AI Governance and What Marketers Are Missing
Scott Brinker · April 3, 2026
ai governanceai trendsai agentsdata layer
### Introduction to Scott Brinker
(0:05) Hey everyone, I'm David Berkowitz and welcome to another edition of AI Insiders from Architectures AI Marketers Guild. It is a true pleasure to introduce an old friend, Scott Brinker, Mr. Chief Martech himself and someone who's been trying to make sense of so many of the worlds that we're in for so long. And I'd even say Scott that you're usually more
### Staying Ahead in a Confusing World
(0:32) successful than most. Even if I wind up in a haze of confusion every damn day that I don't think is going away anytime soon. So can't blame you for that, can I?
>> Well, why not?
>> Why don't you share? Because I feel most people on this call know you, have at least read some of your stuff, know some of your incredible little
### Scott Brinker's Current Role
(0:57) landscapes and all that, but why don't you at least share a little bit about what you're up to today?
>> Sure. For about eight years, the previous eight years, I was actually building the Martech ecosystem for HubSpot. Which was great fun. But I left that back in September. And this work I've been doing with Chief Martech,
### Challenges of Tracking MarTech Trends
(1:17) an armchair analyst to this industry for 20 years. That always has been a side hustle, labor of love, call it what you will. Decided to actually go full-time with that. Which to your point, David, my goodness, even trying to keep track of all this stuff 24/7, I feel hopelessly behind every day in these announcements. Trying
### Current AI Interests and Inquiries
(1:42) to chip away at it a bit at a time. Doing Martech analyst and advisory work full-time.
>> I mean, obviously, I've seen some of your work. We're all in the AI space, market or not, but what areas specifically are you most interested in right now? Are people calling you most often for?
### New Data Layer Report
(2:08) Not to be promotional on this, but I will share there's a report
>> It'll be a little promotional. We published a week ago with the folks at Databricks. And although this was something sponsored by Databricks, the way the report was researched and written, it applies to any of these data cloud Snowflake, Google BigQuery, whatnot.
### Focus on the Data Layer
(2:34) I would say the two things that I've been most interested in is one, what is actually happening at this data layer? Because for all the excitement around AI, you all know it's so much a function of the data that you feed into it. And for the most part, the data layer of most not Martech stacks, but company tech stacks in general, is still, how do I say diplomatically,
### Creative Possibilities with Better Data
(3:01) immature. There's a lot of opportunity for that to get better. And I think as it does start to get better, it opens up a lot of creative possibilities, not with AI, but all the things we can do with that data and AI. That's one thing I've been focused on a lot. And then the second one is
### Rise of AI Agents
(3:23) clearly this is the year of AI agents. You can't go a LinkedIn post without seeing AI agents. There are many kinds of agents, right? There's the agents we're using behind the scenes in marketing. There's agents we as marketers are deploying that are customer facing, everything from customer service chatbots to shopper concierges to, I'm sure we all
### Buyer-Controlled AI Agents
(3:48) love AI SDRs. But the agent that is actually the most intriguing to me is the category of agents that are not marketer controlled. That they're buyer controlled. We've seen this beginning with the shift from SEO to AEO, and buyers increasingly leveraging ChatGPT and Bard and Gemini and all this to take a very different
### Evolving Marketer-Consumer Engagement
(4:15) control over their journey. But I think that's one example of multiple kinds of AI agents that are starting to pop up. We're starting to see some hints of this in the email space, AI control of the inbox that again, it's changing the relationship or the channels and the mechanisms by which marketers and consumers, customers
### Humans vs. Bots
(4:41) engage with each other. And that's very nascent, but it's probably the area I'm most fascinated by.
>> Well, it touches on something that I keep going back to and everyone I work with, they have to hear a little bit of my soapbox on this, and that's especially now, in this increasingly agentic era, that there's
### The Two Audiences: Humans and Bots
(5:07) that we basically have two audiences and it doesn't matter if you're B2B or B2C. It doesn't matter if you're talking to an individual consumer versus a Fortune 500 CEO, that you've got humans and you've got bots. And the humans all have more in common with each other. I have way more in common with Tim Cook or Oprah or some person who's running some startup out of Lagos
### Bots as an Alien Species
(5:36) now. We all have some shared characteristics and have way more in common with that Lagos founder who I've never met than I do with bots, right? And the bots, whether they're coming from Claude or they're coming from Google or they're coming from G2 or any other company, they all behave in a more similar way as this kind of alien species. And so it's
### Acknowledging Different Targets
(6:09) if you start acknowledging that you've got these two very different targets, then I feel a lot of what you have to do unfolds from there. Does that gel with what you're seeing or you have any holes to poke in that?
>> Well, to your point, and you know this better than anyone, marketers have always been doing
### Marketers Adapting to Google Bot
(6:32) this now for one, two plus decades, where there was our human audience and then there was the Google bot.
>> Now granted, it was the Google bot, but obviously the whole SEO industry, we put a lot of effort into actually mastering how do we talk to both of those audiences simultaneously.
### The Challenge of Diverse Bots
(6:52) I think what's perhaps both interesting, challenging at the moment, is it is no longer one bot. It is this increasingly diverse set of bots.
>> Although to me, there still is one notable difference is that with the Google bot, when the idea was that a human would land on that page that surfaced from Google's list of links,
### AI Reinterpreting Information
(7:18) it still had to be readable, right? An SEO expert who wanted to overoptimize keyword stuffing and all this stuff, if they made it unreadable, then the conversion couldn't happen. And now when most of that info is being reinterpreted and synthesized by the LLMs and by other AI tools, then we don't even know what version of this the human's
### Inconsistent AI Models
(7:44) actually going to see later.
>> That's a fair point. And again, there's a lot of diversity Gemini does that is different than ChatGPT is different than Claude. And worse, they're not even consistent within themselves. New model anytime. They're constantly evolving. And that combined with the lack of visibility,
### Early Stages of AI Agents
(8:07) it does make it quite a game. I don't know. I keep thinking about that movie Dodgeball. He puts the blindfold on, and they're like, "God, and it's going to be hard for him to see." It just that's a little bit what it feels at the moment. But to be honest, I still feel that's very early steps. What is
### Buyer-Side Agents and MCP
(8:28) intriguing is the notion that we are going to see agents on the buyer side that actually doing more than reading content, and synthesizing from that. But to the degree that we're able to expose things that they're able to do through companies that are exposing some of their things through MCP in the B2B space, are already seeing this
### Practical Examples of Agents
(8:52) thing where people have their work agents and they're, "Oh, can you fix this scheduling thing for me?" They go off and do it. And it's some of these things that are, I think, relatively new channels.
>> But are you, do you have any either B2B or B2C? Do you have any practical examples that you've used, you've seen, you've worked with other companies that
### MCP for Newsletter Management
(9:14) are doing in some way?
>> Most of the stuff that I've been doing with MCP has been back on the orchestration with things across your stack. For instance, I use Beehive for delivering my newsletter, and they released their MCP server. So, I'm in Claudius. I'm thinking about, hey, can you check on this thing? What
### New Interaction Channels for B2B
(9:36) happened with this audience? How's this open rate been changing? Oh, I had this sponsor here. I had this go. The fact that it can basically go behind the scenes, get that information from Beehive, this is a new channel in which Beehive interacts with me that's not through their app. We see a lot of those examples today on the B2B side,
### Consumer-Side Agent Emergence
(9:58) but I don't see why that can't start to emerge on the actual consumer side. Obviously, what Chetch tried to do here with instant checkout that they've pulled back on, but even then, what Google's trying to do with UCP, it all feels very science fair project stage right now, but it does seem to be pointing directionally into this thing of people will be
### Google Search Console and Base 44 Super Agent
(10:21) leaning on these agents to actually do things, not synthesize content.
>> That's what you're saying with the Beehive MCP, it brings to mind something that I've found especially useful lately because one of these odd things, as I've been building more, I've been learning or relearning Google Search Console because now I have to go and make
### Base 44 Super Agent Functionality
(10:46) sure that these sites are potentially visible. And I use B 44 for so much of my building. Base 44 past couple weeks released a super agent tool to work across your sites. And one of the things it'll do is connect to Google Search Console directly. It can then set up error monitoring. It can fix a lot of the errors itself and then tell you what to do with your
### Powerful Code Rewriting
(11:13) individual properties and then it'll often rewrite code for me to better accommodate what Google Search Console is doing. So that I can then go from where I'm actually building this stuff and no longer look at Google Search Console itself. This stuff feels pretty powerful and useful.
>> And when you think about it,
### Rapid Advancement of MCP
(11:40) it's been a matter of months. Depending on some of the ones that were really forward-leaning, what about 14 months since the entire notion of an MCP protocol was introduced. The speed at which this stuff is moving and advancing is pretty wild.
>> I got an email literally an hour ago saying, "Your
### The "Lake Wobegon" Effect in AI
(12:05) super agent learned 130 new tricks with all of its new integrations." So, I'm about 130 of those behind by the time I got on this call, right? I've been thinking about this lately. The Lake Wobegon effect, all the children are above average. I feel we collectively now are in the inverted Lake Wobegon effect. Everyone feels below average and trying to keep up with this stuff.
### Credible AI Startups
(12:30) So maybe that's me.
>> Do you have some because there's so, I could literally launch an AI startup that might not actually be very good but looks credible on the surface by pulling some stuff together with these code and tools right now, right? And I can give it
### Distinguishing Credible AI Businesses
(12:52) a nice domain and logo and all this stuff and it can look a very credible business. But it might not actually do anything. Do you even have some threshold for what you'll pay attention to, or will you look as far out on the fringes as possible? How do you draw some line in your world?
>> We have generally gone
### Validating Small AI Companies
(13:18) pretty close to the edge. We try and validate that things actually is a legitimate company. There's signals you can have the ability to contact it. What's their presence on LinkedIn? You can cross validate this stuff. But we look at a lot of companies that are very small because I feel there's lots of cases where you get these great
### Value of Curated Landscapes
(13:40) curated landscapes of hey, these are the top 20 products you want to pay attention to, and I actually find those things very useful because they are what we typically think of as the head of the tail or the main things. But almost by accident, I've ended up in this mode with the MarTech landscape, really looking far down the
### Analyzing the Long Tail of MarTech
(14:02) long, long tail, which trying to look at individual companies on that is not very useful to anyone. But actually looking at it in aggregate and how it evolves over aggregate and in which categories and which ones stick around and how quickly do they churn, and what's the percentage rate that grows, it continue. We're in the middle of working on our big
### Preview of MarTech 2026 Report
(14:24) State of MarTech 2026 report now, and it's full of insights of, okay, for individual companies, not super interesting. Patterns in aggregate, very interesting to start to see what these dynamics are showing.
>> Are there any patterns you give us a preview of?
>> One of the things that's actually very interesting is this. Don't spread this one on
### CMS and E-commerce Takeoff
(14:49) LinkedIn, we'll be between us, but one of them is there's quite a takeoff in the CMS and the e-commerce space. E-commerce had a nice wave around the pandemic era, just because of the big shifts, but it's starting to settle down. CMS has largely been a pretty static category for quite some. You could argue
### Category Acceleration and Churn
(15:14) it's actually the oldest Martech category out there. But we saw significant acceleration in both of those categories, and almost double acceleration, because not only did they grow, but they actually had a lot of churn. A lot of older companies have basically left the space or the sector, exited, acquired, or caught on fire,
### Growth Driven by Rethinking CMS
(15:37) whatever happened. So actually, for there to be growth, there had to be that much more to cover up the deficit from those who exited. And when you start to dig into that, and you look at, it makes sense that it's some of the things we're talking about on the CMS side. People are rethinking, okay, how do we manage
### Generative AI and AI Agents in CMS
(15:58) the website in this environment where both, hey, we can leverage generative AI as part of what we're doing here, but also, oh, we've got these AI agents of various kinds and flavors are engaging with us. How do we think about serving that audience? How is that built into this? That was interesting. That was not on my bingo card to see that clear and crisp of
### Content Marketing Category Decline
(16:23) a renewal of the CMS category.
>> Okay. Interesting. Is there anything that was top of mind the past few years that is not as prominent right now? What's on its way out?
>> Well, the category that suffered the most this past year was actually the content marketing category, which part of it
### LLMs and Content Creation
(16:57) is because it had a, it was the category that actually had some of the fastest takeoff around 2023, just because as soon as LLMs were out here, people were, oh my goodness, we could wrap this and do all sorts of content things with it. And a whole bunch of people did. But playing out a few years down the road, there's a big difference between
### Sustainability of AI Wrapper Businesses
(17:21) wrapping an LLM, and as you're saying, hey, I can wrap this. I can put it out a website and it looks credible, okay, this is actually a sustainable business. Even when you're talking about small businesses, still is it a sustainable business? Is there a competitive offering there? And probably not a surprise, but saw
### Survey on AI Use Cases
(17:43) a pretty big exit from that category. The other thing I would say is this isn't about the landscape, but in parallel to the landscape, we ran a pretty in-depth survey of 70 marketing use cases with 28 marketing ops Martech leaders. And what we were primarily asking about was, okay, for this use case, are you using AI within an existing SaaS platform? Are you
### AI Use Case Insights
(18:13) using a new AI native tool for this capability? Have you created something of your own with AI? Or no, we're not using AI, or we're not doing this use case. And that's, and of course, then we split out the data between B2B and B2C. And that's proving very interesting. But one of the things that came out was, boy, a rush of those
### Lack of Adoption for First-Gen AI Features
(18:40) co-pilots, particularly in the content space that you saw so many of the Martech SaaS vendors. I think they haven't gotten the adoption. It's almost at the point where we're now making fun of if you've got the little sparkles on something inside your, it's okay, that
### Disappointing Adoption of Early AI Features
(19:02) was stuck on. Not quite sure what my point was on that, but it was a little bit surprising that for all the effort that so many of the SaaS companies put into that first generation of AI features on their products, those generally haven't been the things that have actually gone adoption. Well, we've got two questions on the CMS
### Standouts in CMS
(19:29) front. And one is, is there anyone who stands out in the CMS field? The big brands are the ones you really know. The long-tail folks I don't have off the top of my head. But again, often, when we publish this, there'll be an interactive map. You can zoom in on them.
>> I often find the long tail again
### Long Tail CMS Innovators
(19:51) interesting. Not the odds of any one of those actually growing up to be the next Sitecore is low. But it's interesting to look at those who have nothing to lose, who are coming in with completely fresh eyes and no backwards compatibility. Even how they think about that space and those capabilities, because that becomes interesting patterns that
### CDP Adoption and Trends
(20:15) actually we might adopt even if we don't adopt that tool.
>> Gotcha. And then Earl is asking, is the trend you're seeing with CMS similar to what you're seeing in CDPs as well, since most B2B marketers are in early stages of their CDP adoption?
>> I think with the CDP side of it, what's happening right now that's
### The CDP Landscape
(20:39) fascinating is we had a whole bunch of people enter that space, obviously. In fact, almost every Martech company was, and among everything else, we're a CDP. But even among those companies that were legitimate CDPs, if you could use that adjective, it was hundreds and hundreds. But they span quite a range
### CDP Capabilities and Shifts
(21:04) of capabilities. Some were very down to the metal databases, and others, quite frankly, were more engagement platforms. Since over the past year, we've seen a couple of interesting things here, most the exits of CDP companies. So those that had scale that exited, they actually moved to the engagement layer. Which makes sense, in particular, with
### Data Warehouses vs. Pure Play CDPs
(21:30) this other thing that people are finding using these data warehouses, Snowflake or Google or Databricks or things like that, actually becomes the easier way for them to get a lot of the raw data. It doesn't solve the engagement problem, which is hence why they still need that. But this idea of a pure play CDP that doesn't have engagement,
### Importance of Data Layer and Attribution
(21:56) they're still out there, but that category seems to be fading pretty rapidly.
>> And that makes total sense, but I was curious because only because you mentioned the data layer in the beginning, and for anybody, as well as an expert, anybody who does the data, how you define your attributions and all the all the
### Engagement as a Focus
(22:14) measurements that you include influences how you integrate your platforms. So that's why I was curious to see if there's been any trends that you've been seeing outside of the CMS. But it seems to make sense that engagement would be the focus. Now you can track all those interactions. But still the data layer, probably the most important component to
### The Semantic Layer in CDPs
(22:34) identify, I would argue.
>> Well, I think one of the things at that data layer, that both the data clouds have been doing this, but also even some of the CDPs, I don't know, Hidoch, is a composable CDP. This concept of a semantic layer does become essential, because it's almost the dog that catches
### Governing Data Overload
(22:57) the car. It's our original problem was we can't get the data from across our. Then we get all the data from across our, we're nobody, we have too much and now we have no AI to make sense of it all.
>> That is actually, I would say, the area where the most active and
### Context Engineering
(23:14) interesting things are happening is, okay, now that we've got the data flowing, how do we govern it? How do we make sense of it? And that is still arguably where there is a CDP-ish role to be played, of listen, out of all that crazy sea of data, how do I package up the pieces of that that are relevant to particular experiences or campaigns?
### CDPs and Context Engineering
(23:36) In the AI world, people talk about this as context engineering, because prompt engineering is 2024. Oh my god, so context engineering. And in a lot of ways, I think that's what those CDPs were way ahead of their time, of, oh yeah, let's bundle up the context from a data perspective to do this particular execution.
>> We got a one from John. Any trends
### Vertical Market Focus
(23:59) around companies becoming more vertical market focused? And if so, what vertical markets are proving most popular? Are there any surprises?
>> That's interesting. We don't categorize it by verticals. So I don't have hard data on that. Everything I have is anecdotal. I will say I feel that narrative has had more strength, has been more popular
### Horizontal Platforms Preferred
(24:26) among the VCs, who are trying to find where can we actually put money now that has a chance to play out? And less of generally what I hear when I talk to marketers. Even if they're in a particular vertical, they still tend to largely be using horizontal platforms that they adopt, adapt to their needs. If anything, I think the twist is
### AI Changing Build vs. Buy
(24:52) now that AI is starting to change the build versus buy equation. Again, I'm still on the camp. I don't think you should build your own CRM. But this ability, because things are now opening up with stuff like MCP, is you're seeing more and more cases where companies have their commercial platform, Salesforce, whatever it is
### Custom Business Applications
(25:13) there. But then on top of that, they're, actually, we want to have our custom version of, okay, how are we going to manage a particular sales pipeline, or how do we do lean scoring on this stuff? And so in some ways, it feels that's leapfrogging a bit of the idea of prepackaged vertical market applications. What's
### Tailored "Business of One" Applications
(25:36) even more tailored than a vertical market application? It's business of one application. It would have been insane for most companies to do even a couple years ago.
>> Yep. It's now, I would still again caution, you can definitely get over your head, but that seems to be where the more likely direction is moving.
### Staying Updated on Dynamic Space
(26:02) And then Adam's wondering, oh, it's a great meta question here. What tools, publication systems do you use to stay on top of this whole dynamic space?
>> Didn't you miss my disclaimer at the beginning? I can't stay on top of it.
>> I don't know, I, honestly, it's on top of it. I follow a ton of people on
### Insightful Newsletters and VCs
(26:28) LinkedIn and X and all that, and take a very heterogeneous set of things that I see flowing through that stream. If there are newsletters I do regularly subscribe to, there's actually a subset of folks out there in the VC community that I find insightful on this. One of them is a guy named Jaman Bell here. I'll put it in the chat, who does Clouded Judgment
### Recommended Newsletters
(26:55) is a newsletter that's good. And then there's Tomas Tongas, who's now got his own firm, Theory Ventures. Those are a couple of the ones that pretty much every time I get one of their news, I'm, it's actually an insightful perspective.
>> And someone's asking if you save
### Saving and Analyzing Research with AI
(27:22) what you find into a platform, a notebook, LM, or something that.
>> Can I, can I expand on that?
>> Oh, go for it. So you said you read and follow all research online. I'm wondering if you're saving what you find, and then if you feed that somehow into some AI model so that we can
### Not Yet Using AI to Process Research
(27:46) extract insights from it.
>> That's a great question. The honest answer is no. I probably should. This is, again, I'll speak for myself, but I definitely feel on more than a few occasions the old dog new tricks. I've forcing myself to learn new things. I'm deep in Claude code.
### Overcoming "Old Dog, New Tricks" Syndrome
(28:10) I'm building some fun stuff there. But there's so many things that have changed, and there's so many things that I still have on autopilot, without having even stopped to think, oh, I should try an entirely different approach to this. And one of them is the way in which I consume writing out there.
### AI Governance Confidence and Ownership
(28:31) >> If you, I can show you a way to do that.
>> Okay. Let's follow up on that here. I'll
>> And then Peter was asking, Scott, your recent report surfaced only 8% of organizations feel confident in their AI governance, but adoption is accelerating. So what do you, who do you think should own the orchestration and governance layer? And to give credit where
### Clarifying AI Governance Study
(28:55) credit is due, that was citing a study that was done by SAS, who was one of the sponsors in that report.
>> So do you think it's high or low then based on, is there bias, is what you're saying?
>> No, no, no. I wanted to give them credit. I'm sure if they wanted to bias in a particular way, 8% is probably not a
### AI Governance Must Be Corporate-Wide
(29:22) good bias. I, to be honest, as much as I'd been an advocate over the years for marketers and marketing ops and Martech to control a lot of their destiny, there's a set of things right now that I think have to be corporate-wide. I think AI governance is one of those. Honestly, the data layer,
### CMO-CIO Relationship for Infrastructure
(29:48) marketers have to be responsible for their piece of the data layer, but it's some of these things where, as an organization, I've been at this for many a decade, there was a time when it was a barrier to what marketing needed to get done. But the world's moved on a lot. And I
### CIO Support Accelerates CMO Goals
(30:10) think in companies where there's a healthy relationship between the CMO and the CIO, there's so much infrastructure stuff that the CIO is able to provide to the CMO that accelerates what the CMO wants to do with their org. But anyways, I feel AI governance is one of those things, that's it's got to be
### Evolving AI Governance Best Practices
(30:33) corporate-wide. And it's a problem because again, you don't, it's such a new thing. Who has experience? How many, what's the best practices of that? These things are being written as we stumble through it.
>> A slight twist on that one over there. I was in a recent round table and the chief HR officer from a global bank was talking about their
### HR's Role in Agent Governance
(30:57) their new role in governance and onboarding agents as they are, as part of the personnel process. So agents go through the same type of training, evaluation and reviews and roles, and have a similar governance to the humans. I'm wondering about seeing that HR convergence with it, in that capacity.
### Opposing Reactions to HR Agent Governance
(31:23) >> I have two opposing reactions to that. One is that actually sounds smart. Of why you would want alignment on those things, and hey, it's a great opportunity for HR to reinvent itself in this next stage. On the other hand, the other reaction I had, which is a very visceral one, is I'm still in that camp where the degree to which certain leaders seem to treat
### Human-Agent Fungibility Concerns
(31:49) humans and agents as very fungible resources. It doesn't sit well with me. So I those metaphors still caused me to twitch a bit. But leaving that aside from an organizational perspective, and it sounds a pretty reasonable way to think about it.
>> And thanks, Peter. Mark's been waiting so
### Question on AI "Wrappers"
(32:17) patiently. Come on. Come on up.
>> Hey there, Scott. Had a question for you about wrappers. It seems the interface is confusing for a lot of marketers, especially solo practitioners and consultants. We've been talking a lot about enterprise. I'm wondering what are your thoughts on wrappers, and are they being used by marketers? And do you report on?
### Defining AI Wrappers
(32:42) >> When you say wrappers, you mean a product that's okay, I don't know, I'm trying to make up something here, but hey, I want something to help me build a campaign, rather than do that step by step myself in Claude. Oh, I've got something that's a little bit of a guided.
>> Wrappers are models that
### User-Friendly AI Wrappers
(33:00) are built for the interface is easier for the user. For instance, there's some for law firms, there's some for medical offices, there's some for consultants, etc. And they have APIs going to the various AI models. Okay. So, it's cheaper and you get access to 20 models, but you're restricted.
>> A couple ways I could answer that. One is
### Human and Organizational Limiters
(33:29) having observed this over decades, the rate of change in technology and the rate of change of organizations, man, that gap widens, and it's now almost insane. You almost can't see across the chasm. The limiter on all this stuff is the human and organizational side. And so while there are people who might look from a
### Value of AI Wrappers for Users
(33:54) technical level and say, hey, you don't need that wrapper, you could do it, I actually think those wrappers serve a great role if they're able to take a set of folks who aren't ready to dig into that, and this is a way that they can get value out of it and use it. Usually where the push back on wrappers is isn't the fact that they're
### Defensibility of Wrapper Businesses
(34:17) non-useful to users, because I think there's actually a lot of cases where they clearly are. It's that from being in the business, from being in the wrapper business, usually the big question is, okay, well, how defensible is that? And at what point in time do the Frontier Labs absorb that?
### SAS Opportunities in Context
(34:40) >> Like everything else.
>> Yes, like everything else. But actually, it's funny. I've been writing a lot about this, around all this stuff around context. And I think there's a lot of opportunity in the SaaS space for them to lean into the strength of what I think they've always had, which has been very, very good at framing the context of
### Domain Expertise for Context Delivery
(35:03) particular kinds of work and activity. I think one of the things that goes into being good at delivering context is you have to have that domain expertise. And again, I'm hesitant to say this, because Sam Altman will come out with something six months from now and
>> completely prove me wrong. But I think it's very hard for those frontier
### Frontier Labs vs. Domain Expertise
(35:29) labs to move in the direction of developing the domain expertise and domain-specific interfaces and things. And I don't see that's to their advantage to do that. They're in such a position to win at the horizontal layer below that. So I still think there's a lot of value out there for that. But
>> Do you report on them?
### Tracking AI Wrappers
(35:49) >> What
>> Do you report on wrappers? There's so many of them out there. I
>> I don't actually have even a way to
>> Right.
>> It's a weird continuum. What's a wrapper, right?
>> So no, we don't track that
### AI Native Companies as Wrappers
(36:02) specifically, but
>> Okay. You could say a lot of the AI native companies that have been born in these past three years, hand wavy back, you could say the vast majority of those are a kind of wrapper. Or actually, see, the thing I like about this is language is such a wonderful thing. The way in which people are now starting
### Wrappers vs. Harnesses
(36:24) to talk about this, oh, is this not a wrapper, it's a harness. The whole thing around Claude code. Oh, no, no, this is a harness for it. And again, to some degree, I actually think that's right. And there is proprietary insight, which apparently Anthropic just accidentally leaked to the whole world, but there is proprietary IP and how do you structure
### Next Questioner
(36:46) a harness to do a particular task well, even if they're all sharing the same underlying LLM.
>> All right, who else has got something for Scott? These are fun.
>> Are you daring me? Because I will. Okay, Earl, you could take the rat sock is dangerous.
>> No, no, no, not at all. Not at all. Everything you're saying, Scott, I was
### AI-Only Solutions and Adoption
(37:06) curious about all these new companies that are coming out with their AI only solutions. And I know that especially in our crowd, we're very interested in it, especially to see where it's going. But I keep going back to what I learned about product and product development, and I'm not seeing the adoption that you get from the civilian side or the technical side in
### Crossing the Chasm with AI Tools
(37:30) terms of all these different tools. For example, David will talk about Vibe coding all day. I can talk to you about custom GPTs all day. But we are very selective crowd, and I'm wondering, think in terms of crossing the chasm. How many people are actually using these in their companies? Companies are pushing them and investors are pushing for adoption. But how many
### Ownership of AI Adoption in Enterprise
(37:50) actual people in their, especially enterprise organizations, are actually pioneering these things and pushing these things forward? Back to the question earlier, who is going to own the AI adoption process? That's the thing I'm struggling with. Who do you talk to when you want to talk about AI to help them with their AI?
>> Okay. So there's a few things.
### Two Questions on AI Adoption
(38:11) Two different questions. What do we think about the civilian adoption? And two, who do we think should be owning the AI adoption process internally at enterprise organizations? Two separate questions. For the first one, there's a place where we have data, and then there's a place where at the moment all I have is anecdote. The anecdote is what I
### Lack of Adoption for New AI Products
(38:32) hear anecdotally is most of these AI features, and then also a lot of these new AI products, they're not getting adoption, for a variety of reasons. Again, I think even the weekly active users on
### Breakthrough AI Products
(39:17) things like ChatGPT or Claude, when you see breakthrough products like Lovable, And so I think it's interesting, for the most part, people don't want to have to learn new stuff if they don't have to. But out of the thousands of attempts that are throwing things against the wall right now, there
### Bifurcation of AI Product Success
(39:43) do still tend to be several dozen or so that, oh no, this catches on and people realize, oh wow, I can do this and this is great, and then they tell two friends and they tell two friends and so on and so on. So anyways, it's a bifurcation. I think it is
### Who Owns AI Adoption?
(40:04) possible for there to be great takeoff, but only for 0.2% of what's in the market right now.
>> Gotcha.
>> What was the second half of that question?
>> Yes. Second half was, who do you see, at least from your research and your anecdotal stories that you've had, who has been the owners of the AI adoption?
### AI Adoption Ownership Roles
(40:28) Because, obviously, it depends on the size of the company, whether it's SMB or enterprise, I get that. But as you said, it can be anybody from the CMO to the CIO, maybe even if they have a CPO, chief product officer, that might be a person involved, who knows? But what have you seen from your research?
>> So I see three roles.
### CEO and CIO Roles in AI Adoption
(40:50) This is CEO, who basically tells everybody they need to use AI. That's about as helpful as it gets.
>> You saw what happened with a couple companies about that. Better to rehire them back.
>> It's a wacky time. The second role is, it is generally the CIO or the IT organization that's going ahead and getting the enterprise licenses and
### Individual Teams Drive Genuine Adoption
(41:11) manage them for the Frontier Labs and some of these major platforms. However, neither one of those things actually speaks to genuine adoption, much less actual genuine impact and outcome.
>> Everywhere I see it, I'm trying to think of other exceptions. It happens so much down in the individual teams.
### Zapier CEO's AI Framework
(41:37) There's a few companies that have been forward about this, oh, what's his name? The CEO of Zapier. The whole Zapier company has been obviously core, their product has been very much on the frontier of this. But he's been publishing his, okay, and this is the framework we use of what we're expecting
### Lack of Top-Down AI Guidance
(42:00) from people. And it's not at that level of the CEO saying use AI. It's, no, no, actually these are the different kinds of things in the use cases, and how do we measure it? But he almost stands out because that is such an exception now that in most companies, there isn't enough of that guidance top down to say that anyone is taking ownership
### CRO Interest in AI Adoption
(42:20) of adoption.
>> That makes total sense. I think that companies that have a C-level executive for revenue, for example, a CSO, a CRO, would be interested in that, considering that they can make a big deal from their data, from they're getting that they're getting from their sales teams. But I don't see it in the market.
### Sales Team Tool Adoption
(42:41) >> Learning new tools is not a sales team. It's kind of big, acquire me from my experience.
>> And they just gotten excited about Gong. Okay, we think we finally got our arms around this.
>> Baby steps. Got woo. I'll take it.
>> Oh, and then Clay, everyone's, of course, we go to market, we've got Clay as if that's our
### Brand Rebranding Surprise
(43:01) magical thing. But
>> I think they rebranded, they rebranded the consumer product to Mesh, I think recently.
>> They rebranded, I believe.
>> Seriously? Look it up.
>> Emsh. They rebranded the consumer product.
>> Hang on. I have to Google this.
### Difficulty Keeping Up with AI Changes
(43:22) >> You Google me your own.
>> Oh my god.
>> I would have sworn that was an April Fool's joke.
>> No, but okay. All right. I'll have to, all right, I once again go back to my disclaimer at the beginning of this thing. I can't keep up with all of this.
>> Nobody can. Not even AI can. Meanwhile,
### Agentic Commerce Inquiry
(43:41) Yogish, you want to chime in?
>> Hey Scott, good to see you again. I had a question that's a slightly different one. In your, I know you were sharing a little bit of tidbits from your upcoming release of your new report for the year. I was curious if you were seeing anything around agentic commerce showing up this year in your analysis.
### Lack of Agentic Commerce Adoption
(44:04) >> The short answer is not a lot. We actually, and this is why anybody who says they know the future in this market, we were actually expecting to see more of that, because at the end of last year, all the big announcements from OpenAI and then Google, and everyone's headed into that, and then it's generally turned out for the
### Consumer Readiness for Agentic Commerce
(44:29) most part, consumers aren't ready for this in a lot of the cases. But part of this depends on how you define agentic commerce too. Because there's these things, I call them the shopper concierges. There are these AI experiences, whether it's a dedicated app or something like this, and you could, depending on how loosely you want to
### Science Fair Stage of Agentic Commerce
(44:54) define agentic, you can see that. But actually having agents go and do these things for me, it's still everything we're seeing right now, it's still science fair.
>> Got it. A quick follow up on that. Based on what you're seeing, at least from the larger LLMs, OpenAI and Google's Gemini, I'm curious to hear
### AI's Impact on Consumer Discovery
(45:17) what your thoughts are on what you're seeing in the space.
>> I think what OpenAI discovered in there is there is this massive shift that has been happening of consumers starting to truly use AI for discovery, and also for evaluation and weighing different options and stuff that, which, to be honest, again, in two years, the
### Industry's Slow Absorption of AI Shift
(45:46) whole nature of how people do discovery and evaluation online is shifted. And to be honest, that hasn't fully, I think the industry, we collectively haven't even fully absorbed that and learned how to deal with that well. And I think when OpenAI was retrenching away from this, they're, okay, this is we're going
### Merchant Resistance to Ceding Discovery
(46:12) to focus on for the core business. I think that's where the balance seems to be. That's obviously also one where the merchants involved are willing to cede that, because they'd already had to cede that once before. Google, not happy having to relearn this all again, but open to the possibility that they will not fully control the discovery channel, unless if
### Walmart's Stance on AI Discovery
(46:35) you're Amazon, they won't control it. But what was it, the Walmart head of AI had that quote at an investor conference when they were, the OpenAI thing was letting ChatGPT do that, that was a temporary moment in time. Walmart's, the hell we are going to cede over, and it become your fulfillment
### Consumer Trust and Merchant Readiness
(47:01) service on the back end here, my friend. So it's both the lack of consumer trust, plus the fact that I think the merchant community was maybe caught a little bit off guard when this first hit. All the ones I talked to, they're, we're not going to, we're not going to walk into that if can help it.
>> I get that. And one thing that I find fascinating in
### Merchants Creating Walled Gardens
(47:26) this whole space right now is that the speed at which some of these merchants are investing into their own, call it walled gardens, in a way, is going to create an interesting mix for brands in terms of how they'll be able to engage across different protocols, right? And I wonder if that's going to open up a new space for startups to be able to offer for potentially new
### Retail Media Networks and Fragmentation
(47:50) solutions. I don't know. I'm speculating here a bit, but it feels that that's where things are going. So,
>> Well, in some ways you could say this is what's happening with AntTech. This explosion of these little retail media networks is actually now once again we've got a fragmentation
### New Opportunities from Market Fragmentation
(48:09) in a market. And so that creates, we were headed towards what the duopoly, and it still largely is a duopoly. But now is enough interesting things happening in this fragmented space that you're starting to see the emergence of software vendors who are great, we can help in that environment. So,
### Upcoming MarTech Report
(48:33) Scott, any other things that you're excited about coming out next next few months? Anything else we should all be paying attention to?
>> We'll have that State of MarTech report out at the beginning of May. So, we now distribute that free and ungated. So, whenever that's ready. But, no, thanks for inviting me to have this chat with you. These are great questions. I love this conversation.
### David's Gratitude and Future Guests
(49:15) Welcome by anytime. We'll make sure to share your latest state of things report with the community. And for everyone, we've got. I almost get embarrassed sometimes when I'm now working on booking things, and I'm, well, you're an amazing guest. Let's look at July. So it's a fun fun problem to have, but it means that we
### Upcoming Events and Holidays
(49:39) got a lot of conversations coming, including with at least a couple of folks on this call today. So appreciate you all coming by. Stay tuned for more. Check the Luma for a lot of what we have scheduled, and a few more things we probably need to add. And anyone in New York next week, we're going to be doing a belated first Wednesday, because tonight is the first Wednesday,
### Holiday Greetings
(49:58) but also the first night of Passover for a lot of folks celebrating here. So happy Passover and happy Easter to everyone celebrating the next few days. And we will see you all very soon.
(0:05) Hey everyone, I'm David Berkowitz and welcome to another edition of AI Insiders from Architectures AI Marketers Guild. It is a true pleasure to introduce an old friend, Scott Brinker, Mr. Chief Martech himself and someone who's been trying to make sense of so many of the worlds that we're in for so long. And I'd even say Scott that you're usually more
### Staying Ahead in a Confusing World
(0:32) successful than most. Even if I wind up in a haze of confusion every damn day that I don't think is going away anytime soon. So can't blame you for that, can I?
>> Well, why not?
>> Why don't you share? Because I feel most people on this call know you, have at least read some of your stuff, know some of your incredible little
### Scott Brinker's Current Role
(0:57) landscapes and all that, but why don't you at least share a little bit about what you're up to today?
>> Sure. For about eight years, the previous eight years, I was actually building the Martech ecosystem for HubSpot. Which was great fun. But I left that back in September. And this work I've been doing with Chief Martech,
### Challenges of Tracking MarTech Trends
(1:17) an armchair analyst to this industry for 20 years. That always has been a side hustle, labor of love, call it what you will. Decided to actually go full-time with that. Which to your point, David, my goodness, even trying to keep track of all this stuff 24/7, I feel hopelessly behind every day in these announcements. Trying
### Current AI Interests and Inquiries
(1:42) to chip away at it a bit at a time. Doing Martech analyst and advisory work full-time.
>> I mean, obviously, I've seen some of your work. We're all in the AI space, market or not, but what areas specifically are you most interested in right now? Are people calling you most often for?
### New Data Layer Report
(2:08) Not to be promotional on this, but I will share there's a report
>> It'll be a little promotional. We published a week ago with the folks at Databricks. And although this was something sponsored by Databricks, the way the report was researched and written, it applies to any of these data cloud Snowflake, Google BigQuery, whatnot.
### Focus on the Data Layer
(2:34) I would say the two things that I've been most interested in is one, what is actually happening at this data layer? Because for all the excitement around AI, you all know it's so much a function of the data that you feed into it. And for the most part, the data layer of most not Martech stacks, but company tech stacks in general, is still, how do I say diplomatically,
### Creative Possibilities with Better Data
(3:01) immature. There's a lot of opportunity for that to get better. And I think as it does start to get better, it opens up a lot of creative possibilities, not with AI, but all the things we can do with that data and AI. That's one thing I've been focused on a lot. And then the second one is
### Rise of AI Agents
(3:23) clearly this is the year of AI agents. You can't go a LinkedIn post without seeing AI agents. There are many kinds of agents, right? There's the agents we're using behind the scenes in marketing. There's agents we as marketers are deploying that are customer facing, everything from customer service chatbots to shopper concierges to, I'm sure we all
### Buyer-Controlled AI Agents
(3:48) love AI SDRs. But the agent that is actually the most intriguing to me is the category of agents that are not marketer controlled. That they're buyer controlled. We've seen this beginning with the shift from SEO to AEO, and buyers increasingly leveraging ChatGPT and Bard and Gemini and all this to take a very different
### Evolving Marketer-Consumer Engagement
(4:15) control over their journey. But I think that's one example of multiple kinds of AI agents that are starting to pop up. We're starting to see some hints of this in the email space, AI control of the inbox that again, it's changing the relationship or the channels and the mechanisms by which marketers and consumers, customers
### Humans vs. Bots
(4:41) engage with each other. And that's very nascent, but it's probably the area I'm most fascinated by.
>> Well, it touches on something that I keep going back to and everyone I work with, they have to hear a little bit of my soapbox on this, and that's especially now, in this increasingly agentic era, that there's
### The Two Audiences: Humans and Bots
(5:07) that we basically have two audiences and it doesn't matter if you're B2B or B2C. It doesn't matter if you're talking to an individual consumer versus a Fortune 500 CEO, that you've got humans and you've got bots. And the humans all have more in common with each other. I have way more in common with Tim Cook or Oprah or some person who's running some startup out of Lagos
### Bots as an Alien Species
(5:36) now. We all have some shared characteristics and have way more in common with that Lagos founder who I've never met than I do with bots, right? And the bots, whether they're coming from Claude or they're coming from Google or they're coming from G2 or any other company, they all behave in a more similar way as this kind of alien species. And so it's
### Acknowledging Different Targets
(6:09) if you start acknowledging that you've got these two very different targets, then I feel a lot of what you have to do unfolds from there. Does that gel with what you're seeing or you have any holes to poke in that?
>> Well, to your point, and you know this better than anyone, marketers have always been doing
### Marketers Adapting to Google Bot
(6:32) this now for one, two plus decades, where there was our human audience and then there was the Google bot.
>> Now granted, it was the Google bot, but obviously the whole SEO industry, we put a lot of effort into actually mastering how do we talk to both of those audiences simultaneously.
### The Challenge of Diverse Bots
(6:52) I think what's perhaps both interesting, challenging at the moment, is it is no longer one bot. It is this increasingly diverse set of bots.
>> Although to me, there still is one notable difference is that with the Google bot, when the idea was that a human would land on that page that surfaced from Google's list of links,
### AI Reinterpreting Information
(7:18) it still had to be readable, right? An SEO expert who wanted to overoptimize keyword stuffing and all this stuff, if they made it unreadable, then the conversion couldn't happen. And now when most of that info is being reinterpreted and synthesized by the LLMs and by other AI tools, then we don't even know what version of this the human's
### Inconsistent AI Models
(7:44) actually going to see later.
>> That's a fair point. And again, there's a lot of diversity Gemini does that is different than ChatGPT is different than Claude. And worse, they're not even consistent within themselves. New model anytime. They're constantly evolving. And that combined with the lack of visibility,
### Early Stages of AI Agents
(8:07) it does make it quite a game. I don't know. I keep thinking about that movie Dodgeball. He puts the blindfold on, and they're like, "God, and it's going to be hard for him to see." It just that's a little bit what it feels at the moment. But to be honest, I still feel that's very early steps. What is
### Buyer-Side Agents and MCP
(8:28) intriguing is the notion that we are going to see agents on the buyer side that actually doing more than reading content, and synthesizing from that. But to the degree that we're able to expose things that they're able to do through companies that are exposing some of their things through MCP in the B2B space, are already seeing this
### Practical Examples of Agents
(8:52) thing where people have their work agents and they're, "Oh, can you fix this scheduling thing for me?" They go off and do it. And it's some of these things that are, I think, relatively new channels.
>> But are you, do you have any either B2B or B2C? Do you have any practical examples that you've used, you've seen, you've worked with other companies that
### MCP for Newsletter Management
(9:14) are doing in some way?
>> Most of the stuff that I've been doing with MCP has been back on the orchestration with things across your stack. For instance, I use Beehive for delivering my newsletter, and they released their MCP server. So, I'm in Claudius. I'm thinking about, hey, can you check on this thing? What
### New Interaction Channels for B2B
(9:36) happened with this audience? How's this open rate been changing? Oh, I had this sponsor here. I had this go. The fact that it can basically go behind the scenes, get that information from Beehive, this is a new channel in which Beehive interacts with me that's not through their app. We see a lot of those examples today on the B2B side,
### Consumer-Side Agent Emergence
(9:58) but I don't see why that can't start to emerge on the actual consumer side. Obviously, what Chetch tried to do here with instant checkout that they've pulled back on, but even then, what Google's trying to do with UCP, it all feels very science fair project stage right now, but it does seem to be pointing directionally into this thing of people will be
### Google Search Console and Base 44 Super Agent
(10:21) leaning on these agents to actually do things, not synthesize content.
>> That's what you're saying with the Beehive MCP, it brings to mind something that I've found especially useful lately because one of these odd things, as I've been building more, I've been learning or relearning Google Search Console because now I have to go and make
### Base 44 Super Agent Functionality
(10:46) sure that these sites are potentially visible. And I use B 44 for so much of my building. Base 44 past couple weeks released a super agent tool to work across your sites. And one of the things it'll do is connect to Google Search Console directly. It can then set up error monitoring. It can fix a lot of the errors itself and then tell you what to do with your
### Powerful Code Rewriting
(11:13) individual properties and then it'll often rewrite code for me to better accommodate what Google Search Console is doing. So that I can then go from where I'm actually building this stuff and no longer look at Google Search Console itself. This stuff feels pretty powerful and useful.
>> And when you think about it,
### Rapid Advancement of MCP
(11:40) it's been a matter of months. Depending on some of the ones that were really forward-leaning, what about 14 months since the entire notion of an MCP protocol was introduced. The speed at which this stuff is moving and advancing is pretty wild.
>> I got an email literally an hour ago saying, "Your
### The "Lake Wobegon" Effect in AI
(12:05) super agent learned 130 new tricks with all of its new integrations." So, I'm about 130 of those behind by the time I got on this call, right? I've been thinking about this lately. The Lake Wobegon effect, all the children are above average. I feel we collectively now are in the inverted Lake Wobegon effect. Everyone feels below average and trying to keep up with this stuff.
### Credible AI Startups
(12:30) So maybe that's me.
>> Do you have some because there's so, I could literally launch an AI startup that might not actually be very good but looks credible on the surface by pulling some stuff together with these code and tools right now, right? And I can give it
### Distinguishing Credible AI Businesses
(12:52) a nice domain and logo and all this stuff and it can look a very credible business. But it might not actually do anything. Do you even have some threshold for what you'll pay attention to, or will you look as far out on the fringes as possible? How do you draw some line in your world?
>> We have generally gone
### Validating Small AI Companies
(13:18) pretty close to the edge. We try and validate that things actually is a legitimate company. There's signals you can have the ability to contact it. What's their presence on LinkedIn? You can cross validate this stuff. But we look at a lot of companies that are very small because I feel there's lots of cases where you get these great
### Value of Curated Landscapes
(13:40) curated landscapes of hey, these are the top 20 products you want to pay attention to, and I actually find those things very useful because they are what we typically think of as the head of the tail or the main things. But almost by accident, I've ended up in this mode with the MarTech landscape, really looking far down the
### Analyzing the Long Tail of MarTech
(14:02) long, long tail, which trying to look at individual companies on that is not very useful to anyone. But actually looking at it in aggregate and how it evolves over aggregate and in which categories and which ones stick around and how quickly do they churn, and what's the percentage rate that grows, it continue. We're in the middle of working on our big
### Preview of MarTech 2026 Report
(14:24) State of MarTech 2026 report now, and it's full of insights of, okay, for individual companies, not super interesting. Patterns in aggregate, very interesting to start to see what these dynamics are showing.
>> Are there any patterns you give us a preview of?
>> One of the things that's actually very interesting is this. Don't spread this one on
### CMS and E-commerce Takeoff
(14:49) LinkedIn, we'll be between us, but one of them is there's quite a takeoff in the CMS and the e-commerce space. E-commerce had a nice wave around the pandemic era, just because of the big shifts, but it's starting to settle down. CMS has largely been a pretty static category for quite some. You could argue
### Category Acceleration and Churn
(15:14) it's actually the oldest Martech category out there. But we saw significant acceleration in both of those categories, and almost double acceleration, because not only did they grow, but they actually had a lot of churn. A lot of older companies have basically left the space or the sector, exited, acquired, or caught on fire,
### Growth Driven by Rethinking CMS
(15:37) whatever happened. So actually, for there to be growth, there had to be that much more to cover up the deficit from those who exited. And when you start to dig into that, and you look at, it makes sense that it's some of the things we're talking about on the CMS side. People are rethinking, okay, how do we manage
### Generative AI and AI Agents in CMS
(15:58) the website in this environment where both, hey, we can leverage generative AI as part of what we're doing here, but also, oh, we've got these AI agents of various kinds and flavors are engaging with us. How do we think about serving that audience? How is that built into this? That was interesting. That was not on my bingo card to see that clear and crisp of
### Content Marketing Category Decline
(16:23) a renewal of the CMS category.
>> Okay. Interesting. Is there anything that was top of mind the past few years that is not as prominent right now? What's on its way out?
>> Well, the category that suffered the most this past year was actually the content marketing category, which part of it
### LLMs and Content Creation
(16:57) is because it had a, it was the category that actually had some of the fastest takeoff around 2023, just because as soon as LLMs were out here, people were, oh my goodness, we could wrap this and do all sorts of content things with it. And a whole bunch of people did. But playing out a few years down the road, there's a big difference between
### Sustainability of AI Wrapper Businesses
(17:21) wrapping an LLM, and as you're saying, hey, I can wrap this. I can put it out a website and it looks credible, okay, this is actually a sustainable business. Even when you're talking about small businesses, still is it a sustainable business? Is there a competitive offering there? And probably not a surprise, but saw
### Survey on AI Use Cases
(17:43) a pretty big exit from that category. The other thing I would say is this isn't about the landscape, but in parallel to the landscape, we ran a pretty in-depth survey of 70 marketing use cases with 28 marketing ops Martech leaders. And what we were primarily asking about was, okay, for this use case, are you using AI within an existing SaaS platform? Are you
### AI Use Case Insights
(18:13) using a new AI native tool for this capability? Have you created something of your own with AI? Or no, we're not using AI, or we're not doing this use case. And that's, and of course, then we split out the data between B2B and B2C. And that's proving very interesting. But one of the things that came out was, boy, a rush of those
### Lack of Adoption for First-Gen AI Features
(18:40) co-pilots, particularly in the content space that you saw so many of the Martech SaaS vendors. I think they haven't gotten the adoption. It's almost at the point where we're now making fun of if you've got the little sparkles on something inside your, it's okay, that
### Disappointing Adoption of Early AI Features
(19:02) was stuck on. Not quite sure what my point was on that, but it was a little bit surprising that for all the effort that so many of the SaaS companies put into that first generation of AI features on their products, those generally haven't been the things that have actually gone adoption. Well, we've got two questions on the CMS
### Standouts in CMS
(19:29) front. And one is, is there anyone who stands out in the CMS field? The big brands are the ones you really know. The long-tail folks I don't have off the top of my head. But again, often, when we publish this, there'll be an interactive map. You can zoom in on them.
>> I often find the long tail again
### Long Tail CMS Innovators
(19:51) interesting. Not the odds of any one of those actually growing up to be the next Sitecore is low. But it's interesting to look at those who have nothing to lose, who are coming in with completely fresh eyes and no backwards compatibility. Even how they think about that space and those capabilities, because that becomes interesting patterns that
### CDP Adoption and Trends
(20:15) actually we might adopt even if we don't adopt that tool.
>> Gotcha. And then Earl is asking, is the trend you're seeing with CMS similar to what you're seeing in CDPs as well, since most B2B marketers are in early stages of their CDP adoption?
>> I think with the CDP side of it, what's happening right now that's
### The CDP Landscape
(20:39) fascinating is we had a whole bunch of people enter that space, obviously. In fact, almost every Martech company was, and among everything else, we're a CDP. But even among those companies that were legitimate CDPs, if you could use that adjective, it was hundreds and hundreds. But they span quite a range
### CDP Capabilities and Shifts
(21:04) of capabilities. Some were very down to the metal databases, and others, quite frankly, were more engagement platforms. Since over the past year, we've seen a couple of interesting things here, most the exits of CDP companies. So those that had scale that exited, they actually moved to the engagement layer. Which makes sense, in particular, with
### Data Warehouses vs. Pure Play CDPs
(21:30) this other thing that people are finding using these data warehouses, Snowflake or Google or Databricks or things like that, actually becomes the easier way for them to get a lot of the raw data. It doesn't solve the engagement problem, which is hence why they still need that. But this idea of a pure play CDP that doesn't have engagement,
### Importance of Data Layer and Attribution
(21:56) they're still out there, but that category seems to be fading pretty rapidly.
>> And that makes total sense, but I was curious because only because you mentioned the data layer in the beginning, and for anybody, as well as an expert, anybody who does the data, how you define your attributions and all the all the
### Engagement as a Focus
(22:14) measurements that you include influences how you integrate your platforms. So that's why I was curious to see if there's been any trends that you've been seeing outside of the CMS. But it seems to make sense that engagement would be the focus. Now you can track all those interactions. But still the data layer, probably the most important component to
### The Semantic Layer in CDPs
(22:34) identify, I would argue.
>> Well, I think one of the things at that data layer, that both the data clouds have been doing this, but also even some of the CDPs, I don't know, Hidoch, is a composable CDP. This concept of a semantic layer does become essential, because it's almost the dog that catches
### Governing Data Overload
(22:57) the car. It's our original problem was we can't get the data from across our. Then we get all the data from across our, we're nobody, we have too much and now we have no AI to make sense of it all.
>> That is actually, I would say, the area where the most active and
### Context Engineering
(23:14) interesting things are happening is, okay, now that we've got the data flowing, how do we govern it? How do we make sense of it? And that is still arguably where there is a CDP-ish role to be played, of listen, out of all that crazy sea of data, how do I package up the pieces of that that are relevant to particular experiences or campaigns?
### CDPs and Context Engineering
(23:36) In the AI world, people talk about this as context engineering, because prompt engineering is 2024. Oh my god, so context engineering. And in a lot of ways, I think that's what those CDPs were way ahead of their time, of, oh yeah, let's bundle up the context from a data perspective to do this particular execution.
>> We got a one from John. Any trends
### Vertical Market Focus
(23:59) around companies becoming more vertical market focused? And if so, what vertical markets are proving most popular? Are there any surprises?
>> That's interesting. We don't categorize it by verticals. So I don't have hard data on that. Everything I have is anecdotal. I will say I feel that narrative has had more strength, has been more popular
### Horizontal Platforms Preferred
(24:26) among the VCs, who are trying to find where can we actually put money now that has a chance to play out? And less of generally what I hear when I talk to marketers. Even if they're in a particular vertical, they still tend to largely be using horizontal platforms that they adopt, adapt to their needs. If anything, I think the twist is
### AI Changing Build vs. Buy
(24:52) now that AI is starting to change the build versus buy equation. Again, I'm still on the camp. I don't think you should build your own CRM. But this ability, because things are now opening up with stuff like MCP, is you're seeing more and more cases where companies have their commercial platform, Salesforce, whatever it is
### Custom Business Applications
(25:13) there. But then on top of that, they're, actually, we want to have our custom version of, okay, how are we going to manage a particular sales pipeline, or how do we do lean scoring on this stuff? And so in some ways, it feels that's leapfrogging a bit of the idea of prepackaged vertical market applications. What's
### Tailored "Business of One" Applications
(25:36) even more tailored than a vertical market application? It's business of one application. It would have been insane for most companies to do even a couple years ago.
>> Yep. It's now, I would still again caution, you can definitely get over your head, but that seems to be where the more likely direction is moving.
### Staying Updated on Dynamic Space
(26:02) And then Adam's wondering, oh, it's a great meta question here. What tools, publication systems do you use to stay on top of this whole dynamic space?
>> Didn't you miss my disclaimer at the beginning? I can't stay on top of it.
>> I don't know, I, honestly, it's on top of it. I follow a ton of people on
### Insightful Newsletters and VCs
(26:28) LinkedIn and X and all that, and take a very heterogeneous set of things that I see flowing through that stream. If there are newsletters I do regularly subscribe to, there's actually a subset of folks out there in the VC community that I find insightful on this. One of them is a guy named Jaman Bell here. I'll put it in the chat, who does Clouded Judgment
### Recommended Newsletters
(26:55) is a newsletter that's good. And then there's Tomas Tongas, who's now got his own firm, Theory Ventures. Those are a couple of the ones that pretty much every time I get one of their news, I'm, it's actually an insightful perspective.
>> And someone's asking if you save
### Saving and Analyzing Research with AI
(27:22) what you find into a platform, a notebook, LM, or something that.
>> Can I, can I expand on that?
>> Oh, go for it. So you said you read and follow all research online. I'm wondering if you're saving what you find, and then if you feed that somehow into some AI model so that we can
### Not Yet Using AI to Process Research
(27:46) extract insights from it.
>> That's a great question. The honest answer is no. I probably should. This is, again, I'll speak for myself, but I definitely feel on more than a few occasions the old dog new tricks. I've forcing myself to learn new things. I'm deep in Claude code.
### Overcoming "Old Dog, New Tricks" Syndrome
(28:10) I'm building some fun stuff there. But there's so many things that have changed, and there's so many things that I still have on autopilot, without having even stopped to think, oh, I should try an entirely different approach to this. And one of them is the way in which I consume writing out there.
### AI Governance Confidence and Ownership
(28:31) >> If you, I can show you a way to do that.
>> Okay. Let's follow up on that here. I'll
>> And then Peter was asking, Scott, your recent report surfaced only 8% of organizations feel confident in their AI governance, but adoption is accelerating. So what do you, who do you think should own the orchestration and governance layer? And to give credit where
### Clarifying AI Governance Study
(28:55) credit is due, that was citing a study that was done by SAS, who was one of the sponsors in that report.
>> So do you think it's high or low then based on, is there bias, is what you're saying?
>> No, no, no. I wanted to give them credit. I'm sure if they wanted to bias in a particular way, 8% is probably not a
### AI Governance Must Be Corporate-Wide
(29:22) good bias. I, to be honest, as much as I'd been an advocate over the years for marketers and marketing ops and Martech to control a lot of their destiny, there's a set of things right now that I think have to be corporate-wide. I think AI governance is one of those. Honestly, the data layer,
### CMO-CIO Relationship for Infrastructure
(29:48) marketers have to be responsible for their piece of the data layer, but it's some of these things where, as an organization, I've been at this for many a decade, there was a time when it was a barrier to what marketing needed to get done. But the world's moved on a lot. And I
### CIO Support Accelerates CMO Goals
(30:10) think in companies where there's a healthy relationship between the CMO and the CIO, there's so much infrastructure stuff that the CIO is able to provide to the CMO that accelerates what the CMO wants to do with their org. But anyways, I feel AI governance is one of those things, that's it's got to be
### Evolving AI Governance Best Practices
(30:33) corporate-wide. And it's a problem because again, you don't, it's such a new thing. Who has experience? How many, what's the best practices of that? These things are being written as we stumble through it.
>> A slight twist on that one over there. I was in a recent round table and the chief HR officer from a global bank was talking about their
### HR's Role in Agent Governance
(30:57) their new role in governance and onboarding agents as they are, as part of the personnel process. So agents go through the same type of training, evaluation and reviews and roles, and have a similar governance to the humans. I'm wondering about seeing that HR convergence with it, in that capacity.
### Opposing Reactions to HR Agent Governance
(31:23) >> I have two opposing reactions to that. One is that actually sounds smart. Of why you would want alignment on those things, and hey, it's a great opportunity for HR to reinvent itself in this next stage. On the other hand, the other reaction I had, which is a very visceral one, is I'm still in that camp where the degree to which certain leaders seem to treat
### Human-Agent Fungibility Concerns
(31:49) humans and agents as very fungible resources. It doesn't sit well with me. So I those metaphors still caused me to twitch a bit. But leaving that aside from an organizational perspective, and it sounds a pretty reasonable way to think about it.
>> And thanks, Peter. Mark's been waiting so
### Question on AI "Wrappers"
(32:17) patiently. Come on. Come on up.
>> Hey there, Scott. Had a question for you about wrappers. It seems the interface is confusing for a lot of marketers, especially solo practitioners and consultants. We've been talking a lot about enterprise. I'm wondering what are your thoughts on wrappers, and are they being used by marketers? And do you report on?
### Defining AI Wrappers
(32:42) >> When you say wrappers, you mean a product that's okay, I don't know, I'm trying to make up something here, but hey, I want something to help me build a campaign, rather than do that step by step myself in Claude. Oh, I've got something that's a little bit of a guided.
>> Wrappers are models that
### User-Friendly AI Wrappers
(33:00) are built for the interface is easier for the user. For instance, there's some for law firms, there's some for medical offices, there's some for consultants, etc. And they have APIs going to the various AI models. Okay. So, it's cheaper and you get access to 20 models, but you're restricted.
>> A couple ways I could answer that. One is
### Human and Organizational Limiters
(33:29) having observed this over decades, the rate of change in technology and the rate of change of organizations, man, that gap widens, and it's now almost insane. You almost can't see across the chasm. The limiter on all this stuff is the human and organizational side. And so while there are people who might look from a
### Value of AI Wrappers for Users
(33:54) technical level and say, hey, you don't need that wrapper, you could do it, I actually think those wrappers serve a great role if they're able to take a set of folks who aren't ready to dig into that, and this is a way that they can get value out of it and use it. Usually where the push back on wrappers is isn't the fact that they're
### Defensibility of Wrapper Businesses
(34:17) non-useful to users, because I think there's actually a lot of cases where they clearly are. It's that from being in the business, from being in the wrapper business, usually the big question is, okay, well, how defensible is that? And at what point in time do the Frontier Labs absorb that?
### SAS Opportunities in Context
(34:40) >> Like everything else.
>> Yes, like everything else. But actually, it's funny. I've been writing a lot about this, around all this stuff around context. And I think there's a lot of opportunity in the SaaS space for them to lean into the strength of what I think they've always had, which has been very, very good at framing the context of
### Domain Expertise for Context Delivery
(35:03) particular kinds of work and activity. I think one of the things that goes into being good at delivering context is you have to have that domain expertise. And again, I'm hesitant to say this, because Sam Altman will come out with something six months from now and
>> completely prove me wrong. But I think it's very hard for those frontier
### Frontier Labs vs. Domain Expertise
(35:29) labs to move in the direction of developing the domain expertise and domain-specific interfaces and things. And I don't see that's to their advantage to do that. They're in such a position to win at the horizontal layer below that. So I still think there's a lot of value out there for that. But
>> Do you report on them?
### Tracking AI Wrappers
(35:49) >> What
>> Do you report on wrappers? There's so many of them out there. I
>> I don't actually have even a way to
>> Right.
>> It's a weird continuum. What's a wrapper, right?
>> So no, we don't track that
### AI Native Companies as Wrappers
(36:02) specifically, but
>> Okay. You could say a lot of the AI native companies that have been born in these past three years, hand wavy back, you could say the vast majority of those are a kind of wrapper. Or actually, see, the thing I like about this is language is such a wonderful thing. The way in which people are now starting
### Wrappers vs. Harnesses
(36:24) to talk about this, oh, is this not a wrapper, it's a harness. The whole thing around Claude code. Oh, no, no, this is a harness for it. And again, to some degree, I actually think that's right. And there is proprietary insight, which apparently Anthropic just accidentally leaked to the whole world, but there is proprietary IP and how do you structure
### Next Questioner
(36:46) a harness to do a particular task well, even if they're all sharing the same underlying LLM.
>> All right, who else has got something for Scott? These are fun.
>> Are you daring me? Because I will. Okay, Earl, you could take the rat sock is dangerous.
>> No, no, no, not at all. Not at all. Everything you're saying, Scott, I was
### AI-Only Solutions and Adoption
(37:06) curious about all these new companies that are coming out with their AI only solutions. And I know that especially in our crowd, we're very interested in it, especially to see where it's going. But I keep going back to what I learned about product and product development, and I'm not seeing the adoption that you get from the civilian side or the technical side in
### Crossing the Chasm with AI Tools
(37:30) terms of all these different tools. For example, David will talk about Vibe coding all day. I can talk to you about custom GPTs all day. But we are very selective crowd, and I'm wondering, think in terms of crossing the chasm. How many people are actually using these in their companies? Companies are pushing them and investors are pushing for adoption. But how many
### Ownership of AI Adoption in Enterprise
(37:50) actual people in their, especially enterprise organizations, are actually pioneering these things and pushing these things forward? Back to the question earlier, who is going to own the AI adoption process? That's the thing I'm struggling with. Who do you talk to when you want to talk about AI to help them with their AI?
>> Okay. So there's a few things.
### Two Questions on AI Adoption
(38:11) Two different questions. What do we think about the civilian adoption? And two, who do we think should be owning the AI adoption process internally at enterprise organizations? Two separate questions. For the first one, there's a place where we have data, and then there's a place where at the moment all I have is anecdote. The anecdote is what I
### Lack of Adoption for New AI Products
(38:32) hear anecdotally is most of these AI features, and then also a lot of these new AI products, they're not getting adoption, for a variety of reasons. Again, I think even the weekly active users on
### Breakthrough AI Products
(39:17) things like ChatGPT or Claude, when you see breakthrough products like Lovable, And so I think it's interesting, for the most part, people don't want to have to learn new stuff if they don't have to. But out of the thousands of attempts that are throwing things against the wall right now, there
### Bifurcation of AI Product Success
(39:43) do still tend to be several dozen or so that, oh no, this catches on and people realize, oh wow, I can do this and this is great, and then they tell two friends and they tell two friends and so on and so on. So anyways, it's a bifurcation. I think it is
### Who Owns AI Adoption?
(40:04) possible for there to be great takeoff, but only for 0.2% of what's in the market right now.
>> Gotcha.
>> What was the second half of that question?
>> Yes. Second half was, who do you see, at least from your research and your anecdotal stories that you've had, who has been the owners of the AI adoption?
### AI Adoption Ownership Roles
(40:28) Because, obviously, it depends on the size of the company, whether it's SMB or enterprise, I get that. But as you said, it can be anybody from the CMO to the CIO, maybe even if they have a CPO, chief product officer, that might be a person involved, who knows? But what have you seen from your research?
>> So I see three roles.
### CEO and CIO Roles in AI Adoption
(40:50) This is CEO, who basically tells everybody they need to use AI. That's about as helpful as it gets.
>> You saw what happened with a couple companies about that. Better to rehire them back.
>> It's a wacky time. The second role is, it is generally the CIO or the IT organization that's going ahead and getting the enterprise licenses and
### Individual Teams Drive Genuine Adoption
(41:11) manage them for the Frontier Labs and some of these major platforms. However, neither one of those things actually speaks to genuine adoption, much less actual genuine impact and outcome.
>> Everywhere I see it, I'm trying to think of other exceptions. It happens so much down in the individual teams.
### Zapier CEO's AI Framework
(41:37) There's a few companies that have been forward about this, oh, what's his name? The CEO of Zapier. The whole Zapier company has been obviously core, their product has been very much on the frontier of this. But he's been publishing his, okay, and this is the framework we use of what we're expecting
### Lack of Top-Down AI Guidance
(42:00) from people. And it's not at that level of the CEO saying use AI. It's, no, no, actually these are the different kinds of things in the use cases, and how do we measure it? But he almost stands out because that is such an exception now that in most companies, there isn't enough of that guidance top down to say that anyone is taking ownership
### CRO Interest in AI Adoption
(42:20) of adoption.
>> That makes total sense. I think that companies that have a C-level executive for revenue, for example, a CSO, a CRO, would be interested in that, considering that they can make a big deal from their data, from they're getting that they're getting from their sales teams. But I don't see it in the market.
### Sales Team Tool Adoption
(42:41) >> Learning new tools is not a sales team. It's kind of big, acquire me from my experience.
>> And they just gotten excited about Gong. Okay, we think we finally got our arms around this.
>> Baby steps. Got woo. I'll take it.
>> Oh, and then Clay, everyone's, of course, we go to market, we've got Clay as if that's our
### Brand Rebranding Surprise
(43:01) magical thing. But
>> I think they rebranded, they rebranded the consumer product to Mesh, I think recently.
>> They rebranded, I believe.
>> Seriously? Look it up.
>> Emsh. They rebranded the consumer product.
>> Hang on. I have to Google this.
### Difficulty Keeping Up with AI Changes
(43:22) >> You Google me your own.
>> Oh my god.
>> I would have sworn that was an April Fool's joke.
>> No, but okay. All right. I'll have to, all right, I once again go back to my disclaimer at the beginning of this thing. I can't keep up with all of this.
>> Nobody can. Not even AI can. Meanwhile,
### Agentic Commerce Inquiry
(43:41) Yogish, you want to chime in?
>> Hey Scott, good to see you again. I had a question that's a slightly different one. In your, I know you were sharing a little bit of tidbits from your upcoming release of your new report for the year. I was curious if you were seeing anything around agentic commerce showing up this year in your analysis.
### Lack of Agentic Commerce Adoption
(44:04) >> The short answer is not a lot. We actually, and this is why anybody who says they know the future in this market, we were actually expecting to see more of that, because at the end of last year, all the big announcements from OpenAI and then Google, and everyone's headed into that, and then it's generally turned out for the
### Consumer Readiness for Agentic Commerce
(44:29) most part, consumers aren't ready for this in a lot of the cases. But part of this depends on how you define agentic commerce too. Because there's these things, I call them the shopper concierges. There are these AI experiences, whether it's a dedicated app or something like this, and you could, depending on how loosely you want to
### Science Fair Stage of Agentic Commerce
(44:54) define agentic, you can see that. But actually having agents go and do these things for me, it's still everything we're seeing right now, it's still science fair.
>> Got it. A quick follow up on that. Based on what you're seeing, at least from the larger LLMs, OpenAI and Google's Gemini, I'm curious to hear
### AI's Impact on Consumer Discovery
(45:17) what your thoughts are on what you're seeing in the space.
>> I think what OpenAI discovered in there is there is this massive shift that has been happening of consumers starting to truly use AI for discovery, and also for evaluation and weighing different options and stuff that, which, to be honest, again, in two years, the
### Industry's Slow Absorption of AI Shift
(45:46) whole nature of how people do discovery and evaluation online is shifted. And to be honest, that hasn't fully, I think the industry, we collectively haven't even fully absorbed that and learned how to deal with that well. And I think when OpenAI was retrenching away from this, they're, okay, this is we're going
### Merchant Resistance to Ceding Discovery
(46:12) to focus on for the core business. I think that's where the balance seems to be. That's obviously also one where the merchants involved are willing to cede that, because they'd already had to cede that once before. Google, not happy having to relearn this all again, but open to the possibility that they will not fully control the discovery channel, unless if
### Walmart's Stance on AI Discovery
(46:35) you're Amazon, they won't control it. But what was it, the Walmart head of AI had that quote at an investor conference when they were, the OpenAI thing was letting ChatGPT do that, that was a temporary moment in time. Walmart's, the hell we are going to cede over, and it become your fulfillment
### Consumer Trust and Merchant Readiness
(47:01) service on the back end here, my friend. So it's both the lack of consumer trust, plus the fact that I think the merchant community was maybe caught a little bit off guard when this first hit. All the ones I talked to, they're, we're not going to, we're not going to walk into that if can help it.
>> I get that. And one thing that I find fascinating in
### Merchants Creating Walled Gardens
(47:26) this whole space right now is that the speed at which some of these merchants are investing into their own, call it walled gardens, in a way, is going to create an interesting mix for brands in terms of how they'll be able to engage across different protocols, right? And I wonder if that's going to open up a new space for startups to be able to offer for potentially new
### Retail Media Networks and Fragmentation
(47:50) solutions. I don't know. I'm speculating here a bit, but it feels that that's where things are going. So,
>> Well, in some ways you could say this is what's happening with AntTech. This explosion of these little retail media networks is actually now once again we've got a fragmentation
### New Opportunities from Market Fragmentation
(48:09) in a market. And so that creates, we were headed towards what the duopoly, and it still largely is a duopoly. But now is enough interesting things happening in this fragmented space that you're starting to see the emergence of software vendors who are great, we can help in that environment. So,
### Upcoming MarTech Report
(48:33) Scott, any other things that you're excited about coming out next next few months? Anything else we should all be paying attention to?
>> We'll have that State of MarTech report out at the beginning of May. So, we now distribute that free and ungated. So, whenever that's ready. But, no, thanks for inviting me to have this chat with you. These are great questions. I love this conversation.
### David's Gratitude and Future Guests
(49:15) Welcome by anytime. We'll make sure to share your latest state of things report with the community. And for everyone, we've got. I almost get embarrassed sometimes when I'm now working on booking things, and I'm, well, you're an amazing guest. Let's look at July. So it's a fun fun problem to have, but it means that we
### Upcoming Events and Holidays
(49:39) got a lot of conversations coming, including with at least a couple of folks on this call today. So appreciate you all coming by. Stay tuned for more. Check the Luma for a lot of what we have scheduled, and a few more things we probably need to add. And anyone in New York next week, we're going to be doing a belated first Wednesday, because tonight is the first Wednesday,
### Holiday Greetings
(49:58) but also the first night of Passover for a lot of folks celebrating here. So happy Passover and happy Easter to everyone celebrating the next few days. And we will see you all very soon.