Best of LinkedIn: MarTech Insights CW 34/ 35

00:00:00: Provided by Thomas Allgaier and Frennus, based on the most relevant LinkedIn posts about Martek in calendar weeks thirty-four and thirty five.

00:00:07: 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.

00:00:17: you can find more info in the description.

00:00:19: so glad to be jumping into

00:00:20: this.

00:00:20: yeah same here.

00:00:22: Because, you know if you are a strategic B to be marketing professional.

00:00:25: listening to this You're probably feeling how incredibly noisy the landscape is right now.

00:00:30: Oh absolutely It's overwhelming

00:00:31: Right.

00:00:32: so our mission for this deep dive is just cut through all that fluff.

00:00:36: We're gonna synthesize The top martech trends That have been blowing up across linkedin recently.

00:00:41: yes

00:00:41: specifically looking at How ai Is completely restructuring the tech stack?

00:00:46: Uh the evolution of customer data platforms and most importantly what This actually means for you, the humans operating these systems?

00:00:53: Exactly.

00:00:54: Because for the longest time building a Mar-Tec stack felt like I don't know...building a

00:00:58: house?

00:00:58: Yeah that's good way to put it.

00:01:00: You buy this software and then in its designated room wire up And just stays there.

00:01:05: It is predictable!

00:01:06: Login, give an instruction, log out.

00:01:09: The software has historically been totally passive.

00:01:12: It literally sits there waiting for human operator to push button.

00:01:16: But looking at conversations over last couple of weeks foundational architecture is well, it's cracking.

00:01:23: It's like the house has started to rearrange its own furniture.

00:01:25: that

00:01:25: Is such a wild thought but it's true.

00:01:27: The biggest tectonic shift in the market right now is AI moving away from just being A chat interface yes?

00:01:33: It's no longer Just a window you log into.

00:01:36: You know.

00:01:37: ask for a clever subject line.

00:01:38: far From it.

00:01:39: it's becoming a headless autonomous agent That just acts across your entire infrastructure.

00:01:44: Which brings us to the absolute blockbuster launch of The Week, which is Claude Force.

00:01:49: Oh man yeah huge news!

00:01:51: Salesforce and Anthropic announced this massive partnership where Claude is now at the default AI across the whole Salesforce ecosystem.

00:02:00: but the crazy part for me isn't just integration it's architectural inversion right?

00:02:06: Salesforce basically turned itself into a plugin inside of Claude.

00:02:09: It

00:02:10: really fundamentally flips how we interact with enterprise software.

00:02:13: Like, you don't have to navigate through some giant labyrinth of CRM menus to use the AI anymore.

00:02:19: Thank goodness for that.

00:02:20: Right

00:02:20: The CRF is now basically acting as a passive data source and uh... A permissions layer just sitting quietly behind Claude.

00:02:28: And they launched.

00:02:28: it was something like thirty seven pre-built sales skills.

00:02:31: Meaning the AI is primary interface.

00:02:33: now, traditional software just engine room?

00:02:36: Exactly and that changes entire integration game.

00:02:39: Aaron Bird pointed out something really massive about this regarding model context protocol.

00:02:46: Yeah,

00:02:46: I saw that.

00:02:46: If you haven't been tracking MCP it basically renders traditional sauce integration catalogs completely obsolete

00:02:52: Totally obsolete!

00:02:53: Think about the old way we made tools talk to each other.

00:02:56: Oh... It was in the nineties.

00:02:57: You submit a roadmap request Wait six months for engineers To build a custom API bridge And then second one platform updates its code The whole bridge collapses.

00:03:06: Exactly incredibly brittle.

00:03:09: But MCP acts as a universal translator for these AI agents, right?

00:03:13: He used the example of mural.

00:03:15: you know that visual collaboration tool and inflection dot IO which is B to be marketing platform.

00:03:21: okay yeah

00:03:22: so both of them have an mcp server.

00:03:24: now That means an agent can just grab an email layout or marketer built in mural understand all its components hand it directly to inflection as a live executable asset.

00:03:35: Without either company ever having to build custom integration for the other.

00:03:39: Precisely, they just both speak this standardized schema so that AI reads their capabilities without any custom coded bridge.

00:03:46: It honestly sounds like magic.

00:03:47: But, okay.

00:03:48: Let's unpack this because if an AI agent is essentially a super smart digital intern.

00:03:53: we have a massive problem.

00:03:55: you wouldn't let a brand new intern email a million of your top clients on their very first day without giving them the company rule book right?

00:04:02: I would hope not.

00:04:03: So If We Just Let These Headless Agents Loose Via This Protocol They Could Go Completely Rogue.

00:04:09: Yeah and that Is The Exact Friction Point The Industry Is Hitting Right Now.

00:04:13: So I am.

00:04:14: Prakash Bajira Brought This Up

00:04:15: Brilliantly.

00:04:16: What Did He Say?

00:04:17: Well, everyone agrees autonomous agents need context to do their job safely.

00:04:24: But the problem is nobody actually agrees on what context means.

00:04:28: right because if you just dump a massive blob of raw customer data into an agent's prompt

00:04:34: You get complete garbage fragmented unusable garbage Because Context isn't Just knowing who The Customer Is.

00:04:41: yeah he broke it down Into Three distinct Layers Starting with platform context.

00:04:46: Okay,

00:04:47: this is the physical reality of what your stack can actually execute.

00:04:50: So I think channel throttles GDPR consent states frequency caps.

00:04:54: Oh wow.

00:04:55: so if the agent doesn't have platform contact it might write a brilliant highly personalized campaign that Technically just cannot be sent.

00:05:03: or worse

00:05:03: It violates a frequency cap and spams you're most valuable accounts like Five times in an hour.

00:05:08: Ouch!

00:05:08: Yeah, that's the classic intern trying to print five hundred brochures without having the printer password.

00:05:13: Exactly The intent is there but the capability is missing.

00:05:16: So what's the second layer?

00:05:18: The Second Layer Is Brand Context.

00:05:20: This is basically the governance layer Your tone of voice Legal disclaimers The claims you are absolutely not allowed To make about a competitor.

00:05:28: And I'm guessing That as a layer Almost nobody actually has formally written down In machine readable format.

00:05:34: yet

00:05:34: Spot on almost no one.

00:05:36: And then finally, you have the ML inference context.

00:05:39: Which is what?

00:05:40: That represents what your internal machine learning models currently predict about this specific user.

00:05:47: like are they high churn risk?

00:05:49: What's the statistical next best channel for them?

00:05:52: got it?

00:05:53: So if a company tries to take the easy route and just collapse all three of those distinct layers into one giant vector database for the AI,

00:06:01: The whole system goes stale at the speed-of-the-fastest moving layer.

00:06:04: Oh that makes total sense right

00:06:06: because engineering owns the platform rules marketing and legal owned the brand guardrails And those data models.

00:06:11: they change every single day based on user behavior.

00:06:14: Yeah They have to be managed separately but the agent needs to access them instantly and simultaneously.

00:06:20: And even if you do manage to nail all three layers, the AI still has to have a feedback loop.

00:06:26: Khan Smith made this crucial point about memory recently... Oh

00:06:29: yeah!

00:06:30: The learning loop?

00:06:30: Yeah because right now marketers are spending most of their week just on execution.

00:06:35: If an agent automates that execution but doesn't It's literally just making the exact same bad decisions.

00:06:42: Just way faster,

00:06:43: which helps no one right.

00:06:45: if a senior marketing manager looks at an AI recommendation and manually overrides it The system has to remember the underlying reason for that override.

00:06:53: Yeah That human override is infinitely more valuable than the original AI Recommendation.

00:06:58: If an agent senses decides and acts but never updates its parameters based on human correction?

00:07:05: It's just a liability.

00:07:06: totally what I think brings us perfectly to the actual foundation powering all this memory and context, which is data itself.

00:07:13: Yeah agents are only as smart of info they can query

00:07:15: Exactly!

00:07:16: And evolution we're seeing in customer data platforms right now.

00:07:19: CDPs just moving at breakneck speed

00:07:21: It really does.

00:07:22: it's a complete paradigm shift from passive data collection active autonomous decisioning.

00:07:29: Yeah, the latest Forrester B to C CDP wave for Q three.

00:07:33: twenty-twenty six just dropped.

00:07:34: yes and analysts like David Chan and Joseph Stanhope really broke down the massive realignments happening there.

00:07:41: The big takeaway is the meteoric rise of players Like amperity in high touch.

00:07:46: yeah That was huge to see.

00:07:47: meanwhile a lot of the legacy monolithic players are either dropping tiers or just falling off the report completely.

00:07:53: I think Imperity was actually the only vendor on the entire wave that earned a customer feedback halo.

00:07:59: Which indicates significantly above average customer satisfaction, yeah?

00:08:03: Yeah And Barry Latimer synthesized exactly the mechanism driving this shift

00:08:07: which is well.

00:08:08: The

00:08:08: industry isn't just slapping the word agentic onto legacy CDPs.

00:08:11: The real shift is moving toward the open data lake.

00:08:14: Oh okay because of the old paradigm you basically had to build an audience list package it up and physically copy that data into a proprietary vendor.

00:08:22: silo, right?

00:08:23: Exactly.

00:08:24: You pushed the list to its destination... Yeah!

00:08:26: ...and just left the email platform to figure out the experience And by the time this data arrived It was usually already stale.

00:08:33: The new paradigm leaves the data exactly where it lives in your central warehouse.

00:08:38: The MarTechStack sits on top of open lake and makes real-time decisions about highly specific experiences customers should get at those exact moments

00:08:47: In the business.

00:08:47: impact of that is just staggering.

00:08:49: I know Latimer pointed out, that Ulta Beauty saw a ninety-five percent cut in production time using movable ink.

00:08:54: Ninety five percent?

00:08:56: That's incredible!

00:08:57: And WOP saw at ten percent cross cell lift...in six weeks…using high touches AI decision and capabilities.

00:09:06: Because they didn't have to spend eighteen months begging IT to build a custom data pipeline.

00:09:12: The data was already sitting in the warehouse, marketing just pointed the AI at it and turned it on?

00:09:17: It

00:09:17: completely flips traditional resource allocation A deal.

00:09:21: Abbas noted that recent market forecasts showing data warehousing and AI are projected grow two-to-two and half times faster than traditional customer engagement platforms over.

00:09:33: Wow.

00:09:34: Yeah, the budget is moving away from the orchestration layer that just sends the message and it's pouring into the intelligence layer that decides what the message should actually be.

00:09:44: but this introduces a massive trap for marketing leaders.

00:09:47: okay here's where it gets really interesting.

00:09:49: Do we actually need all this data flowing in real time, or is real-time just a vanity metric that's aggressively draining our budgets?

00:09:58: That is the exact question Jonathan Moran addressed with his concept of The Freshness Spectrum.

00:10:04: I like that!

00:10:05: Right

00:10:05: because everyone hears about AI querying the warehouse and instantly assumes oh i need all my data streaming in real-Time.

00:10:12: but Real Time has often just massive self inflicted MarTech tax.

00:10:15: so these Real Time Data Pipelines are incredibly expensive to build and even more expensive to maintain.

00:10:22: So do you need milliseconds of latency sometime?

00:10:26: Like for fraud detection,

00:10:27: exactly?

00:10:28: or an abandoned shopping cart where the customer is literally still on the website.

00:10:32: in those cases real time as non-negotiable.

00:10:34: but if I'm say building an audience segment For a standard Thursday newsletter Or just doing weekly lead scoring

00:10:41: then fresh enough meaning data updated hourly or even daily is perfectly fine.

00:10:47: Right, building low latency streaming infrastructure for marketing use cases that could easily run in batch processing Is frankly like buying a commercial jet just to commute the grocery store.

00:10:57: it's

00:10:58: overkill

00:10:58: complete Overkill.

00:11:00: you're over engineering for speed You don't actually need.

00:11:02: yeah make your underlying data architecture infinitely more complex and prone to breaking.

00:11:07: That is such a vital reality check.

00:11:09: Stop paying a premium to stream data in milliseconds if the campaign doesn't even execute for three days.

00:11:15: Hey, real quick!

00:11:16: If you're finding these insights valuable for your own stack strategy take a second to subscribe to this show right now.

00:11:21: so don't miss our future.

00:11:23: deep dives into B-to-B marketing.

00:11:25: definitely do that because it all comes down to aligning your technology with the actual reality of.

00:11:41: With data architectures shifting toward the warehouse and the Martek landscape officially bloated beyond comprehension, I think we're looking at over fifteen thousand five hundred different tools out there now.

00:11:51: It's absurd!

00:11:52: it is And CFOs are desperate to consolidate.

00:11:55: The boardroom directive is always the same Cut Tools To Save Money.

00:11:59: But Nital Shah raised a brilliant counterpoint to this.

00:12:02: Dropping software licenses Is often entirely cosmetic

00:12:05: Cosmetic?

00:12:06: How so?

00:12:07: Well

00:12:08: If you cut three distinct tools and replace them with one unified platform, but you don't actually fix the underlying governance in broken workflows.

00:12:15: You haven't solved anything!

00:12:16: You just squeeze that exact same broken workflow into a smaller cheaper box

00:12:20: Exactly.

00:12:21: And if you drop these new autonomous AI agents onto that broken undocumented process The agent executes this mess faster at scale With total unwavering confidence

00:12:33: Which is terrifying because it's not about The logo on the software, it's about the organizational fit.

00:12:40: Franz Reimersma recently published some fascinating research on this-

00:12:43: Oh I saw that!

00:12:44: He looked across nine hundred and fifty three real world marketing stacks And...the data proves something that vendors absolutely do not want you to hear.

00:12:54: Which is there?

00:12:55: no such thing as a universal best in class technology which

00:12:58: Is really tough pill to swallow for procurement teams You know?

00:13:01: They just wanna consult a quadrant by the highest rated tool and assume that problem is solved.

00:13:06: Yeah, but RhymerSmith's data shows outperformers don't just buy a tool with most features they align tech in their specific operational context.

00:13:15: In email marketing research actually showed that high performing teams use fewer features on their platform But they possess significantly higher internal skills to leverage those core features.

00:13:25: Makes sense because an off-the-shelf platform is explicitly engineered for the average team in any given category.

00:13:32: Yeah, has to be!

00:13:33: To maximize its addressable market.

00:13:35: but your team as an Average you have your own complex sales handoff requests.

00:13:39: yeah Your own Byzantine legal approval loops your own unique data gaps.

00:13:44: Daniel Chishti took this reality to it's logical extreme recently.

00:13:48: What did he do?

00:13:49: He actually built four custom marketing tools himself.

00:13:52: This year Four.

00:13:53: Yeah, he said he got tired of renting someone else's roadmap.

00:13:58: He looked at his team as specific highly nuanced workflows and realized that forcing His team to bend their processes To fit the rigid template of an average vendor's platform was causing more friction than The tool is even solving.

00:14:11: wow yeah when you buy complex software That your team isn't ready for it's literally like strapping a jet engine onto A unicycle.

00:14:18: I love that analogy but

00:14:19: problem Isn't lack of thrust right?

00:14:21: The problem is that your organizational chassis is going to instantly disintegrate under the pressure.

00:14:25: So what does this all mean?

00:14:26: Are we just blaming the tech when the organization is a real bottleneck?

00:14:30: Scott Brinker has been drawing the same chart, it's essentially Brinker's law.

00:14:36: Technology changes exponentially but organizations change logarithmically.

00:14:42: The technology curve is shooting straight up while human adaptability just crawls along a slow steady incline.

00:14:49: And the injection of AI agents has just cranked up that exponential tech curve to an absurd degree.

00:14:55: Yeah, The gap between what technology in your stack makes possible and What you're marketing team is actually capable?

00:15:02: Of absorbing operating and governing Is widening at a terrifying rate.

00:15:07: so it's no longer a procurement question of what should we buy.

00:15:10: It's a survival question of can our organization Actually adapt to run this without breaking the pipeline exactly?

00:15:16: And if you try to bolt an exponentially improving AI onto a logarithmically adapting organization, the people sitting in the middle are the ones who get

00:15:24: crushed.

00:15:24: The human toll of the exponential curve?

00:15:26: It is absolutely brutal out there!

00:15:28: Jay Schwiddelsen shared a reality check from a new NAC report and the stats paint very different pictures than the utopian AI marketing we see on vendor demos.

00:15:37: What were their numbers?

00:15:37: Eighty-two percent of marketing teams report spending at least half of their entire week strictly on production.

00:15:45: Wow!

00:15:46: Not high-level strategy, not creative thinking just grinding out campaigns and trying to get things out

00:15:51: the door.

00:15:52: And this massive production burden is happening concurrently with these so called AI revolution that was supposed free up all this time

00:15:59: Because AI isn't a magic wand.

00:16:01: yet The same report found that eighty-eight percent of marketers say AI generated content still requires substantial, heavy human editing before it's even remotely usable.

00:16:11: Eighty eight percent?

00:16:12: Yeah.

00:16:12: and sixty percent says it takes at least four different people just to review or produce one standard marketing email

00:16:19: which points directly back to the missing brand context we talked about earlier.

00:16:22: The AI can write an e-mail instantly sure But because it doesn't intuitively know your legal disclaimers or specific brand voice, the human team has to spend just as much time editing output they would have spent writing from scratch.

00:16:35: And this friction is causing massive role confusion.

00:16:38: that's actively breaking teams.

00:16:41: Sean Amreili and Borna Grinbaum both raised massive red flags recently about how companies are treating marketing ops in RevOps as synonyms.

00:16:49: Oh yeah I see this constantly on job boards.

00:16:52: Yeah, looking for a Rev Ops wizard to run our Marketo instance

00:16:56: right and it sets the new higher up for immediate spectacular failure because you are conflating two entirely different altitudes of operation.

00:17:03: completely

00:17:04: different

00:17:04: marketing.

00:17:05: ops or momops owns The highly granular systems inside the marketing department.

00:17:09: They build the complex lead scoring models map the field values define specific logic For when a lead gets handed to sales It is deep architectural work.

00:17:19: Which is where people get incredibly confused with RevOps, because RevOps sits entirely above that.

00:17:24: yes they look at the entire pipeline math across marketing sales and customer success.

00:17:28: They care about whether the revenue engine holds up end-to-end not weather.

00:17:32: a specific trigger campaign fired correctly on a Tuesday

00:17:36: right?

00:17:36: And if you staple those two jobs together for one salary The pipeline eventually breaks, because the person is too busy putting out strategic revenue fires to focus on the granular logic of the automation platform.

00:17:48: And when nobody's focused on the granular logic... ...the entire system begins to rot from inside-out!

00:17:55: Ashley Langford highlighted this beautifully.

00:17:57: What would her take?

00:17:58: She pointed that complex instances like Mercado don't fail as the tool is inherently flawed.

00:18:05: They fail from neglected ownership.

00:18:07: Oh, that's the classic scenario.

00:18:09: The in-house expert leaves for a new job and the Institutional knowledge of how the logic works just walks out the door with them

00:18:15: exactly.

00:18:16: And six months later you have fifty active workflows named something like Untitled campaign.

00:18:22: final four Donna touch and nobody knows why a certain trigger fires.

00:18:26: So leadership looks at it calls the system a mess and asks for a six figure rebuild.

00:18:30: But it wasn't a mess It was just abandoned the logic decayed

00:18:35: And this specific type of data neglect is absolutely catastrophic when you introduce AI into the mix.

00:18:41: Shivangi Oosti made a crucial point that AI personalization fails instantly if it's running on unknown, unvalidated CRM data...

00:18:50: Because you cannot personalize what you don't accurately know.

00:18:53: If your core job title field is filled with values like NA, Unknown and as DF

00:19:00: An AI agent doesn't magically fix that.

00:19:02: No!

00:19:02: It just takes your bad guesswork automates it and makes you confidently wrong at a massive, highly visible scale.

00:19:09: So if our data is in neglected mess and our teams are spending all their time just trying to get campaigns out the door what happens when we hook autonomous AI up this shaky foundation?

00:19:18: It upgrades from an operational headache.

00:19:22: Mypastore brought up a sobering point about governance that every single marketing leader needs to internalize.

00:19:27: I'm ready.

00:19:28: Marketers will spend millions of dollars in countless hours obsessing over collecting and structuring proprietary customer data, but they rarely stop to ask if their martech vendors are using the exact same hard-earned data to train AI models who effectively help their competitors.

00:19:44: Wow!

00:19:45: So your proprietary intent data The signals you paid to capture and refine could be secretly making a competitor's AI engine

00:19:53: smarter.

00:19:54: Exactly, did anyone actually read the terms of service when they clicked accept?

00:19:59: To turn on that cool new AI subject line generator?

00:20:02: probably not suddenly.

00:20:03: this isn't a marketing operations issue anymore.

00:20:05: This is illegal issue.

00:20:06: It's a security issue.

00:20:08: it has a finance governance issue.

00:20:10: yeah in a world of headless agents Model context protocols and open data ecosystems stringent governance is literally the only thing standing between your brand.

00:20:20: an absolute chaos.

00:20:21: It's like we built that house.

00:20:22: We talked about at the very beginning But we forgot to put locks in the front doors.

00:20:26: And now the smart appliances are inviting the neighbors into look at our financial documents.

00:20:30: That is exactly what it's like The capabilities of the technology or vastly outpacing our operational maturity To safely manage it.

00:20:38: which actually leads me to a final provocative thought for you to mull over as you look at your own stack today.

00:20:44: Okay,

00:20:44: let's hear it!

00:20:45: This is inspired by strategist Carolyn Healy.

00:20:47: I want you to ask yourself this question if you turned off every single AI tool your team uses today what actual AI capability or underlying operating architecture would remain inside your organization?

00:21:01: Oh man, that is a remarkably heavy question to end on.

00:21:04: Because if the answer's nothing... ...if it all just vanishes when you cancel a SOF subscription then you don't actually have an AI strategy.

00:21:10: Precisely!

00:21:11: You just have a bunch of rented tools.

00:21:12: The goal in this era isn't adoption.

00:21:21: If you enjoyed this episode, new episodes drop every two weeks.

00:21:24: Also check out our other editions on Field Marketing, Channel Marketing and Partner Ecosystem AI & B-to-B Go to Market ABM & Social Selling.

00:21:32: Thanks for listening everyone!

00:21:33: Thank you so much for joining us on This Deep Dive.

00:21:35: Don't forget to hit subscribe So never miss an update.

00:21:38: And as we head back To your own Martech stack today Just remember to check who or what is currently rearranging Your furniture.

00:21:45: See ya next time.