Best of LinkedIn: Martech Insights CW 44/ 45

00:00:00: provided by Thomas Allgeier and Frenus based on the most relevant LinkedIn posts about MarTech in calendar weeks, forty-four and forty five.

00:00:09: Frenous 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.

00:00:18: you can find more info in the description.

00:00:20: great so we are essentially unpacking The absolute most critical MarTech trends have been bubbling up across linkedin recently.

00:00:28: Yeah, specifically for B to be marketing professionals

00:00:31: and our mission For this deep dive is just cut through all the noise.

00:00:34: exactly

00:00:35: no fluff No over explaining.

00:00:36: The basics Just looking at the real business impact of what these frontline practitioners are actually seeing in the trenches.

00:00:42: Yes So you can adapt your own strategy?

00:00:44: And I mean we really have to start with his massive tectonic shift that just happened.

00:00:49: Oh

00:00:49: yeah a billion dollar pivot

00:00:51: right.

00:00:52: So imagine you've spent the last two years carefully building this beautiful Modular tech stack

00:00:57: like picking specialized independent tools specifically so you don't get trapped in a monolithic vendor suite.

00:01:03: exactly.

00:01:04: You go to sleep feeling totally secure about your architecture and then you wake up And Salesforce just bought the crown jewel of your stack for five point six six billion dollars in cash.

00:01:13: Oh, cash no debt.

00:01:14: that's wild.

00:01:16: That's the exact nightmare scenario.

00:01:18: A lot of marketing operations folks are waking up to this month with Salesforce acquiring brace

00:01:22: vendor luck and just literally happened you overnight.

00:01:25: Yeah, and you look at what Ryan Warren And Nikhil Neema or saying about this in a statement they're making is just crystal clear.

00:01:32: real-time engagement Is now officially?

00:01:34: The most expensive highly valued layer of the entire tech stack

00:01:38: which is huge.

00:01:39: I know Al Iverson and Sanjipoo we were echoing that too saying it's all about fusing that deep customer intelligence with in-the moment execution.

00:01:49: Right, but hold on because as a practitioner this actually terrifies me a bit.

00:01:53: Why?

00:01:54: I mean it's powerful combo

00:01:55: It is!

00:01:55: But if real time engagement really the crown jewel doesn't folding historically independent player like braze into massive sales force monolith triggers some major vendor lock and anxiety for you.

00:02:08: Oh sure You bought a tool to stay agile and now your customer journey logic is owned by the exact suite you were trying to bypass.

00:02:15: Right, And that tension is exactly what Stefan Gopé's highlighting right now.

00:02:20: He makes this brilliant distinction.

00:02:22: he says vendor lock-in Is no longer just a contractual annoyance where you're just stuck paying a bill for three years?

00:02:29: It's deeper than that.

00:02:30: Exactly.

00:02:30: it's mutated into fundamental architectural risk.

00:02:34: because think about how these stacks are built today.

00:02:36: You might legally own your raw customer data, sure.

00:02:39: Yeah but if you're segmentation logic and your complex journey workflows like Your actual decisioning models are deeply hard-coded into a proprietary platform

00:02:49: Then...you don't really own your marketing engine at all?

00:02:51: Exactly!

00:02:52: Your carefully designed strategy is suddenly At the mercy of someone else's corporate acquisition roadmap.

00:02:57: Which brings up Gabriel Severian point.

00:03:00: He brought up a really grounded example.

00:03:02: looking at Burger King

00:03:03: Oh right, I saw that.

00:03:04: Yeah

00:03:05: he said point blank.

00:03:06: if their braze renewal was hitting his desk Right now He absolutely would not rush into signing a three-year deal

00:03:12: because you have to step back and ask those hard questions Like what do i actually own here?

00:03:16: All right If I rip this tool out does my customer memory in my logic still function?

00:03:22: or Do I have to rebuild my entire marketing brain from scratch?

00:03:25: And the market is aggressively reacting to that exact fear right now.

00:03:29: The composable ecosystem, you know those vendors building modular tools that plug directly into your data warehouse.

00:03:35: they are moving fast to counter-the-sweets.

00:03:37: Yeah

00:03:37: Nathan Benz and I pointed out timing here which just almost cinematic

00:03:41: It really was.

00:03:42: Literally one day before sales force embrace acquisition went public Databricks dropped the beta for customer lake

00:03:48: Right.

00:03:49: Let's actually define what this means because Lakehouse data sounds like some heavy engineering jargon.

00:03:56: Yeah, let's break it down.

00:03:57: so traditionally you have to pipe your data out of your data warehouse and force into a customer data platform or CDP

00:04:03: which just creates the duplicate copy that gets stale almost instantly.

00:04:07: exactly Customer lake flips that completely.

00:04:10: It is a cdp built directly on top of governed lake house data

00:04:14: meaning in application reads the data right where already lives Right.

00:04:18: It separates storage from computing power used analyze

00:04:23: Which fundamentally changes the speed of execution.

00:04:25: I mean, Michael Trapani reported early results in this and they're just staggering.

00:04:30: Because marketing teams aren't waiting on data engineers to build pipelines anymore?

00:04:33: Exactly!

00:04:34: They are building audiences for two hours instead of two months plus seeing a forty percent reduction in CDP costs because they weren't paying to store data twice

00:04:43: And clocking at ten point conversion lift on their own campaigns.

00:04:47: It's a direct existential threat to the monolithic suite model, and the composable players are moving further up the funnel too.

00:04:54: Tejas Monohar announced that Hightouch is adding enterprise message delivery directly into their agentic marketing platform.

00:05:00: So now if you're running lifecycle campaigns You can find the opportunity build the audience And actually hit send on the message all end-to-end

00:05:09: While remaining entirely composable and independent at the core of your architecture.

00:05:14: It's just a fascinating strategic war.

00:05:16: And Franz Riemersma frames this beautifully with what he calls the openness paradox.

00:05:21: I love his concept.

00:05:22: Yeah, his thesis is basically that making your tech stack more open like intentionally designing it so customers can easily unplug and leave Is ironically The best strategy to make them stay.

00:05:34: because when you give an enterprise marketing team the freedom To connect whatever they want build without artificial constraints They just generate ROI significantly faster.

00:05:43: right Closed proprietary technology is designed to protect the revenue a vendor already has.

00:05:49: Open Technology expands what the customer's actually capable of building

00:05:53: Make it easy-to-leave and you eliminate the reason they'd ever want too, that's brilliant.

00:05:58: But this war over architecture like sweets versus composable stacks It really exposes the rotting foundation at the center.

00:06:05: our current setup?

00:06:06: Oh absolutely And nowhere.

00:06:08: where does that rot more obvious than in the core marketing automation platform

00:06:12: The MAP.

00:06:12: Yeah, Ian Harris surfaced a statistic that should just blow the mind of anyone in B-to-B.

00:06:17: Yeah was it again?

00:06:19: We are operating in a ten to twenty billion dollar market and yet B-To-B marketers have fewer than

00:06:26: eight

00:06:26: legitimate MAP options to choose from

00:06:29: Eight options for a twenty billion Dollar industry.

00:06:32: That is an incredibly starved Market

00:06:34: It Is!

00:06:35: And The reason its so stagged In is that the architecture underlying those systems is effectively fossil.

00:06:40: at this point

00:06:41: John Miller who literally co-founded Marquette out two decades ago, is openly arguing that legacy maps are fundamentally just if then rules engines.

00:06:49: And we really need to look closely at why that breaks at scale.

00:06:53: and If Then Engine relies on strict Boolean logic.

00:06:56: Right

00:06:56: it only knows binary conditions

00:06:58: Exactly.

00:06:58: Did they click

00:06:59: the link?

00:06:59: Yes or no?

00:07:00: Do you have consent?

00:07:01: yes or No

00:07:01: exactly?

00:07:02: It's completely incapable of handling ambiguity or context Or any sort human judgment

00:07:08: Which means every time a marketer encounters an edge case They have to write another rigid rule.

00:07:12: Like if they click the link, but they are an enterprise account and they haven't visited in thirty days then do this

00:07:18: And you multiply that by ten years... ...and your instance just becomes this tangled fragile web of logic

00:07:25: That marketing ops teams are utterly terrified to touch because one wrong move breaks the whole machine.

00:07:30: Right!

00:07:31: John Miller's crucial point here is that slapping a AI chatbot interface onto that legacy framework doesn't actually solve the problem.

00:07:39: No, it just takes those rigid fragile rules and executes them faster.

00:07:43: It doesn't fundamentally change how the platform reasons.

00:07:46: To borrow an analogy...it's kind of like dropping a Ferrari engine which represents this incredible dynamic AI capability into a wooden horse-drawn carriage.

00:07:55: Right!

00:07:55: The Carriage being the twenty year old Rules Engine.

00:07:58: If you rev that Ferrari engine…the Carriage doesn't become race

00:08:01: car It just violently rattles itself apart

00:08:03: Exactly.

00:08:04: AI is not a feature you just add to an MAP, it's a complete teardown of what an MAPP even is.

00:08:10: Because AI native platforms replace those pre-written hard coded rules with semantic reasoning.

00:08:16: in the moment

00:08:18: They analyze person and account history The intense signals And they dynamically calculate next best action.

00:08:25: Okay but let put yourself on the shoes right now.

00:08:29: Sure You're marketing ops leader.

00:08:31: You have spent years mastering all the weird quirks and workarounds of your current platform.

00:08:37: Is there massive disruption in ripping out the central nervous system?

00:08:40: if you're marketing engine actually worth the supposed ROI for switching to an AI-native tool?

00:08:46: Well, the bar for that kind of disruption is incredibly high and Moni Oloyere actually ran a stress test on this exact premise.

00:08:54: Oh right she evaluated those four merging challengers

00:08:56: Yeah Pominga fave inflection IO and conversion And she looked at them strictly through the lens of a battle-tested practitioner.

00:09:04: Then what did

00:09:05: you find?

00:09:06: Her takeaway was a pretty harsh reality check.

00:09:09: having exciting AI technology just isn't enough.

00:09:12: Out of the four she tested, She found only conversion offered a compelling time-saving use case strong enough to justify The sheer operational headache Of migrating away from a legacy system.

00:09:23: So you can't just sell the AI dream.

00:09:25: You have to solve a tangible pain point immediately

00:09:27: Exactly.

00:09:28: But when look at the challengers who are trying To solve those pain points Their underlying philosophy is radically different.

00:09:34: Like Fave Right?

00:09:35: Yeah Built by John Miller and Nick Bonfiglio

00:09:38: They are completely tearing down the concept of single-lead MQL scoring.

00:09:42: Which is huge, because we need to explain why MQL Scoring really fails.

00:09:46: in modern BtoB A traditional map scores an individual Right.

00:09:51: So if a CFO downloads white paper she gets ten points.

00:09:54: If director attends webinar he get twenty points.

00:09:57: But they belong at same buying committee.

00:10:00: Single lead scoring just misses that entire web of relationships.

00:10:03: Exactly, and Nick Bonfiglio notes that Fave replaces with buying group playbooks.

00:10:09: So the system uses AI to dynamically assemble the org chart And map out in a buying committee?

00:10:15: Then it measures engagement based on skills you actually tune rather than some black box algorithm.

00:10:22: We're also seeing shift where human marketers are no longer clicking buttons at all.

00:10:26: Oh yeah, Amit Datu is talking about this.

00:10:28: He's building infrastructure where autonomous agents are the ones building and optimizing campaigns.

00:10:33: He

00:10:33: pointed out that AI-native companies aren't even buying legacy MAPs anymore.

00:10:38: They're skipping generation of software entirely.

00:10:40: Yeah they just stitching together landing page builders in data warehouses And then using a single text command to an AI model like Claude To orchestrate whole thing instantly

00:10:50: Which was wild!

00:10:52: Even massive incumbents reading writing on wall here.

00:10:56: Amon Batra broke down Salesforce's pricing for marketing cloud next.

00:11:01: Right, they're launching a growth tier at fifteen hundred dollars a month and an advanced here at thirty two hundred fifty.

00:11:06: But the critical takeaway from Omons breakdown is that this isn't just an upgrade to their existing Cloud.

00:11:12: No it's completely distinct new architecture built on unified data and AI agents.

00:11:18: but this pivot agent everywhere introduces massive fatal flaw into ecosystem.

00:11:24: It really does because whether you buy an AI native map from a startup or you migrate to and incumbents new architecture, your AI agent is entirely blind without context.

00:11:35: it's the classic garbage in-garbage out problem but completely weaponized.

00:11:40: Onamoro shared a cautionary tale that perfectly illustrates how dangerous this is.

00:11:44: Oh!

00:11:44: The ski boots example.

00:11:46: Yes

00:11:46: So imagine an AI agent tasked with sending promotional offers.

00:11:51: A customer, let's call her Maya recently bought skis and is now browsing boots on the website.

00:11:56: Right

00:11:56: so they?

00:11:56: AI correctly interprets that behavioral data as high purchase intent

00:12:00: And it automatically fires off a twenty percent discount code to try and close this sale

00:12:04: Which on-the-surface looks like phenomenal hyper personalized marketing right?

00:12:09: Until you look in the wider context The agent had absolutely no visibility into the customers service database.

00:12:16: Maya currently has an open Highly escalated service claim because her skis arrive completely damaged.

00:12:22: Oh, man.

00:12:23: So she is furious waiting for support

00:12:26: and the brand is aggressively trying to upsell her in or inbox with a discount code.

00:12:29: That's awful but The data that agent use was perfectly accurate.

00:12:33: She did buy skis.

00:12:34: She was looking at boots

00:12:36: exactly.

00:12:37: But because the AI operated in the silo it lacked the fundamental business logic that a human marketer just intuitively knows

00:12:45: which is Never send a discount to an account with open, unresolved service claim.

00:12:49: Right And that knowledge exists within the company but it usually lives in human beings head

00:12:54: Or its buried into some random PDF slide deck from last year's kickoff.

00:12:59: It is completely inaccessible for machine.

00:13:01: Which exactly why Scott Breaker and Andy Gu are pushing this concept of brand context model?

00:13:07: Yeah because for thirty years digital marketing architecture has been built around web pages.

00:13:11: We format information using HTML For Human Ice To Read.

00:13:14: But an AI agent doesn't want to navigate a scavenger hunt of URLs and pop-ups, to understand your company's rules.

00:13:21: No!

00:13:21: A brand context model fundamentally changes that instead of maintaining thousands of fragmented webpages you build one canonical machine readable knowledge base

00:13:31: so you structure your content uh...your brand voice guidelines your capabilities in your business rules into vector databases

00:13:38: where concepts are mathematically linked by meaning.

00:13:41: And then every surface, whether it's a website rendering for human or an API feeding data to an autonomous agent draws from that exact same centralized truth.

00:13:51: And if your AI tools aren't connected to a governed real-time messaging model like that they will confidently hallucinate and distribute you old outdated story at massive scale

00:14:01: which is terrifying.

00:14:03: but wait let me stop you there As a marketing ops person if you tell me I have to perfectly scrub, unify and centralize all my CRM data.

00:14:10: Yeah And perfectly map my content into machine readable brain.

00:14:14: before i can deploy AI Let me tell ya the eugenic future is dead on arrival.

00:14:18: It's because it's impossible.

00:14:20: We've been trying clean B-to-B CRMs for decade.

00:14:23: They are still swamp of duplicate records in conflicting fields.

00:14:26: Are we really saying that we need pause marketing just two years?

00:14:30: Well Is the defining operational debate splitting the industry in half right now?

00:14:35: You have two very militant camps on this.

00:14:38: Okay, break them down.

00:14:38: for me

00:14:39: Camp one is the clean first side.

00:14:42: Sarah McNamara shared a really compelling stress test run by Clay to prove this exact point Right.

00:14:47: they took two different AI agents and tasked with building pre-call briefs For sales reps.

00:14:52: Yeah And One agent was hooked directly up To standard messy enterprise CRM.

00:14:57: The other was built on Clay's data layer where the underlying data had already been merged, deduplicated and cleaned.

00:15:04: And they planted a massive deal threat like hidden risk factor inside of duplicate account record to see what would happen?

00:15:11: And the results were incredibly stark!

00:15:13: The agent running on clean merge data caught that threat thirty out of thirty times...

00:15:18: ...and did it in just twenty three seconds.

00:15:19: Wow

00:15:20: But the general-purpose agent that was forced to navigate the raw, messy CRM it only caught threat eighteen times.

00:15:27: It took a full sixty seconds process and completely missed conflicting risk fields because just didn't understand needed cross reference multiple duplicate records for same company.

00:15:38: So Sarah's conclusion is simple an Agent cannot flag a risk never sees

00:15:43: Makes total sense.

00:15:44: but then you have Camp two who argue agents can absolutely handle this mess.

00:15:49: Phil Gamash highlighted a very controversial take from Arjun Pillai on this.

00:15:53: Oh,

00:15:53: I read that.

00:15:55: Pillai's thesis is spending months debating which system gets to be the perfectly clean central brain?

00:16:01: It just a massive waste of time

00:16:02: Right He argues.

00:16:04: modern large language models are increasingly capable pulling context completely disconnected systems reasoning through conflicting data and figuring out truth dynamically.

00:16:14: it kind like an upgrade robot vacuum analogy.

00:16:17: we always hear

00:16:18: like this.

00:16:20: The old way of thinking is that you have to pick up all the toys and perfectly organize the furniture before pre-organize the data.

00:16:37: It just bumps into the duplicate records, maps the edges of conflicting fields, cross references time stamps and basically builds its own spatial awareness on your messy CRM On The Fly.

00:16:49: By the way if you are finding this breakdown to the architectural battles driving Martek valuable take one second to subscribe to The Deep Dive right now so that we don't miss our future analysis.

00:17:05: Because relying on that LiDAR-style AI to navigate a messy CRM requires a massive amount of computing power.

00:17:13: Oh, for sure!

00:17:13: And the vendors are not going to just absorb that compute cost.

00:17:17: they're gonna pass it onto you

00:17:19: Which brings us into very harsh reality check.

00:17:22: The grace period where companies were playing around with AI pilots is officially over.

00:17:27: Yeah we're staring down the barrel of major financial reckoning.

00:17:31: in twenty twenty seven

00:17:32: Venkatesh Amurthy brought the receipts on this.

00:17:35: He pointed to some really sobering data from Forrester and Gartner.

00:17:38: What are they projecting?

00:17:39: They project that a third of all enterprise AI initiatives will be significantly scaled back, And over forty percent of agenic projects Will be flat out cancelled by the end of twenty-twenty seven.

00:17:50: Forty percent

00:17:51: Just canceled

00:17:53: just canceled

00:17:53: Yeah because when you let agents run wild in a messy system The API calls and compute costs just skyrocket.

00:18:00: Plus, the AI makes decisions in a black box that marketing ops can't audit.

00:18:04: So business value becomes entirely unclear

00:18:07: And we're already seeing software landscape contract.

00:18:11: in response to this Mary Wallace noted that net growth and MarTech tools has practically flatlined

00:18:16: Yeah sitting at just zero point seven percent growth.

00:18:20: These new AI native platforms are driving brutal consolidation Point.

00:18:24: solutions that only do one specific thing can no longer justify their budget line.

00:18:29: when an agentic platform could replicate the feature set dynamically.

00:18:32: But if marketing teams are aggressively cutting tools and consolidating their stacks, how these vendors making billions back?

00:18:39: Right they're spending fortunes on compute to power agents.

00:18:43: Are we all going get ambushed with our annual renewals come up?

00:18:47: Lisa Cole is sounding the alarm.

00:18:49: exactly this scenario Your next software renewal is going to fundamentally break the traditional SaaS model.

00:18:55: Because we're witnessing a rapid shift away from fixed seat licenses, right?

00:18:59: Exactly!

00:19:00: It's moving toward variable usage-based pricing.

00:19:03: Look at HubSpot.

00:19:04: They are now charging for two of their new AI agents based on volume of completed tasks

00:19:09: And that's built onto your standard seat license.

00:19:12: Yep and sales forces metering work they do in credits.

00:19:16: So... The concept of a Seat License Is basically just cover charge to get inside the club.

00:19:21: Exactly!

00:19:22: The credits are the meter running on every single drink you order inside.

00:19:26: So if a marketing ops manager doesn't tightly govern how many tasks their agents are executing, that variable compute spend will devour their entire program budget in a month.

00:19:35: Which is why Lisa heavily advises against signing long term multi-year agreements right now If you don't fully grasp how this pricing model's going evolve.

00:19:44: She says she needs to push for twelve months terms or negotiate specific clauses that allow you to shed expensive human seat licenses as the AI agents take over the execution

00:19:53: work.

00:19:54: So how do you actually survive an environment where every single action costs a compute credit and buyers are completely fatigued by AI generated noise?

00:20:03: Well, Ateer Abdul Rauf predicts that twenty-twenty seven will be the year that elite marketers win buy subtraction.

00:20:11: Subtraction meaning we have to stop equating volume with value

00:20:15: Exactly.

00:20:16: Okay,

00:20:17: a tear pointed to some alarming data from snob showing that outbound email open rates have plummeted twenty three percent just since Twenty-twenty Three.

00:20:25: Wow!

00:20:26: Twenty-three percent.

00:20:27: so you cannot Just pump Out more AI generated emails To make up the difference?

00:20:32: No

00:20:33: Subtraction is A painfully counterintuitive strategy for marketers Who are infected with this.

00:20:38: see what my AI can do fever but it's really The only viable path forward.

00:20:43: So what does subtraction look like in practice?

00:20:46: It means aggressively pruning your SEO blogs because large language models reward deep topical authority, not a scattershot of two hundred off category posts.

00:20:55: Right it means cutting out bad fit ideal customer profiles that are just draining your compute credits and your sales team's time.

00:21:03: And it probably even means looking at those massive thousand person industry conferences and replacing them with highly curated, intimate dinners for maybe twenty actual decision makers.

00:21:14: You won't get an award for what you delete but ruthless focus is the only way to survive

00:21:19: and operationally surviving this lean era requires a very diagnostic mindset.

00:21:25: Rima Thomas offered a brilliant rule from managing this.

00:21:28: that's her role.

00:21:29: she says when a metric suddenly drops in your dashboard Your first instinct as a marketer is to panic and try to fix the first anomaly.

00:21:36: you see, but you must resist that urge.

00:21:40: Because in an usage-based system blind tweaking literally costs money?

00:21:43: Exactly!

00:21:45: You have to frame your hypotheses first look at which metrics moved while others held steady And use that contrast as your entry point for diagnosis rather than just randomly turning dials on the machine

00:21:56: because every dial you turn might cost you fifty credits

00:22:02: Absolutely.

00:22:03: Well, as we look at this entire landscape I mean the consolidation of tools The architectural wars and handing over execution to AI agents...I want leave you with a provocative thought to

00:22:14: mullover.

00:22:15: Yeah building on that insight from Cassenia Ruffle right?

00:22:17: Exactly We are rushing toward a tech stack That is entirely automated and agentic.

00:22:23: But an AI Agent no matter how advanced it's semantic reasoning Is can only act On data that has been digitized.

00:22:29: Right It can only process what it actually see inside the system.

00:22:34: So, What happens to human offline signals?

00:22:38: The handshake at a private dinner?

00:22:40: Or the whispered concern about budget in hallway track at conference?

00:22:44: The nuance of an under-record phone call In AI perfect attack stack where every digital interaction is instantly scraped and processed and optimized by everyone's agent simultaneously...

00:22:55: The most valuable competitive advantage you have might just be the conversations happening in spaces where AI cannot listen.

00:23:19: Exactly!