportunity to, I don't know, just do it differently. B2B just felt a bit traditional to us. And so we decided to kind of shake it up and found our lane in global tech. And, you know, we work with Microsoft, SAP, pretty much all the Silicon partners, Frontier Labs. We're about 70 people today. And we do pretty much all global work. So that's the long-winded intro. Not long-winded. And you and your partner still, she does half the tasks. You do the other tasks.

[Andrew] Yeah, pretty much.

[Lenny] Pretty much.

[Andrew] All right. Yeah. She's on the ops and client services side. And so I oversee our studio team, which is strategy, content, creative. So we still maintain the two parts of the business, generally speaking, yeah.

[Lenny] Okay. Very cool. So for the audience, I mentioned the kick-ass PR team. And the basic pitch that reached out was that you have a perspective on the legacy, long-storied legacy, you know, of leveraging open-source data, particularly kind of social media data, for strong signals in the market and utilizing that from a marketing perspective versus the new shift of the focus on synthetic sample, right? And building personas and all the variations of that. And the argument being that organic signal is still the most valuable and most beneficial from a marketing standpoint. And yeah. So correct me if I'm wrong. This is a conversation I want to have because I agree. And it's fast-moving. And what we don't often get on this show is that marketing side, right? We talk a lot with kind of inside baseball, the researchy side, the data side, but the marketing side. So now, give me your view, please.

[Andrew] I'll give you more context. So first, we as a company use both organic, real human data and synthetic data. And I think that there's a place for both. But maybe I'll just take us back of like, why would an agency even invest in infrastructure and research and all that kind of stuff? Because, you know, there are research vendors out there, right, that can feed us this stuff. And then we can go and execute what we need to execute. But for us, strategy is like sort of a tip of the spear for the company. I mean, it is like the stickiest part of our business. And I would say that where we found some frustration in B2B is like you've got all these analyst firms, which are great. We work with them. So there's no shade here on the analyst firms.

[Lenny] Forrester, don't get mad, right?

[Andrew] Forrester, IDC, and the rest of them. We love you. We love you. But when we get a paper, okay, like when we're building a strategy and we look up research, and let's just use a really basic example. IT decision makers, what is their priority? Their top challenges, let's say. I could have a paper that was published in September that says the number one challenge is ERP modernization. And it just so happens to be sponsored by SAP. Or in the same week, I could see a paper that the same question is actually cloud modernization sponsored by Google Cloud. And so a lot of the data sets that we're like, oh, that's a good data set. A lot of them are sponsored. And you know very well, and I'm sure the listeners here know very well, you can make the data say whatever you kind of want it to say, right? Like typically, a lot of these paid things is like, well, where do you need to end up? And then you kind of reverse engineer it. Again, no shade here. I think there's still great quality work the analyst firms do.

[Lenny] IDC was a client for many years. So I get it.

[Andrew] You get it, right? But commissioned research, sponsored research is marketing content. It's not data in our view because we couldn't authoritatively look at this and go, hey, we believe this is right. So we said, well, the most pure, organic, form of research, we go back in the day to like ethnographies, right, where you actually would be in somebody's home and you'd be like, hey, how do you use this dish soap? And which way do you put the bottle? And that's what CPG companies did really well. And so the thinking or the thesis was, OK, well, what is that today at digital scale? And to your point, social landscape, it's the largest living data set in the world. And we said, why aren't people using this for research? And the biggest problem was actually the sampling technique. Because if you look at, and you and I were just talking about this a little bit earlier, social listening, if you just crawl outward on a network, you will probably find the most popular accounts, which is an influencer strategy. You can identify influencers through these social listening tools very quickly. But you cannot run statistically significant research. So we came across an academic paper that talked about using a sampling technique called coupling from the past as a way to randomly sample social networks. And we said, oh, that is a really interesting use case. And so that was sort of how we cracked the randomized sampling technique. And so what we're able to do now is we've got a windowed sample of around 20 million people, where it's across X, Reddit, Quora, Blue Sky, others, where we can actually look at a specific topic within a defined period of time. And we actually look at the entire post history of every single account, which is very compute intensive. And we are trying to find, again, how do you even do this more and more efficiently? But we look at the entire post history. So this is how we can also help identify who's a bot, who's not a bot, based on posting behaviors. We look at bio, we look at behaviors, and we use two sets of encoders, bi-encoders, cross-encoders, basically a filtering strategy. And then we're able to give you statistically significant insights based on any topic. So if it's like DORA compliance and financial services, we can look at people in that sector with specific roles and tell you what do they actually find confusing about this. So very different than a put a traditional survey in front of somebody and say, hey, Lenny, pick all that apply, or pick one that you feel strongest about. That means you have to pre-draft, you have to know already what people think. And you can easily sway the audience.

[Lenny] I love that approach. I've always been a big fan of social listening in its broadest terms, right, as an input. As I mentioned to you, my concern has always been quality of that, and that concern is increasing now. Sounds like you found a great way. I never even thought about that, looking at the history. It's the same thing we would do if I responded on the panel side for a survey, right, to filter out the bad actors. So apply it to social. That's brilliant. Very cool. Now, are you – so in the email that’s sent out, we were looking at the juxtaposition between that organic signal, but also the use cases around synthetic. So are you building – and I use synthetic as a very broad term, right, there's multiple dimensions of that. But are you – now once you've identified the organic contributors, building synthetic off of them, so now you get to here's the natural ingestion of information. Now we're going to ask questions off of that to explore things that may not have been stated. Yes?

[Andrew] Great question. Yes. So we are building synthetic audiences. We've got a couple. One of them is on – we call them Camille and Cortex because you're in marketing. You've got to come up with fun names. So Camille is sort of like off of sort of Chameleon. What that one is, it's a synthetic persona that's actually trained on a lot of the data that you and I just talked about. And what we can do is simulate responses to narratives because a lot of the time it's like there's this whole laundry list that product marketing gives you. Well, what's the thing? What's the hook that you hang your hat on that actually make people care and look at? And so that's kind of what we do pretesting on. And so we get – there's a standardized measure. Number one is stopping power because if nobody sees it, it doesn't matter. And then we go through resonance and a bunch of other factors. So we do that on the narrative side. The other side that we're finding really good gains from is actually on creative. And so this is one that – this data set we did not build. We worked with a neuroscience data team that is running this very interesting study, which is they – it started with around 300,000 people that put on these brain wave trackers. Yeah, exactly. Eye trackers. And so they showed these respondents a bunch of static and video ads. And so what you can actually derive from the brain wave signals is focus, engagement, like positive emotion. The one that I actually like the best is cognitive demand because our creative team, God bless them. I think they're wonderful, Cannes award-winning folks. But when you – and again, strategy has to thought spar with them, right? That's our job is like how do we create that healthy tension? But if somebody has to look at this creative and go, whoa, I don't even know what this is.

[Lenny] What the hell is that, right?

[Andrew] Right? Off the chart. Why are they floating? That's not good. So that's where we have – it's called Cortex, which allows you to kind of predict what response rates are going to be. And honestly, I mean in our assessments, it's been 95% accurate because we have pre and post data. We know what actually works. So I think to the – I think efficacy, there's a role for synthetic data, a lot around hypothesis testing, simulation testing. But again, when you want to know – when you're still trying to win humans on a buying committee, and that's a whole other conversation of how do you win agents versus humans. For us, I think the purest source is human data. What do people talk about when you're not in the room? I don't want to put anything in front of them. I don't want to put topics or response options. I just want you to talk. We will sit back. We will listen. We will observe. We will analyze in the aggregate, right? But we can also in Watchtower drill down a specific conversation, which are anonymized. So we don't know it's Lenny as an example, right? Like we would – it's anonymized. But then we can, I think, find more accurate findings. But synthetic data is a really interesting way to scale some of the insights. So that's kind of how we look at it.

[Lenny] I agree. I think that's the consensus that's starting to emerge. And it's been my perspective that, look, there absolutely is a use case for that, especially early-stage ideation, concept testing, all of those things. Lower-risk decisions. The analogy I always use is, look, I have no problem asking a synthetic persona if the widget should be blue or green. I'm not going to ask the synthetic persona, should I build a factory in Sri Lanka to make the widgets?

[Andrew] Right, exactly, exactly. But Lenny, I want to give you one other practical example too just because I find that this is one that just kind of helps people conceptualize it of what can social – on the flip side of synthetic, what can actual organic data do? So we had a client that shall remain unnamed. Household name, everybody knows them. But they were modernizing one of their solutions, right? It's a platform that I don't know what percent of the world uses them, but a big chunk, probably over 50%. And what they were seeing is that commercial customers weren't migrating fast enough to this new platform. And they were like, well, God, what's happening? Why is this happening? So they go to an analyst firm. That shall remain unnamed. And the insight that came back from them was like, it's timing, budget, priority, conflicts, that kind of thing. And all of that is logical, but –

[Lenny] It's not insightful.

[Andrew] No, I mean, what do I talk about? I'm talking about the same stuff that I would have done without using this research. It's funny. So, Lenny, when we ran it, the exact same query, the number one reason why people weren't accelerating migration is the previous one. It was very, very challenging because the platform architecture was totally different, et cetera. So that actually gave us the insight is that that past baggage that people are talking about and scared about is not appearing on any survey that the analyst firm did. So that actually changed the entire narrative and messaging, which is around, this is the easiest one. Right.

[Lenny] We don't suck. We don't suck. We're going to make it easy for you.

[Andrew] This one's going to be easy, I promise. So if you're an analyst firm or you're an analyst working at that firm, if you value your job, are you going to go to the client and be like, hey, I got it. The response we should include as pick all that apply is that because we weren't so hot on the last one. No. So this is where I find that again, you can uncover a lot of this stuff by just observing instead of like, you know, trying to influence anything. It's like a river in a way that where you've got this cup that you could constantly kind of like dip into the stream and test the water and it is a longitudinal data set. So I can I can see, oh, the reading is different than last quarter, than last year, than five years ago. Then that tells us what are the real trends. So anyway, just as a concrete example on the organic side of like, what do you actually do with this stuff? How does it actually impact marketing?

[Lenny] That's a great story. Let's let's step back. You work within tech, you said, you know, Microsoft and big tech companies. So years ago, I was on the board of Google Consumer Surveys when they launched a survey capability. And, you know, I went in and I said you’re Google. You have like everything, you have all of this data, right? And the interesting conversation was, yeah, but we can't use that. It's very partitioned. It's, you know, the left hand often doesn't know what the right hand is doing. And therefore, in that case, so Google was always the biggest customer of Google Consumer Surveys. Because they needed a way. Yes. The ingestion of all the search and all that good stuff. But they also needed a way to have, you know, explicit consented human data with, you know, real real humans. And that was valuable. And now along comes AI. Right. And it doubles for me, at least, it doubled down the perception of, well, now we're just all the data is fair game. And everybody's just merging it all. And it's all, you know, you don’t… The need to ask questions diminishes because you have access to so much data that now is easy to synthesize and utilize in a new way. What I'm hearing from you is, not necessarily. Right? Microsoft, obviously, you know, heavily invested and not to pick on them particularly, but this is the name that popped out. But they still need to understand organic human perspective and perception in a variety of ways, regardless of the scale of other data assets they may have. And ease of use because of AI. Is that?

[Andrew] Yeah, it's a great point, Lenny, because I think that, again, I mean, we're only dealing with the enterprise tech sector. They all are sitting on mounds and mounds and mounds and mounds of data. To your point. It's not a knock on anybody just because when you get that big, this is what happens. It's so siloed. You don't even know sort of which data can you use compliantly, which data you can’t. And so, I would say that in enterprises, the problem isn't really this lack of data as it is like— so it's not collecting. The challenge isn’t collecting data. It's connecting the data. How do you do that in a really seamless way and how do you make it accessible? You know, again, the challenges that we run into or that we see with our clients is like, sure, they've got CRM databases and all this kind of stuff. But it's human error in terms of input. Like if a customer is logged 10 times, how do you know which is the right record? And so these are like the common problems that we typically see. And so that's why they're like, look, we need a almost like a different view. Someone else has a sort of a different data set that can help us connect it. Now, the other thing that we're doing, though, is I'll give you another practical example. So for some of our clients, they have persona marketing teams. So these are teams that are dedicated to winning the technology office, the marketing office, the sales office, all the different lines of business. And so what we will do is use Watchtower to find insights specifically against these personas. It could even be industry specific. But we fuse that with the data assets that we work with the research teams at our clients and the data teams at our clients. And we fuse these data sets together. And then we create in this specific example, these persona playbooks that, you know, when you look at it's like 200 slides. But what we decide to do is like, well, why don't we create an agent that allows somebody to query this data? Like, I don't need 200 slides. What I need to know is, let's say if I'm an event marketer and I've got an event for CFOs, what's the narrative? Like, what do I lead with? And help me inform the content strategy. So I think agents and AI are just a very useful way to query data and access insights. And so we're kind of democratizing it that way. But, yeah, it really started at this particular client with this, again, on the persona side. And it was meant to just inform them. But then the social team got a hold of it. They're like, wait, how do I get insights? And the event team and the – so in the aggregation and the analysis, making it accessible, AI is super, super powerful, very useful there.

[Lenny] Yeah, yeah, I agree. As we sit here recording on 9/15, in the last week, I started experimenting with GrokBot.

[Andrew] Yeah, yeah. I mean, oh, my God. I hadn't done that yet, right? It's crazy.

[Lenny] Right. That, seeing the agents that can do stuff, right? It's not just go find this information and help me put it together. It's like, oh, you're, like, building things for me. Okay, that's – you're taking action on my behalf. To your point of the – I think that the utility of data, the ability to unlock value creation from multiple data sources is just exponentially increasing, right? The technology just does that. That's what it does. But more and more, in my mind, it is reiterating the need to – where real business value is generated by insight, right? Because the insight is what drives revenue. So, to your point, the persona. We're targeting – we're getting the right message to the right person at the right time, and then they take the right action as a result of that. And the – and sometimes the broad, generic kind of means is okay. Middle-aged guy in Kentucky drives a Suburban. Well, he probably – whatever. He likes Chick-fil-A, right? Whatever. Easy analogies, right? Easy things, right? But in a complex business environment, particularly in B2B, to be able to understand, wait, in their market, you know, this – whether it's a typhoon in Taiwan that's impacted the chip display or chip – whatever. There's so many different variables that have a downstream effect. And understanding, oh, there was – in real time, there was a typhoon in Taiwan yesterday. Uh-oh. You know, supply chain is going to be disrupted. That's going to impact, blah, blah, blah, blah, blah, all these downstream effects. And then building from there, now how do we change our marketing and sales process to understand that? We're not going to get that just through the blah of other data flowing in.

[Andrew] Yeah. No, I agree. I think it's how do you – I think you nailed a key point here, which is are you creating real value or are you just creating other data that kind of sits and doesn't get actioned on? And so just to give you another sort of practical example of, I think, the value that can get unlocked with AI for these account-based marketing programs, which basically is you've got a set number of accounts. I've got to go win, let's say, these 500 accounts, right? JPMorgan Chase being one of them. Well, again, you go back to the problem that a lot of these companies have, which is data fragmentation. I don't know where it is. The data is. I don't trust the data, et cetera. And so what we're able to do in an agentic way is actually say, okay, you need JPMorgan Chase? Okay, so here's what an agentic workflow would look like. We intake the company. We've built a workflow where we can cross-reference against five different databases that look at intent, that look at technographics, that look at different personality assessments based on members of the buying committee. We can pack that all together to give this sales rep a direct view of this account. JPMorgan Chase is surging on this. Here are the people on the buying committee. Here's Lenny's personality assessment based on LinkedIn voice. And then we can actually take the assets that are already developed by marketing, curate them to say this is what is most likely to win Lenny and team, different assets for different people. We can even customize these assets, pack that up in 48 hours. That's the kind of new, I think, era that we're in right now is being able to take data from insight all the way through to your point on action. Because the sales rep doesn't give a shit about this data set. Give me the tool that I need to go have the conversation. And that's kind of where we're trying to push to. I mean, we are doing that today, by the way. That's not theoretical. That's a real example that's in production at scale. But I think what's interesting for us is people tell us, they're like, hey, you guys are kind of like this research firm. But then you're also the agency thing. And you're able to take it from one end of the field to the other. And I think that's what's missing in a lot of the scenarios that we see is there's lots of great data. Nobody's doing anything with it.

[Lenny] Yeah. I was, right before this, was in a conversation about what does it look like to federate data with a framework around its utility for decision. So based on use case. Because so many companies are sitting on, to your point, they're sitting on data that's a useful input. But it's neglected often. Anyway. There's a whole other conversation we could have around all of that.

[Andrew] That's another episode.

[Lenny] There we go. So you built an agency 20 years ago, roughly. Did I get that right?

[Andrew] Yeah. Yeah. 2006.

[Lenny] It sounds like you are much more of a tech company at this point. Right. So that process from a service-based, human-driven into now the human-moderated, human-orchestrated, but technology-driven organization. What has that been like in the trenches as you've been adapting, particularly in the last two to three years, your business model and structure?

[Andrew] It's a really good question. I would say as a general trend from what I see in the data, agencies are generally flat or slight decline. Because right now you've got AI generally eating some of that commoditized work. Yeah.

[Lenny] The creative testing.

[Andrew] Fortunately, knock on wood, we've had consistent double-digit growth over the past few years. And I think the reason why is we really leaned into this. Let me take us back a few years, sort of pre-ChatGPT moment. We do work with some clients on their... Everyone's got a different name for it. But basically it's a team that works on technologies that won't be commercialized for another two or three years. Because it requires this level of technical acumen to go, if we're talking about orbital space access for compute and stuff like that, who's going to get that stuff? And so generally we get these hard problems like that. So our team gets plugged in. So anyway, pre-ChatGPT moment. We're kind of like, oh, this is an interesting technology that I think is going to completely disrupt and create new opportunity for agencies. And so this is where we started experimenting early on this stuff. And it's like, well, how do we bring agentic workflows into the mix? And so we built a technology department in-house because we figure we are probably going to be more of a technology company than an agency. And so the past couple of years have just been like, it's crazy. It's fast. We kind of had to split our AI workloads into lanes. One of them is AI-powered services that are productionized, that we've proven works. And so unit economics has to change because people are not going to pay the same thing for human-generated craft as they will for AI-generated craft. The value prop's different. And so this is where I'm also thinking about lawyers and professional services in more general. It's like, why is my legal bill still this and not this? And so we were like, this is happening whether we want it or not. So we need to productionize it and actually create almost like a product strategy around this. And so what you would lose on potentially margin on any individual thing, you kind of make up for in scale.

[Lenny] In volume, yeah.

[Andrew] Exactly. So that's one side of the house. The other side of it is what we call frontier work, which I mean, I know everything's frontier, frontier, frontier. But frontier work for us is this is like the work that has just never been done before, that is kind of scary, that comes with risk and all that. So I'll give you an example. One of our clients was saying to us, I want to take this idea of personalization to a whole other level, data-backed and all this stuff. So we came up with this idea that what if we created this one-on-one content engine? Meaning if somebody were to download an e-book, because B2B marketing loves e-books and white papers and all the rest of it. So what if that one e-book could not be a static e-book but be a living asset? Meaning if this e-book was on data center modernization and Lenny, you downloaded it versus I downloaded it, we actually should get different versions of this. Because we're coming at it from different ways. And so we created an agentic workflow where somebody requests an asset. It actually does research on the person, like the individual and the company, and will change the narrative or the positioning of the paper. It will change the— core content of product: unchanged. Because it's like the product does what it does.

[Lenny] But the context changes.

[Andrew] Thank you. Context changes. The case study that gets selected changes. Because if you're JPMorgan Chase, I want to see a financial services case study where you understand compliance. I don't care about manufacturing. Or I don't care about the FED or SLED. That's not my game. Also, your call to action should be specific to where you are. If I'm very serious at JPMorgan Chase and I'm like, I'm ready to make a decision and so forth. Okay, well my call to action, maybe don't send me to a webinar. Send me to a local workshop that's happening or a one-to-one offer kind of thing. So that same e-book is now 50,000 copies of that e-book, which is efficient to do. It's just as efficient as if you were to customize five or 10 e-books. But now you have 50,000. That is what we call frontier. So we co-fund that stuff. It was Bezos that said, if you know it's going to work, it's not an experiment. I'm paraphrasing. But this actually has to be an experiment. There has to be a, I think this will work. I don't know if it's going to work. And then we fund up to 50% of the production cost. But we share the learnings. We share the IP, et cetera, et cetera. So that's kind of how we split the two. And I would say that's been a driving force behind our growth. And do we call ourselves an agency? I don't know. Does agency, the word agency, just change over the next few years of what an agency does? Because traditional means of traditional definitions of agency I think is just going to die? I don't know. What are you seeing on that? How do you see agencies in this whole thing?

[Lenny] Well, so again, audience, Andrew and I had not met prior to this. But I am fundamentally a consultant. And I have, to answer your question, I wear lots of different hats, involve lots of different organizations in a variety of ways. And one of my primaries, I own a— co-owner of a consultancy. And the same way clients come to us, here's the problem, help us figure it out. And increasingly, the process, the workflow is agentically driven. So that's, again, my experimentation with GrokBot, the next level for myself. So I think what we think of as agencies, whether that is a marketing agency or a research agency or a consultancy, strategy consultancy, it's all the same model. It's a human-driven model that has always had inherent scale limitations. And the technology reduces the scale limits. The bandwidth is expanding. I'm busier than I've ever been in my entire career because I can take on more work. But the work itself, I am an orchestrator. I am the facilitator. I am the IP creator and scaler. And that's what I'm hearing from you, is a very similar process. That the human insight, intuition, that weird step that happens in our brain that we can't even explain, where we jump from point A to Z, right? It's like, okay, I just know these things are connected, you know? That, I don't think that we're there yet. The inference component of AI is not there because it doesn't process information the way that we process information. So that's where that orchestration component, the aha moments that come in for us, of like, hey, can we do this? Whether it's for a client demand or even an internal idea for a concept or a product, that's that orchestration facilitation process. The everything else, the grunt work, that's done via agents. But I'll tell you, here's the interesting thing. I was reluctant to use AI initially when it first entered because I thought this is going to devalue me. So, right? Because I'm selling time. And fundamentally, an agency sells time, right? That's the multiplier. What I found is actually that is not the case. I know, and sorry, we'll get, I think you'll get what I'm trying to, what I'm going through all this. I sell value. I sell impact. And that has actually gone up in price. So it's not tied to the process. The unit economics of that to navigate for a large organization is really freaking hard. So, you know, to transform that way is tough. Smaller organizations, just me and a couple guys, we can make that transition. You've got, you're smaller compared to, you know, Publicis, right? Whatever. You can pivot and adapt. But the big guys, I think, are really struggling to change the infrastructure, the organizational structure, their business model, their pricing, all of those things to adapt to being the value creator from an orchestrator versus the doer.

[Andrew] I think we share a similar view there because I think you hit on a couple key points, which is the value creation bit, right? And there's the taste and the judgment that becomes so valuable. And, you know, you give one person a set of AI tools, what they can do in a day with the right skills, experience, expertise, category depth, etc., versus you give to a team of people that doesn't really have that. And it'll take them months and they still won't be able to do what this person does in a day or a day and a half. And so I think that when you think about AI, it is very good about finding the mean or the average, right? Well, in consultancy work or marketing work, you're trying to stand out. You're trying to be different. And so I think that without that override of like, huh, interesting idea, AI, but I think we should go here, you retain your role as a thought leader and AI becomes your thought partner and not the other way around.

[Lenny] 100%

[Andrew] Because I see a lot of people do the opposite, which is just like hit the easy button and you get this mediocre AI slop. And so I actually think that the cost of average work goes down, like it plummets. But the cost of really good work, I don't care what field it is, the cost of really good work goes up. And I think that that's what we're all trying to do is how do we lead with our human intuition and get all that experience, that lived experience that an AI model will never have? How do we use that as our advantage? And then to your point on the largest organizations, I mean, of course, I'm biased because I'm on the other side of the fence here, but I think they're in deep, deep, deep shit. Because they, and I just think about our industry's marketing, the Publicis and all the rest of them, they are a body shop. Typically, how they go to market is like, okay, well, we do this RFP. You have this body of work. How many bodies do you need? 100? I'll give you 100. Okay, maybe I might give you 110 or 120 for the cost of 100. And so it's like how many bodies can you stack in a meeting? Well, when people refuse to pay time and materials type of thing, and it's really to value, a value-based pricing, billing, et cetera, I think they're in trouble because the change management, you can have the right idea. You could have the right technology. You could have it all figured out. Good luck trying to change 100,000 people. It is hard enough at 70 people, I'll tell you, right? And same thing, like with a small crew. It's still, yes, you can kind of pick the business up and move it and retrench it, but it's still a little painful. Oh, man, I don't know. In the largest organizations, I think that it's going to take a long time. It's going to take years to have this properly implemented without blowing up the entire thing.

[Lenny] Yeah, I totally agree. Now, on the flip side of that, so one of my business partners, Greg Archibald, shout out to Greg, we have this debate all the time. When AI first started, his default was we're 10 years out. And then as things started, I was like, nope. Timeframes are compressed. But the point that he was making, and folks talked about this as well, is change management, change, buy-in, implementation does take time. That is the human component. And what I see happening at a broader term is that there's still a bit of a mismatch. And maybe you're closer to this since you're on the tech side. The organization, well, here's one. Organizational adoption, enterprise adoption of AI tools obviously is sky high. Impact is still very low, right? ROI is still very low. It's still early in the game. Now, consumer adoption is small, right? Consumers have not changed their behaviors much yet. So there is something in, even though I believe that that change cycle is going to be compressed, there is still time. The solutions have outpaced the human ability to adapt and implement. But I don't know if that is a defensible, certainly not moat, it may be a speed bump for large organizations. It's cold comfort that, okay, well, our clients haven't gotten there yet, so we still have a little time. Okay, but we're still a little time, we're talking a year or two, especially with the amount of capital flowing now into the forward deployed engineers. And, you know, I mean, they're pushing this. And there's too much money at stake. It's going to happen.

[Andrew] Again, I think you hit on a good point, which is the pace of change is no longer the capability of the frontier models. It's just capability of people and change management. And that is what throttles us, right? Not the frontier capabilities. And, you know, even if the models never evolve past today, and we still stay with Fable 5.1 and GPT-6, which is incredible, there are trillions and trillions of dollars of economic opportunity here for everybody, even if the models never evolve past today. So I think that it is around human change and that capacity for change management. And I think what we're seeing is this revolt, in a way, against AI, where, you know, if you see the recent tweet storm, right? It was Jacob Coxon and the rest of it. It's like you've got this anti-AI movement, anti-data center. And so I think that that also will play a role in throttling the speed a little bit. However, and this is why I think this is such an opportunity for entrepreneurs, small and mid-sized organizations, et cetera, is that while the mega, mega enterprises of any space have battling all this change management and politics, you've got these smaller teams that, you know, all rally around a thesis that their alignment there work really, really fast. I think that they become, you know, the new, they displace the incumbents. I mean, look at Anthropic. How many years have they been around? You know, probably going to be a trillion dollar, you know, or close to IPO. They're technically a startup. Like, that's the kind of potential trajectory companies can have. And I think the, again, no knock on Publicis, but I'll just use them as an illustrative example. I think that they will realize that that mega ship, that that Titanic can't move fast enough. And they're going to have to figure out a way to diversify and another strategy to go at this because it's just too hard to do that change management.

[Lenny] Well, they bought, I think the play there has become infrastructure. I mean, Publicis has bought LiveRamp, right? So owning the pipes is part of that. Salesforce that, so I've paid attention, but the rumor is Salesforce is going to buy ListenLabs. ListenLabs, I happen to know, you know, have a deep relationship with them. And it's a two-year-old company, right? The, for, you know, $2 billion. Anyway.

[Andrew] That's the AI interviewers, right? 

[Lenny] Yes. I'll tell you, I'll say this publicly, which I haven't said publicly. The, when they first started, they showed up in my, I do a startup competition and they won the competition in 2024. And the CEO offered me an advisory role for equity. And at that point I didn't take it. Stupidest fricking decision I ever made in my whole life. Anyway.

[Andrew] I’ve made many of those.

[Lenny] Okay. Yeah. I have too. It's probably not these two. If you ask my wife, she would rattle off a whole bunch more that were probably stupider. But the point is, these companies are adapting to become infrastructure sales. You know, the SaaS apocalypse of a few months ago, right? The, sure, where value accrues is shifting. It's being deployed in different ways. And the big companies, their best bet is to really focus on inertia. So switching costs, it's a pain in the ass. So if they can adapt and kind of stay where they are and meet the client where they're going, then they'll be fine. But I love the point that you made on the smaller, nimble companies. What an opportunity to adapt. Where the, you know, to fit that gap. Where the big companies, yeah, we're going to get there. We got the money. Look, we're going to do this, but it's going to take us a year. In the meantime, for buyers, it's like, well, we want a solution today. What a great opportunity for companies like Intercept to step up and say, hey, we got this. And create real value and scale.

[Andrew] Well, you know, just in the spirit of data, one of the things that we actually looked at for the topic of how buying committees use AI, right? Like in terms of how they go find vendors. What was interesting is that in the consideration phase where they're actually looking and evaluating vendors, the number one thing that they're actually looking for is overlooked vendors or vendors they didn't know. And so I found that interesting to me because, and we did, by the way, this is Watchtower powered data, but we also did 10 in-depth interviews as well to get more context. And I remember, because I was part of some of the interviews and sitting in on them, and I remember somebody saying, again, these are all large enterprises. But he's like, I start with the thing I know, the vendor I know, and then I use AI to find everyone else around it that maybe I didn't know. And so traditionally it was, oh, I got to be in the Gartner Magic Quadrant or the Forrester Wave, which still is credibility, right? But buyers are also looking for these more niche-y, still, again, that can meet enterprise requirements, but because they're looking for maybe an advantage, maybe they're trying to use AI to round out their thinking. So I think what an opportunity when I saw that insight, I'm like, wow, that's what people are looking for is overlooked vendors. Huge opportunity for mid-market and non-enterprise.

[Lenny] 100%. So part of what Greenbook does, and you may not know this, we put on events, and we started those events 15 years ago because we were good friends with P&G. And P&G was asking us for recommendations for vendors, and then they said, well, would you put on an event and bring them to us? And we did, so our first Insight Innovation Exchange event was basically a stealth vendor fair for P&G to help them find the new companies. Now, that event, that series is global, and our U.S. event is 1,500 people. But fundamentally, it's still the same. What drives buyers there, in our case, is we're showcasing, yeah, you see all the established guys, but we carve out. That's why we did that startup competition, to showcase new companies. So I agree with you, and there's a whole other topic that we could go into. I want to be conscious of time. We've been going for 51 minutes. I get the sense we could probably go on for another two or three hours if we had the time. It's been a great conversation from my perspective. I really appreciate you sharing your journey and what Intercept has done and how you're adapting. I think it's very inspirational because what you're going through is the same thing companies on the traditional resource side are going through. And one of our goals is to help everybody realize you're not in this alone, and here's folks that are navigating through those things. What did we not cover that you would like to cover?

[Andrew] I was hoping we could cover not just the organic, but synthetic. I think we did that. I think we kind of touched on this. The thing that I hear a lot is, is this game over for us? Like humanity and job displacement, all this kind of stuff. I think I'll just share maybe some two cents on this, because I tend to get this a lot. Because when I do keynotes and what have you, it's probably the number one question I get. And people are like, oh, what about the ghost GDP thing? That paper that went viral a little while ago that was like, basically companies use AI to make everything more efficient, and then everything just goes to shit because you eliminate your customer base. You have no more customers to pay for your products. And I think that why I believe, or how we can ensure that that doesn't happen, is really focusing on coming up with new ideas. So AI is a really good efficiency driver, whether it's data, research, strategy, content, creative, et cetera. But what we need to do is really carve out time to think on innovation. And if you don't have that time, if you aren't carving out that time to parallel this to, I think it was Guy Kawasaki on the whole S-curve thing, we will die on that S-curve if we don't jump it. And the only way to jump it is you got to know what S-curve you're going to jump onto next. So thinking about spending more time on innovation is, I think, what we need to do. And we have, everybody that's in the workforce, category depth, experience, skill set, expertise, don't forget that that has a ton of value. And we are so relevant in this era. So I think it's just like having that right mindset because it is overwhelming. And some days I wake up and I see stuff and I'm like, it's hard not to have that sort of defeatist attitude in a way. Like you kind of go through ebbs and flows. So I think that's just sort of the one final point I would make is that this is the biggest game, the most important game that we will probably play in our professional careers, many of us. And what a time to be on the field, like to see this happen and feel it. And there will be opportunity. It's going to be messy. There's going to be that messy middle where, yes, there's going to be some displacement and jobs will change and all that kind of stuff. But I still think from a research standpoint, the people that know the customer the best, whether it's a human or an agent, is going to have a ton of value because that is how you get the best ideas. So there's always opportunity to innovate. And I'm genuinely excited for what the future will look like in a few years. Though, I also share the concern of like, oh, what's that messy middle? And we've got to make sure that we make our way through it. So that just be my sort of final thoughts.

[Lenny] Right there with you. The analogy I use is, look, we can go Star Trek or we can go Black Mirror. Right. So the potential for the dystopian hellscape is it's there. Absolutely. I don't think we'll get to Terminator level. Right. Hopefully not Matrix level either. But to your point, my trust in humanity and our innate ability to solve problems and to create, because that is what we do as a species. It's what we have always done as a species. You know, from fire to the wheel, to the steam engine, to nuclear power, to the internet age. This is what we do. And I have utter confidence that we will do that again now. To your point, it's going to be messy. My son-in-law works in IT security. He was laid off last November. Found a new job. Just found out he's going to be laid off again. And that stinks. But look, son, we can go do these other things. So it's an opportunity to recreate yourself and to tap into opportunity when that messy middle happens. None of that’s comfortable. I'm utterly sympathetic. I started my own business 25 years ago because I never wanted to be laid off again. Right. Because it happened to me. But we'll get there. So I echo that. Andrew, I really love this conversation. I'm glad that we connected. I hope that we have a chance to talk again, either publicly or privately, because I thought this was a great conversation. Where can people find you? How can they connect with you to have conversations themselves?

[Andrew] Well, first, I mean, this was a ton of fun. So, again, thank you for having me on. But, yeah, you can find me on LinkedIn. Andrew, last name is Au, spelled A-U. Or you can go to interceptgroup.com and find more information there. But, yeah, Lenny, I'd love to stay in touch. And I'm sure we could spend hours chatting on lots of stuff.

[Lenny] Absolutely. I suspect so. Really appreciate it. I'm sure that the audience really appreciates it as well. So with that, let me give a couple shout-outs. Thank you to Emma, our producer, who wrangles me, which is a full-time job in and of itself. To Big Bad Audio, who makes sure that all of this comes out looking and sounding great. To our sponsors, and most of all, to you, our listeners. Because without you, Andrew and I wouldn't have had an excuse to connect like this. So appreciate that. Hopefully, you'll get value out of it, too. But that's it for this edition of the Greenbook Podcast. We'll be back with more soon. Thanks a lot. Take care. Bye-bye.