Best of LinkedIn: MarTech Insights CW 28/ 29

00:00:00: provided by Thomas Allgaier and Frennus, based on the most relevant LinkedIn posts about MarTech in calendar weeks twenty-eight and twenty nine.

00:00:06: Frenness is a B to B market research company that supports enterprise marketing teams in unlocking the full potential of their customer data with the help of AI.

00:00:15: you can find more info.

00:00:18: Our mission for you today is to basically distill all those top completely no fluff Martek trends that are, you know bubbling up right now.

00:00:26: For BDB marketing leaders

00:00:28: exactly we're mapping out three massive architectural shifts for this deep dive.

00:00:33: We've got the whole composable CDP debate AI agents becoming actual core infrastructure and this kind of quiet sneaky rebranding of marketing operations.

00:00:43: Right, because I mean the overarching thread here is decentralization.

00:00:46: for like a decade The standard playbook was just buy an all-in one customer data platform...the

00:00:51: traditional CDP.

00:00:52: Yeah!

00:00:52: The pitch was super alluring right?

00:00:53: Just pump every piece of telemetry Every CRM record into this One centralized box

00:00:58: and it magically spits out a unified customer profile.

00:01:01: Exactly but that era Is rapidly closing.

00:01:04: It really is.

00:01:05: Yeah, I mean Adria Garcia Castellon highlighted this striking stat recently the percentage of companies using a CDP as the central piece Of their stack dropped from twenty seven percent to seventeen percent in just A single year.

00:01:18: wow

00:01:19: that Is a massive drop.

00:01:20: it's a huge architectural defection and at The exact same time reliance on the data warehouse grew To twenty four percent

00:01:28: which you know makes complete logical sense when you think about the friction of that legacy model.

00:01:33: Pumping everything into a package CDP usually meant building this secondary siloed copy of your data,

00:01:39: right?

00:01:39: Paying double for compute costs

00:01:41: exactly!

00:01:41: You're duplicating data that already lives in a cloud warehouse.

00:01:45: and Archie Morinkovitz actually observed huge industry signal here When Databricks launched their customer lake offering.

00:01:51: oh yeah he basically noted Databricks deleted the standalone CDP by just moving those capabilities directly into The Lake House where data already sits.

00:01:59: It's like

00:02:00: if I can use an analogy here, building this custom high-performance engine which is the Composable CDP.

00:02:05: but critical question isn't we forgetting fuel?

00:02:08: The identity resolution.

00:02:09: Yes.

00:02:11: Seger Deerex brought reality checkup warning that these composable architectures they only actually work If you've solved identity internally Which almost nobody has.

00:02:20: No, it's incredibly complex like connecting anonymous web traffic and B to be sales interactions back To a unified account without a common primary key.

00:02:31: that requires matching models most internal data teams just

00:02:34: can't build right.

00:02:35: which brings us to this massive signal shared by John Suarez Davis.

00:02:39: Oh,

00:02:40: this is wild!

00:02:41: Yeah the rumor that Hightouch offered publicist group up to one point two billion dollars for liveramps identity and onboarding assets.

00:02:47: What's

00:02:48: fascinating here?

00:02:49: Is that identity not-the data storage is clearly becoming The ultimate control.

00:02:53: point A one

00:02:53: point Two billion dollar price tag For identity infrastructure tells you exactly what a composable ecosystem is desperate

00:02:59: for.

00:02:59: Exactly they aren't just buying software features.

00:03:02: They're buying the definitive map That connects an anonymous IP address To specific enterprise buying committee.

00:03:08: Yeah, so while you know legacy players like Salesforce Data Three Sixty and Treasure AI are still sweeping up these IDC Marketscape leader titles

00:03:16: because they're deeply entrenched.

00:03:17: right but the actual infrastructure battle is moving downstream.

00:03:21: hmm So if The Warehouse wins and centralizes this identity data it creates a new question What Is Actually Executing Campaigns Against That Data?

00:03:30: Because It's Not Just Software Interfaces Anymore Its Autonomous AI Agents

00:03:35: And this is our second big thing.

00:03:36: Right, so we have to... completely reevaluate how we conceptualize AI.

00:03:40: We used to treat it like a highly efficient intern, you give an isolated task of human reviews the output draft

00:03:47: this email suggests subject line

00:03:49: exactly but Andrea Vigiani pointed out that recent acquisitions like Salesforce buying Finn and Moengage acquiring on pay-proof agents are moving from just being a feature To be in the architecture itself

00:04:01: as before.

00:04:02: we dig into the dark side at that And there is a dark site yet.

00:04:04: so want to quickly mention if your finding these architectural insights valuable, make sure you hit subscribe so that you catch all our bi-weekly deep dives.

00:04:12: Yeah definitely do that because the speed of this AI innovation is actually breaking the enterprise buying process.

00:04:19: Here's where it gets really interesting.

00:04:21: Tony Byrne shared this wild story.

00:04:24: ambitious sales engineers are literally using Claude to hallucinate fake software interfaces for vendor demos

00:04:31: Just completely fabricating a custom front end To win an RFP.

00:04:34: Yes backed by absolutely nothing.

00:04:37: The buyer asks for a complex B-to-B workflow and instead of admitting their platform can't do it, they use an LLM to generate a functional looking React UI.

00:04:46: That is terrifying!

00:04:47: You sign a multi year deal based on vaporware?

00:04:49: Exactly but as dangerous buying fake AI as deploying real ungoverned AI is even worse.

00:04:56: Yeah I Shoria Pandey shared a forester prediction that Ungoverned Ai will cost B-To-B companies ten billion dollars this year alone.

00:05:03: Wait Ten Billion Just

00:05:05: This Year.

00:05:06: That is a stark warning about what happens when adoption outpaces proof.

00:05:10: If an agent has right access to your CRM and it hallucinates the pricing strategy based on messy data,

00:05:16: It could autonomously email hundreds of accounts with massive discount.

00:05:19: Exactly!

00:05:20: The financial damage scales instantly because machines execute at velocity humans just can't catch.

00:05:25: So how do you as B-to-B team protect yourself?

00:05:29: Mark T Level laid out roadmap for this specifically around model context protocol or MCP

00:05:35: the secure bridge for the AI.

00:05:36: Right, he suggests that when connecting AI to marketing automation you have to start with a dedicated read-only API user.

00:05:44: You force the AI to prove its competence in a sandbox before ever.

00:05:48: let it write single piece of data back into system.

00:05:51: It's basically principle of least privilege applied an LLM.

00:05:55: But basic API limits only stop from breaking things.

00:05:58: To make it actually effective, Smith-Garai argues you need a horizontal semantic layer.

00:06:03: Right and ontology?

00:06:04: Yeah,

00:06:05: explicit business rules because without that ontology AI agents are literally just guessing context.

00:06:10: the semantic layer tells the AI what your completely siloed business terminology Actually means.

00:06:15: so it understands that say A high volume of angry support tickets overrides a positive marketing engagement score.

00:06:21: Exactly otherwise The agent might autonomously target a furious customer with an aggressive cross cell campaign.

00:06:28: Which brings us to our final theme, the people.

00:06:32: Because if the stack now requires building bespoke semantic ontologies and restricting API access for autonomous agents The humans running this can't just be software admins anymore.

00:06:44: No!

00:06:44: The whole role of marketing operations is being completely re-architected.

00:06:49: Gene Otz argued that marketing leadership is shifting from traditional commander who buys media.

00:06:55: To

00:06:55: the chief marketing architect?

00:06:57: Yes, The Architect who actually translates and connects the stack to the consumer.

00:07:02: And

00:07:02: we're seeing this shift in reality.

00:07:03: Corey Gabor shared his brilliant example of a three-person Marketing Ops team that trained an internal AI agent they named

00:07:10: Operator Okay I love their name Right.

00:07:12: It takes a Slack form autonomously builds an email in NAC gets it approved then syncs directly with customer.io

00:07:19: Completely removing human from manual handoff.

00:07:22: Entirely.

00:07:23: But wait, doesn't that make the human operator obsolete eventually?

00:07:25: Well no because The Human Operator's Job transitions entirely to governance and alignment.

00:07:30: Which ironically is part of what technology can never solve.

00:07:34: Odd Morton Sernsson interviewed twenty-six marketing ops leaders... ...and uncovered this deep truth….

00:07:41: …the technology was NEVER THE HARD PART!

00:07:42: It's ALWAYS a culture problem always.

00:07:46: Sean Simon observed most implementations fail just because People get emotionally attached to their old tools.

00:07:52: They know how to hide in them, right?

00:07:53: Oh absolutely!

00:07:54: Migrating into a perfectly governed composable stack forces operational transparency.

00:08:00: that makes people deeply uncomfortable.

00:08:02: Change management is human problem.

00:08:04: So if you're the one managing that change Ashley Langford offered this really practical career tip If your evaluating new rev ops or marketing ops role.

00:08:13: Ignore job description.

00:08:14: Look at reporting line authority.

00:08:16: Who

00:08:16: does hiring manager actually report too

00:08:18: Exactly?

00:08:19: That tells you more about your ability to actually get things done than any bullet point in the job spec.

00:08:23: Because historically, the recognition for operations is incredibly low.

00:08:27: Mike Rizzo calls it The Stage Hand Problem

00:08:29: Even with a fancy rebranding of GTM Engineer.

00:08:32: Yeah!

00:08:33: It doesn't matter.

00:08:33: You are invisible because you're good.

00:08:35: Just like A Stage Hand The audience only learns her name when a prop breaks on stage.

00:08:41: In Ops, people only learn your name when the pipeline breaks.

00:08:44: It's a brutal reality but treating your stack as a cohesive product that needs constant architectural iteration instead of just a bunch of vendor logins is defining challenge right now.

00:08:54: Absolutely!

00:08:55: The legacy monolithic systems are fracturing AI demanding totally new governance and whole foundation shifting.

00:09:02: Well, if you enjoyed this episode new episodes drop over two weeks.

00:09:05: Also check out our other editions on field marketing channel and partner marketing AI and B to be go-to market ABM in social selling.

00:09:13: And before we go I want to leave you with this final somewhat provocative thought to mull over.

00:09:18: it builds On a concept from anim around

00:09:20: okay?

00:09:20: What's here?

00:09:21: so in our endless quest to use these composable data layers an AI agents to perfectly optimize every single customer journey and eliminate all uncertainty, are we quietly optimizing away our own serendipity?

00:09:34: Right.

00:09:34: Because those lucky unplanned accidents often yield the quarter's absolute biggest wins!

00:09:40: So how do you engineer a system that is hyper-efficient but still leaves room for?

00:09:56: We'll see you in the next deep dive.