The Future of the Connected Customer Stack | Zack Wenthe & Adrian Bolosan | Signal & Noise LIVE at Advertising Week 2026

Watch on Youtube

Listen on Spotify

What happens to the enterprise marketing stack when AI becomes more than another feature—and starts changing how data, technology, people, and decisions actually work together?

Recorded live at Advertising Week 2026 in New York City at MCM Creative Studios, Rio Longacre and Brett House sit down with two longtime friends of the show and two of the smartest voices in data and marketing technology: Zack Wenthe, Director of Product Marketing & Customer Data Evangelist at Tealium, and Adrian Bolosan, Head of ISV GTM, AMER Enterprise & Emerging at Databricks.

Rio has known Zack and Adrian for many years, and this conversation goes well beyond the usual AI talking points. It’s a wide-ranging discussion about how enterprises are actually adapting their MarTech and AdTech stacks for an AI-driven world—and what’s real versus what’s simply the latest industry hype.

The conversation starts with one of the industry’s favorite words: composability. Has the composable stack actually won? Or has it become another catch-all marketing term? Zack and Adrian explore why enterprises are moving applications closer to their data, the growing role of federated architectures, and why flexibility increasingly matters more than buying another massive monolithic platform.

From there, the discussion gets into the foundation underneath every AI strategy: data. AI is only as useful as the signals feeding it. The group discusses data collection, tag management, governance, identity resolution, consent, context, and why enterprises need to understand what data they actually have before trying to unleash agents on top of it.

They also tackle one of the biggest unresolved questions in marketing technology: Where should decisioning live? In the data cloud? The CDP? The CRM? Or directly inside the activation channel? The answer may increasingly depend on the moment—combining historical intelligence from platforms like Databricks with real-time signals captured at the edge.

The conversation then turns to how AI is changing the marketing organization itself. The traditional marketing technologist is evolving into the marketing engineer—someone who can connect systems, data, agents, models, and workflows while still understanding the business problem being solved. At the same time, marketers are gaining the ability to interrogate enterprise data through natural language instead of waiting on analysts, engineers, or data science teams.

Rio, Brett, Zack, and Adrian also dig into:

• The resurgence of tag management in an agentic world

• Why context may matter more than perfect one-to-one personalization

• Identity resolution and the changing economics of third-party identity

• “Perishable moments” and the need for real-time customer decisioning

• How AI could break down data and organizational silos

• Why human judgment becomes more—not less—important as AI gets better

• The rise of disposable, use-case-specific applications

• Whether the predicted “SaaSpocalypse” is actually happening

• How AI is changing consulting, systems integration, and enterprise technology buying

• Why the companies that win with AI will combine strong data foundations with people who know how to turn technology into business outcomes

  
  

For anyone working in MarTech, AdTech, customer data, enterprise AI, cloud data platforms, advertising, or digital transformation, this is an outstanding conversation about where the enterprise technology stack is actually headed.

A huge thank-you to Tealium, one of the sponsors that helped make our Advertising Week programming possible and an ongoing sponsor of Signal & Noise. We greatly appreciate Zack and the entire Tealium team for their continued partnership and support of the show. And a special thanks to MCM Creative Studios for hosting Signal & Noise LIVE during Advertising Week 2026.

Hosted by: Rio Longacre & Brett House Guests: Zack Wenthe, Tealium | Adrian Bolosan, Databricks Recorded live: Advertising Week 2026 | New York City

#ArtificialIntelligence #EnterpriseAI #MarTech #AdTech #CustomerData #CDP #DataCloud #MarketingTechnology #AIinMarketing #AgenticAI #ComposableArchitecture #DataStrategy #DigitalTransformation #MarketingEngineering #CustomerExperience #Databricks #Tealium #AdvertisingWeek #AdvertisingWeek2026 #SignalAndNoise

---

Transcript:

Miles (00:03.816)

signal and noise. Bringing clarity to data, tech, and AI.

Miles (00:13.742)

Hi, this is Rio Longacre here with Brett House, Sigilin Noise Live at Ad Week 2026 in New York City. And we are in for a real special treat for this episode, Brett. I'm really excited about this one. Totally. We've got two incredible guests, but before we jump in and tell about who the guests are, let's first thank our sponsors. Yeah, so Telium, PubEx, Gray Deal, and our prime lead sponsor, Media Radar, not to mention Nexa.

A little stealth software company, tech company that's gonna be coming to market shortly. But but big thanks to everybody. Media Radar has been hugely helpful from a team support perspective, helping to make this possible. we literally started out last year, one year reunion from Ad Week.

In a tiny corner of the pr and I keep telling this story, but it's it's we commandeered part of the press room and got the press room. Yeah, and and now we're here in this Hollywood production studio with John Hann ham upstairs and and and welcome to the show. This guy's this is gonna be great. Yeah. So we've been looking forward to this one. So our guests are Adrian Bolison, who is I've known for a long time. who leads ISV go to market for Databricks, incredible company, great partner for of mine for many years.

And Zach Wenthe, who's director of product marketing, as well as customer data evangelists at Telium, who's Brett mentioned, is not only a longtime partner of of mine, but one of our sponsors for not only the podcast, but this incredible. A media and an event sponsor. You guys have been great. Yeah, it's been awesome working. So excited to have your your team in later to celebrate with our tequila sponsor. Or best. So so welcome, gentlemen. Thank you. Thank you so much. This set is amazing. So yeah. Thank you so much. Yeah. Yeah. Sorry, I'm not John Hamm.

But you know, hopefully we'll have at least a good conversation. Yeah, it's funny I told my wife that he was probably gonna be in the building today and and she she almost came out came on this trip with me and we had this and she told me one time, I don't know if she's joking or not, that he he's the one person she would leave me for. I think I think about ninety-nine percent of women on the planet Earth have said that at one point or another.

Miles (02:17.76)

It's like, God damn it, you know. And we and we we do talk about I mean people that have not seen Mad Men, if you haven't if you're in the industry, it's a terrific show. It's a lot of of like what is old is new again. You see a lot of those con those themes coming back of like where is the serendipity in advertising? Where is the power of creative, right? We're all about data and science and and you know, precise targeting, right? Where is that part of advertising? And so that's a that's a big conversation. So it's it's sort of

You know, it's it's serendipitous that he's upstairs, right? And that's a big theme. I think of ad week in general and the industry in general, right? Totally. So it's a bunch of stuff we wanted to talk about with you two gentlemen. But the the first thing, it's a term that was been thrown around a lot. Maybe you could say it's been overused, composability or I mean I think modularity is maybe the newer iterations. So is it overused, is it underused, did we finally achieve it? Has it won? Zach, start with you, thoughts.

I I I mean yes, it's a wildly overused term. you know, everything is is composable and that's either been a product stop gap, a a way to say, well, we don't do everything, so we're a composable piece, all the way to there is a legitimate need for flexibility and and and need. So it's like in one way, yeah, I can understand where

The term came from, but I think it's become this blanket term like so many others that but counterpoint, I don't know if it's necessarily only that we don't do everything, so we're just gonna focus on this one or two things. Rather, we don't want to s it it might be a s a business model, a sales model, right? Of we don't we've got we we don't want to sell a monolithic stack because not everybody needs these things that solve 15, 20 different use cases. Maybe they only have one or two. We've seen it on the commerce side, we've seen it on the data side, right? They have one or two use cases.

You get in the door at that level, it makes for an easier, kind of frictionless sales process, right? And then you can expand in in in based on based on client need. So it's maybe the client need that's kind of rising to the surface, saying we don't want to buy your monolithic gated SaaS product, right? Because we're not gonna use eighty percent of it. Yeah. Is that a Yeah, I think that's fair. I think the the one dot is the reverse ETL, which the one dot five was the composable. I think it gave organizations a lot of commercial flexibility.

Miles (04:35.158)

Yeah. Right. To your point, I don't have to come into the door as like this full behemoth type of product that's this huge lift and shift. Why don't you just take bits and bytes of what you need? And it gave them a lot of flexibility and it's kind of like a land and expand. So some of those that kind of brought that composable name moniker out, you would argue they're pretty pretty large C D P players now too, right? So it's like one thing that they once kind of talked about is what they have become. It's kind of it's kind of interesting. I think you made a good point and

Brett, I'm gonna agree with you as well about it being some of composability became a thing. It came out of the whole reverse ETL thing, which I thought was always a strange way to describe it anyway, right? But then And describe that, define that, just so people know what you're talking about exactly. Well composability and no the reverse ETL. Well reverse ES ETL is just being able to like s activate data from I mean theory from a zero copy, right? So you're basically taking it from your cloud data warehouse and you're using like a

a tool to to pass it into your activation channels where you're supposed to be zero copy so it was which is in essence copying the next yeah so so I mean zero copy is being entrenched. Also the worst term but it was yeah yeah that's another term that's been overused and abused and so that was the theory right and like I don't like you could argue like were these solutions really zero copy? What is reverse CTL? Was it just a marketing or product gimmick? And then I found it interesting too a lot of these monolithic platforms

right after composability was the thing, suddenly suddenly they were saying, we're composable too. So which maybe really agree with your point, Brett. Maybe a lot of it was just was just good marketing. Yeah. Or good product marketing. But but is it like has that model one though? I mean I guess I think from a technology standpoint, the definition of composability, at least through my lens, is actually bringing the applications closer to the data. So like regardless of anything else, I think the spaces agreed that that is a more of a composable

Like model? Yep. Regardless if we're talking CDPs or other technical stacks, it's like, can I just bring the app to the data where it lives? Instead of shipping data everywhere. Exactly. Yeah, yeah, yeah, exactly. And and and do you think that that is the is there a case for sort of a federated data architecture to support that where the where or do you do you really fundamentally still need to centralize your data? Speaking of data bricks, we were talking to Eric Foxworthy last night.

Miles (06:55.608)

With stuff like, you know, there there's there's been a moving along, you know, for fifty, sixty years of centralizing data from various repositories. Right. But we now have more federated capabilities, right? Where you can have agents or headless systems go into multiple places to pull, you know, the ERP to pull this specific data, the CRM to pull this specific data. Instead of taking it from ERP and transforming it into the and then trying to bring it in and normalize it and deduplicate it and resolve it, right?

Do you think what you think of that from a data architecture perspective? So like the the genesis, we obviously want more data and do the Databricks lens on Databricks. But because we're open and interoperable, federated like tables and views is heavily used, meaning we don't require everyone to have to do that. You can connect your data via federated if you've made different data warehouse choices or whatnot. And so we think through federation it still allows for like governance, which is our ultimate

A central governance layer over wherever the data sources are. Correct. Right.

Zach, another thing I wanted to dig in and I'd maybe start with you on is, and we talked member, we talked to Jared Tippets about this, who was from Hilton, right? We had a really good conversation with him about you know the importance of not only just tag management, which is obviously putting tags in websites that so you can so you can track things, right? And pull them back into your into your different applications. But the the importance of that, importance of having a good tag management strategy and implementation to to be able to get the right data in order to not only have

Just generally better data, but also if you're going to be creating agents or you or informing, building a governance layer, it starts with the right data. Thoughts on that? Is is that a trend you're seeing? yeah. With type management, it's been one of those things that's been around for for for ages. Yep. And you know, Telium It's overlooked by the way. it's overlooked, very much so. I mean, you and there's there's only a couple vendors in the space, Telium obviously being one of them, but

Miles (08:48.052)

As we've independent tag management company too, right? I think there's just like Adobe and Google and you guys, right? Yes, yeah, definitely. you know, we've gotten into this world where everything that was hot 10 years ago is all of a sudden becoming hot again. It's funny so funny. Real time tag management, yeah. It's you know, and and because people are starting to realize again, the more context, well, that's another overused term, but we can we can talk about and debate that here later. But yeah, the more information what the context layer means. Yeah.

The more information we give to an agent about what's going on with the consumer, the more likely we are to actually get a a agentic response that is meaningful and complete and and you know and understandable. so yes, there is a analytics perspective, there's a advertising perspective, obviously pushing that to you know your ad platforms for conversion tracking and and retargeting and whatnot. But now, I mean realize most of our consumer behavior.

a vast majority of it happens in a digital channel. Our browsing, our searching, our our research, our re reading reviews and and then Like interactions. L L and then buying, right? I mean not most people are not walking into a store, sitting down with a with a sales rep and saying, Tell me about your product, right? That just doesn't I mean though eighty cars eighty percent of all of purchases still happen in stores. Right on the like the region. Yeah. Yeah.

Although during COVID I bought my car over text message. So, you I mean it's the better right. But having that behavior and knowing how long did they look at this, how how you know, what did they where did they go next? LLMs now have the ability to take that and form a story. Yeah. Where a traditional rules-based personalization engine really couldn't say, well, because you looked at these three pages and then you went and looked at

The shipping FAQ. Yeah, it's a it's like a it's a Boolean logic solution that is something that's super dynamic, that's always changing, it needs to be always on. You interpret it can it can make that interpretation and it's predictive analytics of like what's gonna happen next? What is the next best action for us to take based on this consumer path? What you know, how are they, you know, what are they are they in a buying mode, are they in a research mode? And being able to classify that pretty accurately now.

Miles (11:04.684)

Yeah. Is just this resurgence of of tag management being a a tool for that whether it's going right to an LLM or it's going to a data warehouse, you know, as part of the clickstream, you know, there there's just a lot of new applications for it. What are some of the challenges and and maybe both you guys can answer this around associated with with sort of managing tag bloat, right? Like we've said you know, where brands have tags all over the place.

They might go back teams that different leaders, different teams, agencies, and suddenly you have this, it's kind of the classic data house in order, get your data house in order problem, right? Where they've just got data that they don't even know it either exists or or how do how do you actually help brands clean that process up? It's almost like you need to start with a clean slate. Yeah, I think a Gentec has kind of expedited almost hygiene of the intake of data. So it's just like

Well, what do we want the agents to do? What are the ingredients to power these agents? And then they start to look at their inventory of all this data. They're like, what is all this other signal over here? Are we even using this anymore? And then I think that goes back to them using best in class like data collection at the edge, like Telium, to really get that rich sign the the signal that's most important to them, right? And I think a lot of the to the bloat is like, well, we don't need all this anymore. Maybe you continue to collect it, but maybe you're not leveraging it, right? So it

It really starts to prioritize what is the true signal they need to make these decisions. And with clients, do you need to go and clean that up, or can you basically build let's say a semantic layer, a translation layer that says ignore X, Y, and Z so that it can it can, even though there's a lot of noise in your data, it can actually go right into that unstructured data and pull what it needs to make decisions.

Right, is it or or is it w is there a cleansing process that has to happen on the client side. I I think cleansing is always best practice, right? You should if if if you're firing a tag that you no longer are sending data, you know, or using or you are firing a pixel for an ad platform you're not using anymore, then obviously that's just wasted noise and wasted data and bloat on your page and and and so yeah, obviously you should always clean. I you know, I've advocated for a long time that every organization should have a data dictionary. Yeah. You know.

Miles (13:15.016)

What's the what's the data source? What's it for? Where's it going? How do we collect it? What part what side of the the consent equation is it on? Do we collect it after And that's so much easier to do now with AI though. And that exactly. That is one hundred percent. Now I can go and point my, you know, an agent of your choice at, you know, a data store and say, map out what do I have, w what what's the timestamps on these, when were they changed? And you can start to get something that you may have paid a consulting firm

to go do for you in the past. It's funny, like with when C C PA, the California Consumer Privacy Act was was about to pass, I mean it's funny, like my old job, we created this point this offering around helping brands we weren't w weren't offering legal advice, but helping them prepare for it from a technology perspective. It's funny, one of the things we would do was we go to their website and we look at all of the pages. We'd actually look at HTML and look at all of the tags they had and pixels they had. Because you know for those who don't really understand this, like when you if you want to

it all of these platforms and these ad ad campaigns I'll they'll give you a little bit of code you'll put in your page. Every time the page loads, like that that code will fire, right? It'll s it'll trigger pixel that tells someone like they sent traffic there or whatever. And a lot of these things will actually pass information back sometimes, right? so it's fun we'd audit these things and it would be an insane amount of tags in these pages. I mean, which is crappy. Like

You would slow down the loading of these pages. Yeah. Which affects your experience. Internet's pretty fast now, so typically you wouldn't it wouldn't make that much of a difference, but it does slow down page load. You're putting all this garbage code. But think the wild thing was though, they wouldn't even know what half these tags were. Right. And it would be just a laborious process. I think now with the I you can just tell an agent go to go to the HTML of every page, create a table for me with every with every tag and try to figure out where is this tag pointing back to. Yeah, exactly.

Almost like frequency of use, how is this being utilized in other channels, right, that are that are important to me and eliminate the ones that that are irrelevant, right? Yep. Yeah. Right. Yeah, I mean there's a compliance benefit, there's there's obviously just an overhead benefit for a a team managing, you know, that that tag profile and the updates and are we, you know, are we using the 'cause you know the vendors change their their tag protocols all the time, meta and others, you know. I mean, so it's like are we using the right even the right version? Are we sending the right information?

Miles (15:28.706)

Would we get better results if we're sending X, Y, and Z? And so just having somebody who's owns that dictionary, whether it's a center of excellence, a team, whether it's a person, whether it's a department, I mean that that kind of depends on, you know, the how how a company operates. But somebody looking at that every once in a while is is the difference between a compliance problem often and a, you know, a defendable

Yeah. Data position. Yeah, yeah, yeah. Well and you wonder you wonder if that's something that you just automate out of people's lives, right? Because you know, the problem with the dictionary is that often just collects dust. It's like your encyclopedia set on, you know, like in the bookshelf and no nobody ever looks at it again. It's sort of a static piece of content that nobody ever updates, right? So but if you could have this digitized and basically trained, so the agents are sort of constantly updating and and checking.

Against you know a checklist of of sort of data it's a data governance layer that's being agentically powered, basically. Yeah. Right? Based on a data dictionary. So it's got kind of rules of engagement with with this data and with these tags. Aaron Two things that I think are in flux right now because of let's say the the shift towards number one more agentic and things going headless, but also I think the the trend of that we mentioned things generally being more compos composable. I'm defining composable as more like Headless.

Headless or the the fact that you know s solutions will do different things. And maybe it's one company, one tool does several things, but you can you know can rip and replace certain parts out more easily. I think headless is accelerating now. But I think that was a trend anyway. So I think those two trends are really like creating like two things are kind of up for grabs, I would say. I mean, I I would say one of those is is identity. Like where does it sit now? Right? Maybe we start there. Like the role of identity.

Where does it sit? Is it something that branched on their own? Should he use the third party? Like these are all questions that I thought we saw 10, 15 years ago to your point. Yeah, a lot of different identity providers, right? I came came from one with New Star, but there's the Experience, there's the Epsilons, right? There's the Axioms that all kind of came derived from similar database marketing. Back in the day, right, and transferred this over.

Miles (17:37.464)

Do they have any you know, not to be anti-competitive or or anti anti any of these pro providers, but do they have a real right or provenance over identity as a as a as a is it you know, or or can that be democratized across the ecosystem? Your thoughts and I just don't, Adrian. I think you guys know I spent 10 years with an identity company. The I think the value's still there. It depends on the brands, right? So we obviously know that retailers have a lot more first party data than CPGs do.

So they continuously license these huge graphs because they're not direct to consumer, right? So I think that value's there. I think in being able to license identity resolution like one record at a time, the commercials around it are changing, meaning like I don't need like I just need identity when I need it to fill gaps and things like that. So just I think the commercial models of these things are changing. Also the friction to go from zero to identity resolution's changed drastically due to platforms like Databricks. where

organizations no longer have to go through like a bake off period and then the CISO review and then sending all the data out to see what the match rate is and come back The match rate tests. Like this is I still have like nightmares about going through these tests, right? And that stuff is essentially happening because of composability or native identity solutions. Like you can get a read on that in minutes and decide yes, this is going to be valuable or not, right?

And so I think that's the biggest thing that's changed. And I think identity providers will continue to show value because they still have to collect or sync offline to online. I think agentic solutions will help some of that like drastically over the next and I think those same identity companies are using agentic internally to make their own products better, right? So it's like a race for is the brand gonna do it themselves? Like there's you have to have a lot of know-how in develop d like creating graphs. I've been in a lot of conversations like we're gonna develop our own graph.

It's a complex data matching machine learning exercise of like how do we map these tables together in a way that at a certain level of confidence, we're sure that this, you know, whether it's per driving personalized measurement or personalized marketing is almost beside the point, but like are we confident that this household's composed of these people with these devices, right? And and what level of confidence do we have? Is it ninety-five percent or ninety percent? So can can an identity provider bring incremental data to the table?

Miles (19:59.286)

Right, that is going to make the fidelity of a consumer record that much higher. Right. That's really the value. If they can't, and if they're all commodity and they're all pr pretty much the same, then there's not a I don't think there'd be a ton of value, right? Yeah. And I think the black box approach, what you're describing, is gone away. And for brands and advertisers to be able to throttle that based on what the use case is. So if you're going for measurement, you need high fidelity, like

Yeah, know, pr that needs to be ninety percent plus. If you're going for lookalike scale stuff, maybe not so accurate, right? Like you have to dial it back. These were levers that identity companies weren't providing. Like towards the end of my career and experience, like we started to allow them to kind of have those levers. I still have never really seen that in a UI. I've seen a lot of product leaders talk about that, right? Like less accurate to more accurate in terms of a reach or like a targeted marketing goal. Yeah, marketing versus like a transactional emails. Obviously, you'd want

Like like one you'd want to be pretty much a hundred percent accurate marking maybe yeah maybe you're fine with ninety percent. Yeah, yeah. Well, and does a lot of this conversation in down the path in in of of like one to one personalization, which which I think we've made the perspective pretty clear it's bullshit and it's not necessary. There's always a type dream and it was it was never acquired. And we're and I think we're as humans our egos get in the way of being overconfident that we're actually that different from one another, right? And so

To us it was it was, you know, and we've talked about this with a bunch of different leaders, right? Of of the just the power of cohorts, right? When you have a cohort that's tied to a very specific, it could be a a demographic group, it could be a fan base, right? Whatever, there there are a lot of like similarities that are good enough, right? And you don't need the signal that you would need at the person's level, device specific, et cetera. Right. So I mean, our identity solutions, do they thrive in a world where where the one-to-one lie?

sort of perpetuates or do you think they have a play in the cohort based on I think they play in both. Like I I think we as a community we shouldn't give up on the endless pursuit of this unique experience for every person, right? Like that's that's what we're all aiming for, to your point. Like, has anyone gotten there yet? No. But like a gentic is allowing us to definitely go, you know, like I'm not a cohort, I'm a person, right? Like I think we're going that direction. I think identity providers play a part in that.

Miles (22:20.92)

They've really done so more so in the cohort space today, because no one's match rate is a hundred percent. Yeah. So by de facto there's gonna be cohorts in there because you're making up some type of model for the rest of them. Yeah. But but you're saying the promise is that we should get to one to one. I think that should be the endless pursuit. Yeah, but do you think that's there's an ick in surveillance capitalism effect right there when it comes to I think we're hitting this world where where consumers are starting to be aware of

amount of data that's out there. Started I mean it started back with the Cambridge Analytica thing. I think it's obviously the first tipping point. But you know, it's grown and and and people are like, I don't want to give you I want to give you just enough so I can have a good experience, but not too much so that it's yeah, it's creepy. I mean I remember years ago when Marketo was selling, you know, they would their one of their sales tricks was like you'd click around on your website, they call you up and be like, Hey, I saw that you looked at this, this they and they just read back your profile to you. And you're like, as a marketer, I'm like, I know that's what they're doing.

But it's still creepy when they say it. I hated that. Yeah. It did it did not work on me. I just thought it was like just you don't have to You don't have to say it. You don't have to say it. It's just I you know all this on me. Anyone who works on online marketing at your point, they know like I I but I think the normies don't know that. Right. Right. Like my mom does not know that, right? And it cre it would creep the shit out of her when she do that. Yeah. And you're trying to predict in

based on a particular behavior that may not spell any sort of detent intent at all. No. And and we have to realize that the situation that somebody is in is a much bigger predictor of outcome than the person or the demographic. Yeah, yeah. The context not using context in like a data architecture. Right way. Yeah. But the context of where they are, time, place, time of day. What else is going on in their lives. Yeah, they're they're at a live event or they're at home watching television, all of these things.

are gonna increase recep receptivity if you look at them correctly and you say person's gonna want to see something like this at this point. And all these things, Brett, have been totally underused when you look at signal from, let's say, supply the contextual signals and how they're being used, certainly for advertising, right? I think those things have been horrendously underused in terms of applying context to actually the ad targeting and the message cr and the message personalization, for sure. Another thing we wanted to dig, and this is the other thing I think kind of

Miles (24:39.07)

Like back to the future. It's like you know, it's like up for debate again, where does it sit and is decisioning. Like if the data warehouse, the cloud data warehouse ends up being let's say the the you know, the main source of truth and then you have, you know, not zero copy, but you go into less you're not shipping data all over the place anymore, whether it's tag you know, tag management data you're you're collecting from let's say your your own properties or it's paid media data from your from your you know from your from the

you know, paid media channels and platforms, or it's your CRM data. So there's less data moving around. Where does decisioning actually sit now? I mean, you're that argument I think was ten years ago, is like marking automation to marketos like, it's gonna sit in in your marking automation platform. Then your C D people are saying, no should sit it should sit in your C D P then the cell the C R are saying, No, should you sit in your C R right? So that was never decided, but we just forgot about it for a few years. I think it's come back now, right? Now that we're in this now that we've come to this point. Thoughts on that Zach? Like

Where did where where should it sit? Where does it sit? Like I I I may be the contrarian in this group here. I think it's gonna sit at the channel. I think it's gonna sit at the ad layer or or the the ESP. I don't think it sits in the data warehouse. I don't actually think it sits in most CDPs. because I think one is the time delay from a warehouse to an action on a website is too slow to take a decision, calculate said decision.

Push it out for a in-session type experience. Yeah. If you want real time, yeah, there's no way. You know, and then and then email, it's like, well, if I have all the context there, now I have to put all of that open, click, read, you know, scroll rates, all the stuff that say an email channel already has. I have to move that all back to my warehouse. That may be beneficial for some. That may just be extra data and and and and frankly, extra consumption and cost. And does the return of moving it to the data warehouse give you a better decision or not?

And I think for a lot of organizations it's not. Yeah. It just do I send one email or not, they're gonna send the email. Funny you mentioned that before for Adrian, we pass over to you for your your answer on this, about the is it is the juice worth the squeeze? It's funny. I I would have clients I remember this four or five years ago, I'd have so many clients ask me about real-time per like web personalization and they, you know, because one of like one of the platforms had come up with this part of their CDP and they were pushing the hell out of this. And remember I had clients ask me about it, and I remember like some of them didn't even need it.

Miles (27:04.77)

So why would you need this? Like you're selling insurance, right? Like, yeah. You don't need if you have four or five cohorts and you that's all you need. You don't need to like and remember there was no use case and they were saying B2B. Like there's like there's no point in even trying to do this. Though you could maybe do this, like maybe like if you completely re-architected and spent tens of millions of dollars, but like you wouldn't have any return, even close to return investment of this. I remember this discussion I have with many people. I think it's gotten a little easier now with some of the new technologies, but like

Like that was my conclusion with a lot of clients is like, yeah, do we is it worth it? Maybe not. Well, and I think now it's shifted though, because now with AI, AI needs fast stuff. And now you can start to do personalization that is more meaningful, even if it's cohort based, that that ha can happen in session, whether that's text based, that's imagery. you know, or if you're in a re I mean, I I rem I I have customers who are in, you know, like quick service, you know, restaurant or or fast food, you know.

An abandoned cart on a food app is like your your your time to influence them is minutes. Right? You can't wait an hour and send them an email and being like, hey, you left something in your cart. Well, they're already done with lunch at that point in time. What about that burger? Yeah. So you know, you have to be able to and then yeah, obviously, you know, in in insurance sales or you know, something that's that's a long, a long sales cycle, you don't need in session necessarily. but yeah, I think the world of of real time, the world of decisioning.

I don't think we're gonna solve it. I don't I I think it's gonna continue to change. I think it if you look at what's happened in the last week in the world, we've got Jev now, we've got open AI decisions, which are really fast models that are built to give a a decision answer. The the fact that I can do that in milliseconds now and and Telium can push like we can push out a profile and get an answer back right away on what cohort do you belong to?

Or are you a gifter or are you a a a you know, a gift giver or gift buyer, a high value shopper? We don't have to have these like complex models sitting complex models sitting running on a on a on a batch process anymore. Yep. Yeah, so I think defin to Zach's point there's perishable moments that are happening. I th I view what Telium's doing on the edge of like what's happening now. And there's decisioning that hap has to happen there at speeds and SLAs that like Databricks wasn't originally designed to support.

Miles (29:31.372)

And then there's decision on what happened with that customer if they're an existing customer. And like you have models, I think organizations that are marrying those two, they're the ones that are winning, right? So it's really like multiple decisioning based on what did happen previously and what's happening now. And marrying those two together. Some of that's gonna be in real time, some of it's not, because it's just not technically possible. Well, but it's and it's helping to predict sort of next next best. So when you when you arrive back at that set or back in that content experience.

That's when things should update because w you know, whether they've seen you before or not, if you can do some level of data enrichment to say we know the basic demographics, the basic geo, the basic behavioral, contextual, then you're able to show something that's that's more relevant than it was the last time. Right. I mean, isn't that the goal? Yeah, I like I I think that's a good example. I and I was gonna go there too with like so there's what's happening now on the edge, you need you need a decision in like a millisecond or two, right? And you need to decide what cohort are they in, where do they live, what should we show them, how should we influence them to get

drive towards the outcome we we want. And then to your point, like at some point we need to decide, okay, what's your customer lifetime value? What have they done in the past? Do we know them? Have they been a customer before? Like what what's our overall strategy for interacting with them next time they come here? Or or do we want to maybe they came to our website, we know who they are, do we want to maybe, like depending industry, do we want to have someone call them? Or do you want to have someone if it's if it's pharma, do we want to have do we want to have a a you know one of our reps actually go to their office tomorrow?

Right. Because they because they showed up. I think I think putting those two together is like I like that example. Well and it giv and it gives some time for that data enrichment process to go back to the data warehouse to say, hey, do we know is this an existing customer? So I don't go to Zappos, buy boots for my wife for Christmas, and then for and then for fifteen months have me being targeted as a female boots. Yeah, those damn female boots. They think I'm single in Manhattan and thirty two years old, right?

It it's it's remarkable. I'm like there you know and I know what's going on behind the scenes. So we used to talk about this, but that same retargeting problem continues to happen. They're not connecting what would be the warehouse data with the actual shop shopping cart commerce data. Right. Right. And and you know, make the connection and correct it if you if you know that I am I was a buyer on behalf of someone else, for example. Right. But I think that's why there is no one place that decisioning is going to happen.

Miles (31:49.206)

And I think I think we're gonna continue to have let's call it composable decisioning. I'm not just yeah. we have a new term. Yeah, we're gonna hear in single noise, folks. Yeah, yeah, yeah. Headless decision. Headless decision modular decisioning, right? But the flexibility of you're gonna have to make different types of and and realizing there are different types of decisions. Are they strategy decisions? Are they you know big broad cohort decisions? Are they in moment in session decisions? You know, there's gonna be

AI is gonna continue to disrupt and change that. But there are going to be organizations that have the ability to pull out that data together and then there's those who just they don't have the resources. And so they have to pick a lane to to perfect for now and and and go with it. And so and some of them don't need all that information, you know. And so we talk about it live, I almost see like an arc of perishable moments where you're just using the the most data you have available at that time. So if they don't buy something immediately and then you can enrich from the lake house, let's say.

What can you then come back five minutes later, 15 minutes? Obviously, gonna these windows will depend based on the products that we're talking about. But like I'm still guilty of like putting things in a card and abandoning on purpose, hoping that I get the promo code. Like really? Yeah, I do this. I don't if I've ever strategically done that. right? Really? But they will enrich my data and they're like, we know what this guy does. Yeah. Like we'll at best give them Yeah, I'm on a list. But like some still have this older decisioning.

And I'll just wait. He's not a discount shopper. He's he's he's gaming the system here. So sometimes it's like one or two days later, but I'll get the I'll get the 10, 15% off or whatever it is. But like I will consistently do that. Yeah. Yeah, you get like an auto email report. It's like sorry, sucker. We know what you're doing. That is that is funny. It's too funny. One thing I want to talk about a little bit as well, maybe a slight pivot is like the how

AI is changing the marketing organization. And this is something I'd love to get both your opinions on. So I wrote this article that we published in Signal Noise a few weeks ago called The Rise of the Marketing Engineer. Mm-hmm. Right, which Scott Brinker said Scott Brinker, who really liked it. Marketing technologists are more important than ever. Like, and I think, but

Miles (34:15.138)

I guess part of our argument was the marketing technologist is becoming the marketing engineer. And like the there's a role for that in the marketing org's gonna look very, very different because of the headless trend, because of use of agents instead, and because of

AI being able to do a lot of things marketers have always done in platform. And I I find I'm already spending less I'm just not that's not gonna be in platform make messing out with campaigns in in a few years, but I'm spending less time doing certain things and I'm doing them. So what are you seeing as as like you work with lots of different marketing organizations? Like how are they using Telium, working with like how has that changed? What are you seeing? I yeah, I think I think there's gonna be two strata and in you know, we can debate on where that line draws. But you know, there's that marketing engineer, the person who is

wiring it a a lot of this up. And then I think there's going to be the AI marker, the person who is using AI as a as a force multiplier for their job. Where they're they're not the they're not setting up the MCP and the connections, but they know how to leverage them to get more out of their platform. Yeah. And and so but I think both of them are

We're we're moving into this like some of want a UI, some of want a command line. And it's the same it's the same problem that we've had for for years. A developer wants a command line 'cause he's comfortable there. you know, a marketer likes the UI because then they can spot check and they can take a screenshot and save it. You know what mean? And I think it just comes down to flexibility. you know, we have the ability for somebody to

Go into Claude and say, B build me an audience based on X, Y, and Z, and Claude will go to Telium and build set audience. And then you can still go in and look at the audience, make sure it built it right, approve it, push it out. So you can still spot check it if you're comfortable doing that. Yeah. If you still want to use it. Nobody wants to do SQL on the water.

Miles (36:15.098)

The team and everybody's got sort of access to the same data layer, the same data operating system, whatever you want call it, right? The it should give people more ability to derive insights from the same kind of same information. I mean, I think a lot of times we have information silos within brands and organizations where people aren't accessing the same data pools, the same insights. So so they're not able to really derive, you know,

like across the team the same types of outputs because because they're all operating under different assumptions with different data. Right. Right. Yeah. And I think it's just gonna some of it's gonna be, well, what is the what is that brand or that organization's purview or thought about AI, right? Because we may have we may be very AI forward, but not give you access to anything. Yeah. So you're still gonna have to go to Telium and and use the chat experience in there to build set audience because they're not gonna give you clawed access. Or

All the And there's the the the classic story of like the CIO that the board is telling. And and we actually heard this from you know firsthand and this was like a fortune five company. And the board had a mandate to roll out AI, and the CTO and or CIO had

you know, a huge number of projects that he or she was working on, and just didn't have the bandwidth to kind of fully transform the organization in terms of how they were really gonna use it, use this, because he knew that, like he knew what a real AI transformation took from a people perspective, from a tech perspective. And he was just like, can't do it, but I gotta make the board happy. Threw up his arms and just installed Copilot. Yeah. And then and then people asked the most basic question of Copilot, speaking of like data congruence, like.

Tell me the lifetime value of X or tell me the revenue we had in quarter Y, right? And they can't get the basic questions answered, right? And so I'm like, how are you empowering your people with like data if if you're not really creating connected systems? Right. Right? And I'm not saying everybody gets access to everything because there's like a lot of proprietary, important data that goes into either financial models or HR, ERP models that can't be used by anybody in the org, right? They can't. But still, I mean th th that that's not a solution and that's not a transformation.

Miles (38:21.816)

The organization. That's just making the board happy with something that looks like AI. I think I like the the marketing engineer role, like wiring everything up. I think the other piece that we see a lot of marketing teams evolve quite a bit is this more like data practitioner marketer. Yep. So instead of having to submit a ticket to a data scientist to ask the question, they can just do natural language. You're really hiring people that are know how to ask the right questions and then decision off that, right? So

I think that's where that role's changing a lot. Some of these folks, I think they can now just ask those questions. They knew how to ask the right questions. They're able to just move faster because they don't have to rely on a data scientist. They literally have a data scientist at their fingertips. Yep. And so it's like, how are you gonna use I I have I use genie all day long. And sometimes now I'm like, what did I do before that? I don't know. How did I work before? How did I do anything? Like, yeah. and so

I think it's hiring the right folks, the interview process is gonna be very different, right? How are you leveraging these tools to be better at your job and make better? Right. Or a team, right? So that people are using there's rules of engagement, there's there's best practices in terms of how do you nudge or like to to get the right answer out of something that might have come, you know, how do how do you interrogate the answers? And then where do you where's the starting point, right? Oftentimes, you it's like

People start with with the AI doing something for them when in fact they should have come up with the idea themselves, written down on paper, whether it's strategy or a creative plan or a brief or whatever, and then interrogated it with AI, right? If it's not like a data type analysis, right? Like where does human edge you know, judgment begin and AI work and, you know, and and you know, what's the line between those things?

think it's critical. Yeah, I think that what you're describing is more collaboration with yourself almost to come up with a strategy, right? And I'll find myself, you know, missing the first ten minutes of a call because I'm like deep in the I'm like, whoa, whoa, I got a call like had a call start 10 minutes ago. Yeah. And it's almost embarrassed. I'm sure a lot of people do that today, right? Because you just get so deep in your own head and you're like, like the strategy's now starting to take shape because you're kind of iterating with yourself and that's a good point. Like people don't realize this too who aren't using AI a lot. Like it it

Miles (40:38.016)

It does many of the tasks you were doing that required like mindlessness. Yes. But the work you do in order to use AI has to be incredibly focused. Yes. Yeah. Right. So you have to really look at these things. Yeah, you you'll let things run, but then after you get the output, you have to really dig in and look at it and be very discerning and very precise. And I and I And interrogate the hell out of it, right? To get to the right

'Cause oftentimes you're like, This isn't the right answer and you interrogate it until it you nudge it. Yeah, much more productive, but I still need to in order to get tasks done to get them from let's say if a if a if AI is middle to middle, like the beginning and end is probably the g gonna be you, right? The beginning and end have to be very intensely focused in order to give it the right inputs and then take the outputs and then actually maybe go for another round through or do whatever you're gonna do. It takes a lot of concentration. So even though I'm much more productive, I find that the biggest bottleneck is still me. Yeah.

Totally. Like, because yeah, I could it can just run, it can go fast, but I still need to. So I find I stuff piles up and I do what you just described. I'll be sitting there and I'll be like, a call just started five minutes ago. I I always do this, right? I'm terrible at that. Yeah. Yeah. You're giving it more context at which to make a better decision, right? But don't take that first answer as the answer. And don't, I think, is it we you we've talked about this a couple of times today, is like don't take that output, that raw output.

And then send it off to a team member to say, hey, look at this five page analysis that I just did that you largely didn't do. Yeah, people can see through that today. Yeah, and then you're just you're just basically it's like a diffusion of responsibility. You're like you're not doing the hard judgment work, but you're passing it on to someone else and you're wasting their time. I will tell someone if I'm gonna do that, I will tell you, listen, I'm busy.

I ran this through AI. Here's the output. Like it's gonna be rough. Like don't like I did not write this. I'll tell you, Don't burden other people. I didn't write it. Like I'm giving this to you to help you go faster, but this still needs a thorough review and your analysis. And that's all it is is an input to whatever you're doing. I'm I'm very clear when I do that. I and I think that's where the

Miles (42:37.9)

the next shift. I mean, you know, a couple of months ago, you know, the hot trend on LinkedIn everybody was, you know, sharing their skills once, you know, the clicking, yeah, yeah, here's my nausea. Here's nine hundred skills that are that you can replace every marketer in your organization or whatever. You know? And the thing is is I found like I have skills that I like I would love to share with my other team, but they're almost like employees of yourself where like

I know how to manage them. I know how to give them the help your your other employees. And they're not gonna necessarily help no, like co-work or not. It's not it's all Exactly. I I think that's the next like iteration of skills that I think we're gonna start to see more of, which is this idea of like, well, what are ground rules that are universal to a marketing team or to your marketing team? Yeah. And then what are the Rio specific or Brett specific kind of nuances that are in there? And and and I think that's the part that

having having you know all of the data in in in data and giving genie and and having that context like yeah I could we could all use the same prompt but we're we might not want all the same output. And I think so that's that's the layer that is still undetermined. I think it's an important layer too because that is sort of the collective capability of a team. Yeah. Right? That oftentimes gets broken between meetings, communication silos.

Data incongruence, data silos, right? There's all this stuff where people are just, even though they're on the same team, there there's not a multiplication effect. Yeah. Right, in terms of of their ability to say we're not we're more than the sum of our parts. I think AI certainly helps with that with that. What's the two worst things I've seen with AI? Like, or can help with that. I think the two things that bother me the most is one would be just the person who just the LinkedIn like if not this but this, like boom and bullets. It's just

Just come on, you couldn't take five minutes to edit the damn thing. That that drives me crazy. I still see a ton of that, right? and it's so obvious the person like if it if the it's obvious AI wrote it, you shouldn't post it like that. Yeah. Totally. And I think the other thing is I've had noticed this a few times where like there's certain people who get so addicted to AI when they're working that everything they do is in is like Yeah, you're almost like where's the voice of the human being? What are you like what value are you adding, right? If all you're giving me is just

Miles (44:53.804)

regurgitated AI output every time with no thought and in any argument they but it's my AI I've trained it in my with my with my skills. Okay, big deal. But it's still like I can still tell it's like it's still it's still AI slop what you're giving me. I don't care if you had five hundred skill documents you're still giving me something that clearly hasn't like had a human actually review it and give their input on it and like massage it so like another human could actually use it intelligently. I j I drives me crazy. But I think that'll be that big function of that.

AI engineer or that marketing engineer type who will be helping create that layer, try to figure out how to standardize certain tools, how to share, you know. Yeah. That it's gonna be not just, hey, how do I do better personalization or how do I wire up these systems, which is what the marketing technologist was. Now the AI technologist is gonna be more like

How we skill share? How do we exactly like how do you build IP with at the team level that that it that is sort of more than the sum of company level, yeah. It's all of the learning stuff. You're still gonna have your individualized sort of, you know, markdown files and your in your in your individualized kind of treatment of AI. But how do you actually do it where it's gonna compound the effec effects of the team and then at at i even a bigger level or higher level, it's like at the brand level.

They're building a ton of context and a ton of institutional knowledge, right? Which it and this is the stuff that people keep talking about is is kind of theory, but it's kind of critical. If you're gonna go through a transformation where you're truly leveraging this tool to capture and utilize AI effectively as an organization, right? Like you've got to you've gotta create a central learning hub, like a like a knowledge graph.

Right. And and that's gotta be unique to your organization. That's where you're different than Unilever is different than PNG, right? McDonald's is different than Burger King, right? And how do you get people feeding into that so that it becomes this kind of a super intelligence supporting the brand in a way. Right? That is that is brand IP. We're not there yet. I mean the I think the use cases are radically more simple than that within Teams. Yeah, right. Just going back to it's almost I what I think is getting lost in a lot of this stuff through AI is actually the

Miles (46:57.326)

purity of what that brand represents and I've yet to see brands actually create like their own branding MCP that everything has to run through so it keeps the voice of that brand. Because you start to see just if you look at all the decks that people are presenting, like I'm like, these are all kind of like looking the same now. Like

Regardless of what the product or the literature is. The head is the same, the writing's the same. Yeah, totally. Yeah, yeah, totally, totally. A lot of use of purple. There's a lot of purple in a lot of the the design styles. No, and so like that's the human or authentic thing that I think is starting to get lost. Yeah. And if you just look at the new the younger generations, they can like sniff that out much better.

And you see these surveys where they're like, what's really important to you? And they're like authentication of this brand and like and they're like not AI. Yeah, exactly. They're it's it's gonna be like there's hope for us, you know, with the youth. Yeah, yeah, yeah. Yeah, there definitely there's there's definitely a a growing distaste and like a a like a reaction and not not a positive one that I think that we're seeing you know to AI. And I I saw the study is interesting too the other day who was saying like who like which

Who's afraid of like AI destroying the world? And it was it was it was really different among age groups. I thought that was really that's really wide. It looks like the boomers and the millennials agree that it's gonna destroy humanity. Millennials or Generation Z? No, Generation Z actually th was more like Gen X. I thought it was interesting. interesting. And then the theory was the hypothesis, but G Gen X raised Gen Z. What? Gen X raised Gen Z and and Boomers raised millennials. Yeah. Right. So it was like, okay, so there's these shared values? I thought that was really interesting. Yeah. I mean, yeah, that was just one survey.

See never know. But one other thing I wanted to talk about before we go too was the and this is this has been talked about a lot, but I I I think it is worth covering a bit here is the you know SAS pocalypse, right? Or I I'm calling it SAS pocalypse never, right? That thoughts on, I mean that was a big thing maybe seven, eight months ago, and I and I think we we've clearly seen my opinions. I love your before I

Miles (49:00.11)

Tell you every what I think. I'd love to hear your opinions on what are you seeing out there? Like was it was it something was it a real thing companies should be concerned about or was it just fear mongering? Fear mongering. I mean, I I saw I saw a quote and I've I've stolen it and used it. I don't even know who it was from, but it was like we've been able to bake bread for many years, yet we still go to the store and buy bread. Right. And and I like like there's a there's a point in time where it's like, I'm just I you know,

I yes, I know like I use granola as a note taker on, you know, for for transcripts and keeps my ADHD in check 'cause sometimes I just I'm like, wait, what Yeah, yeah, yeah. Your your mind is wandered during that conversation. And yes, I could go vibe code my own and use my own transcription tool. It's ten bucks a month. Like there's a point, you know. What I do think vibe code your tr transcription tool, you could a theory, right? You absolutely could. I actually I started to one week and just to

It was actually just to play with it, not because I actually wanted to like have my own tran it like I was like, I know how that works, I know what it does. 'Cause it's some only s something someone would do if they're a complete nerd and they want to do it not because it's actually practical. Right, exactly. And that's and that's that's how you know how it played out. But I think there's I think there's a messy middle where people will use they may have bought a vendor for one specific thing that doesn't drive a lot of value, but it's it kind of important and

I think those type of solutions will go away because AI can be that integration layer, right? Or it can be that it it can do that one specific thing now via an MCP, right? And so I I do think there will be consolidation and and and But there always is consolidation. But there always is right. Whether it's gonna be AI or something, there always will be. I I think more so what'll happen is I think whether we call it headless, whatever, I think there will be a lot less

UI emphasis going forward. I think it'll be I wanna use my you know, we've already saw it in in apps and things. I want my device on my way. Don't make me use, don't force me to use your way or your paradigm. Yeah, yeah. Which which I think there's an argument in most of the research that I've seen is that the per seat licensing business model of SAS is what's really fundamentally gonna go away. Sure. Right, right, where it's really gonna be use case by use case, capability. It's almost like

Miles (51:21.602)

You're you're acquiring a capability that you're gonna plug into your data operating system, your data operating layer, right? Versus a monolithic software that you need to pay a person per license seat to. I think that's the model that is largely gonna be. That's not going away. People are not gonna vibe code a CRM or not gonna vibe code like

Tag management or any of these things. It's just it's not gonna happen. Yeah, it's been battle tested for too long. It's been tried and tested in the market for too long. Some of them are such business critical. I think the system of record model is gonna change a lot though. Folks are gonna go after and say, well, what actually sits behind a CRM? Why don't we just be really good at that? Other folks other folks can create apps as a skin on top of that. Sure. The reality is you people are pulling like the the or uncovering this great thing that's now on parody that day.

And then the questions are like, well, who's gonna maintain this? Yeah. And if this needs to evolve with our business, who's gonna do all that? And they're like, well, it looks really good today. And it's like, that's that's all that's why people buy and license these things, because it's actually responsible for someone else in case it goes down. Like it's funny. So I like the other day, so we took all our p all of our podcasts, we have all the transcripts from them, right? You it's like a hundred long form podcasts and all the short form ones. I mean, it was like hundreds of episodes, like thousands of hours, right?

And I took all the transcripts and I vibe coded this database to store them all in this front end where you can like query them. And it was it was pretty cool. And you know, it was it was amazing that I could do this in like 30 minutes, right? Like it this would have been this would have taken a you know, an HTML person to do the front end, a database person to do the back end, and I would have had to get a cloud like a cloud account. I mean, yeah, you know, store it all. It just would have taken like weeks to do this, like

minor thing that it's pretty cool, like we can query and find quotes and look for things. So I stuff like that, but it you know that's not like something you use for an enterprise. It's just something like a couple of guys at the podcast are doing, right? I mean I think things like that are incredible and fun and and but but yeah it's not this is not something you could give to like a a big company and say, hey, can you use this for your Right it's not gonna have SOC two certificated. I mean yeah think about all the the the security questionnaires we've all had to fill out or been part I you won't have that. But I think

Miles (53:42.188)

I think one thing that's really interesting, a and and really I think is compelling for a lot of organizations is when it comes to like data exploration or asking questions, being able to quickly spin up an app that may be a transient app, just so I can kind of solve this one problem today and then throw it away. Yeah. Like the idea of disposable apps I think is actually going to grow.

Yeah. we're you know, we're doing an event and I wanna be able to send my SDRs, you know, all this information all in one location, and then when it's done, dumb it's gone. Yeah, and Scott Brinker talks about that with his his he did a re he did research with Databricks. I think that's the report that I'm referring to, right? Where he talked about obviously the growth of sort of the Lumiscape of Martec from two hundred and fifty to fifteen thousand plus. But then he said there's gonna be this sort of hyper long tail with the democratization of sort of front end developers and and full stack developers and people that that can

vibe code stuff on their own, they're gonna build a lot of use case specific apps that solve a specific problem at a specific point in time and that's gonna explode from a Or you take a lovable and you say, okay, well lovable now technically fits in every single category in that Martech thing because you can probably build something from all of those. And maybe you don't want to go, you know, yeah, pay your per seat license to, you know during COVID I was working in CMS and we had some people come to us and we're like, well we don't we need to get information out to our, you know

people working in stores and whatnot and all this information, but we don't want to pay per seat for SharePoint for them. Yeah. 'Cause it's five to twenty dollars per circ, you know, for a bus boy in the restaurant. Like they're like, Can we build something? Like, yeah. Like, you know, and I was I was I was at Aqui at the and this was Drupal and it that's just a database it's just a relational database with content. One hundred percent. Yeah. So I think that's no you were at Aqua. My my brother was one of the first nine, first ten at Aqui. I was not that long ago, yeah. Up in up in Boston, right, with the founder and

They ran into kind of the composability problem because their CMS. while modular wasn't fully composable and adaptable to this kind of future world of like we only need this specific use case. What was the what was the C D P they acquired? Remember that? It was it was Agile Agile one, yeah. Agile one I I wrote the first demo and all of the information for that. So yeah. Yeah, they were one of the first C D Ps I did work with. This is way back. Way back. This is must be like twelve Yeah.

Miles (55:57.592)

fourteen years ago. I even did identity work when I was an experiment with them, just to we all touched. Yeah. It's a it's everyone's like like maximum one or two degrees of separation away from each other in like Martech actually. I think so, yes. 100%. So it it does seem interesting. Like there's gonna be I a bifurcation when we look at like the trend of of of whether or not companies are to go and develop their own you know apps or use case you know, like like CRM, right? Like I've seen CRM stuff being built

and marketing automation for startups like Alex Marut Marutka right at Craftsman Squad. Right. He his COO just spent a weekend and she said, we're not gonna pay for Salesforce or Hubspot. I'm gonna I'm just gonna build this for the specific case use case that we have, right?

And we're gonna run this in our Right. And so she was a full stack developer, right? And and she developed that. So as as a startup, I've seen that and and we've heard that a couple of times from startup leaders and even relatively large startups that are building a lot of these underlying capabilities because they don't want to go and pay per se license with the Salesforce. But when you go to the enterprise level.

And you're dealing with ERP systems and massive accounting. Put the blame on somebody else. Yeah. Who's gonna maintain it too? You need a higher degree of governance and control and in right and in yeah, so I think this the the change there is gonna be slow or nonexistent. But within startups it seems and there's six six million new startups in the US in twenty twenty six. Well saying the difference like

Like a few years ago, startups wouldn't have gotten s Salesforce, they just would have used Excel, right? Like that would have been their like opportunity management as everybody goes. And then you Salesforce, like once you reach a certain point, you can afford it. But now they can just vibe code it until they're ready to actually, okay, now we need something with multiple locations and something that like, you know, that that passes like basic security and privacy things and you know, and whatever. Like like so I I think that's we're gonna see that, right? But at a certain point, so yeah, this the SaaSpocalypse and you look at

Miles (57:56.418)

All of the stocks of the SaaS companies have all completely rebounded. I I think it HubSpot maybe hasn't, right? but all all the others have Salesforce almost back to where they were. So you know it's it's an interesting time. It really is. I mean and you start to see like how how effectively have companies, big brands, medium sized, mid market implemented AI solutions.

that are that are creating value and driving impact at the organizational level and it's and it's a bit harder to point to right and so I think there was a lot of this fear mongering that that it was just AI was just gonna eat up this entire tech ecosystem. And some of these tech players are battle hardened and have proven use cases and and this is not their first pivot. People forget Salesforce has done multiple pivots like they went to the cloud, right? Like like all of that. Like before like they started in 2000, right? People were like that was

In tech that was eons ago. That was before mobile. Well, how many organizations are still running an AS four hundred somewhere in their in their you know, I many of them are. Many of them are. Exactly. So like to say that well cloud takes care of i nothing takes care of anything. Like there's just you know, there's an evolution, but I this idea of like s blank is dead is is like the most that's more that's more overused than composable. So Yeah, yeah, yeah. Totally. They've lean talk about pivots and leaning in, that's like Cloud Force. Like they are like, Hey, like best of both worlds, right? Like

They're very smart. Yeah, and proven for they're durable. Yeah, they're they're there's I mean, they're they're more of a s like sales and marketing organization than a tech organization in a lot of ways this year. But again then again, that's why they're smart. They're good at it, right? So I think that is not surprising that they've successfully it seems like they made this l latest pivot. Yeah. Yeah. Yeah. But any anybody that's worked with like Salesforce Instance.

Integration, you know, you do acquisitions. I've been in a lot of companies that have acquired other companies and you've got multiple instances of Salesforce, no source of truth for customer records, right? And then that integration process is an absolute nightmare. And I've seen it take anywhere between one to three, four years. Well that's right. That's why SIs have Yeah.

Miles (59:52.526)

I do pretty well. I mean I've done a lot of that work in my career. Right. And it's like you you know, speaking of like time to insight, time to action. I mean, you know, they spend so much time data wrangling, you hope that that process can be a lot easier, right? Especially with with more headless systems. Right. Yeah. It's interesting too, like looking at SIs, like the

That's that's a sector that's changing a lot. Right. I'll I I will say that. It's it's changing a lot. And I was I was telling I was talking to Anish, you from the Art the Art of Commerce podcast he interints too. There's an they've been like hanging out with here f doing some onsite filming and I was telling him how like I'm like I've been doing work now with clients where over one weekend I'll crank through what would have taken a a team of three like weeks. Right? And just

The ability to synthesize information so quickly and then help think it through, organize it, and then, you know, and then take that and translate it into output that actually makes sense, which is really what a lot of consulting is. No, I a lot of it's is it is like the the best part of it's the thinking part, like the okay, what does this really mean? And AI is not great at that, but it's good at getting you to that point and presenting it to you so you can then take that.

And all of the busy work, whether it's the note-taking or the analysis or the breaking into parts or organizing, it does that really well. Does that better than better than most junior people and certainly better than team of junior people could do quicker. So so it is changing it and it's I almost think like the the expression I've heard is the the is going from a triangle or pyramid model into like an obelisk or you know pillar model where it's just

you can't justify you don't need as many junior people relative to senior people. So it does make the cost model a little different. And we're actually switching to more like and I'm telling like my you know my teams, let's just like I don't even want to build client sign materials anymore. I don't want them to question how many hours did you really work on this? How many hours like what did that BA do? I don't believe that BA actually had that much, you know, had that much work, which they 'cause maybe they didn't, but let's let's agree on the outcome. Yeah.

Miles (01:02:02.946)

Let's agree on when you need it by and let's agree on how I'm gonna deliver it and then we'll work with you to make that happen. Yeah. I'm I'm finding that's more successful. The value of the output, whether it took you ten hours or a hundred hours, is kind of irrelevant because I need X output, you know. Yeah. And and price based on that outcome, not the amount of of hours, man or artificial hours that are put into that process, right? Yeah, because you know stuff that, you know, y you're as an SI I I I was

before I was in the tech world, I was in the partner world and S I you know, and it's like, well, the value of this is here. I can actually get it done here 'cause I've done it ten times. Yeah. But I'm gonna either have to burn a lot of hours because otherwise I'm cutting my own margin by getting it done fast and making the client happy. Yeah. But then I have a utilization problem. So I yeah, I think outcome based I think consumption, I you know, you're gonna s continue to see the the that idea of, you know, like pay for what you use, pay for what you get. Yep.

The risk is obviously commoditizing that and just driving you know, everything becomes like a well, if you'll do it for thirty thousand, they'll do it for twenty eight. But that's what comes down to like value relationships and expertise, right? 'Cause if you have real expertise, like it doesn't like it's not it's not apples to you know, you're not comparing prices of apples to oranges, print apples to apples and you can see like w where the value is. But also think too something that is helping the industry is despite challenges to the the overall like structure of these organizations.

Which is start it's starting to show, right? Because like there's fewer junior people, that means you're hiring fewer junior people, that means, you know, like your fewer resources that have higher margins on projects. But like that aside, there's more demand for work when you look at there's a reason why anthropic and open AI and and all all of these frontier models and any r a lot all these AI companies, even all AI companies are creating Palantir, you know, the Palantir model of forward deployed engineers. Yep. Because in order to actually operationalize AI, you need experts who can actually

Take it and adapt it to the industry and implement it in and implement use cases that are actually meaningful and effective. Right. Yeah, yeah. Yeah. AI is not a strategy, right? AI is equivalent to is as much a strategy as electricity, right? And and the and you know, or computers, right? Yeah, or computer strategy. We've got an internet tragedy, right? We've got a telecommunic, you know. But with forward-deployed engineers, you're customizing a specific use case.

Miles (01:04:25.174)

At a specific point in time for a company's like current reality, right? And that's that's where the magic happens. That's human judgment. And you're in your automating what you need to automate. But the point is is that that AI on it on its own is just the it's like a median solution that everybody has access to. But I think that's similar to what your Databricks your your model's always been. Let's find great partners, right? Whether they're SIs or ISVs, and let's you know let's they can actually take

the power of this and adapt it to these specific needs and a and actually make an actionable and build right native applications within the within the ecosystem. I mean we have super aggressive goals and from the top down we say we will not get to these goals without our partner ecosystem. Yep. SIs and ISVs being the the key partners for that. Yep. and so that's obviously a huge part of why I'm here. It's all

partner ecosystem rich environment at Adweek. yeah. And they're doing things on the edge that Databricks doesn't do. And the power of us together is how we're accomplishing value going back to what you're saying, delivering for our shared customers. Yep. Yeah. You guys will provide like a stable governed system where the data's flowing. Yes. Right. But it's also protected and in and sort of pseudonymized, right? So you're not dealing with the data movement issues, right? And you just build on top of that ecosystem, right? through through partners, right? That's a good model.

Yeah. Makes a lot of sense. It is a good model. I yeah, when the snowflakes and the databricks came out, it was like, this is interesting. yeah. I mean, you know, like I was a transunion. We were building native identity resolution apps within the snowflake environment, like when when I was when I was over there. So it's certainly a way that yeah, it's like you centralize your data for one. Yeah. even if it's if it's relatively federated, you still have a layer that that communicates with all of your data, right? And then plug in your specific applications for per use case.

Only waiting for you gu you guys to go public. I know. Actually, yeah. No, it's funny. So six years ago I remember my first ad week with Snowflake at the time, and it was very much like Databricks, hmm, what's that? What do you guys do? Yeah. And then just to see like obviously our presence both organizations here is much bigger.

Miles (01:06:26.048)

It's less question of who Databricks is. There's no question anymore. Yeah. Well, it's And you guys get along. I I thought you and Aaron Fox were there. Last night we're going to get into a like a wrestling match. You know, but I was like, no, it was a great conversation about some of this data match. I know, I know. But about data architecture and some of how this is being managed on the client side, on the partner side. And it's it's collectively, I think we're still trying to solve some of the same problems. and it's what's really changed what I love now is it's like, hey, we know our organization uses Databricks.

Can you help me find out who in my company is the champion? Cause I want to connect with them. So like I'm helping like the marketers and the advertisers connect with their technical teams, and that's still the biggest gap. Yeah. Right. The personas don't talk to each other within the own organization. It's it's less of a data silo problem and it's a persona silo problem. Yeah, the databricks probably mostly just sells to IT. Yeah. Yeah. But in a but I think probably your growth area is advertising marketing, right? Yeah.

In in each, there's different same in security or same in office of the CFO. We just don't talk to the we historically have not. Yeah. When we walk in together with like a telium, then it's like a joint conversation. We're able to move much faster. They already worked with the CMO or they had a meeting. Yeah, yes. Yeah. And and Rashad was said something that I thought was really good, not to overquote Rashad Tobacco Wallow, but it's not to It's easy to keep quote. Right. He said he said AI is gonna be like the hydrochloric acid to organizational silos, right? It's gonna burn away.

organizational silos. But I the underpinning of all this is you gotta have I keep saying that word data congruence because I think it's a good one, but you've gotta have shared data and and to create shared insight. Right. And sh and and then obviously you've got to align incentives, right? That's gonna be critical in terms of aligning different teams. But it's it's because of information deserts that you oftentimes have silos, right? People, you know, in s incentive like like information deserts and then incentive misalignment. And those two things

create power imbalances, they create incentive imbalances, and organi how do you expect organizations to change? People need to change, right? The organization itself isn't gonna change. It's the people within the organization that need to force that to happen, right? And that's AI is kind of gonna accelerate that process, right? And and we're seeing a lot of that shift, right? A lot of our customer profile is becoming more technical, right? Yeah, you'd say marketer, but it's it's there's a lot of data data engineers and and and you know, it go into data A and

Miles (01:08:49.688)

Databricks AI Summit this year, you know, that profile is a lot more business. You know, there so there's that that that meet in the middle kind of thing. And we should all we should all adopt like hyphenated job titles. Job titles, you know. Right? Like we you've gotta be like continuously learning and continu and picking up new skills, right? 'Cause if you 'cause if you stay static, right, especially in today's world, whether whatever function you are, right, it's gonna be it's gonna spell

a a negative future for you. Yeah, my job title is gonna be shit poster. Professional shit poster, real log anchor. Right? There we go. Well this has been phenomenal. Thanks guys for joining us. So thanks TLM for being a sponsor of the show. Thanks Databricks for showing up. you know both great conversation. We knew this was gonna be fun. we're running you know towards the end of our our independent content before the event starts, right? The podcast we've got one more. but everybody that joined us and made it this far

enjoy all of our content from live at s it at odd week two twenty twenty six. It's a ton of fun and we'll see you next time. Thank you so much.