Best of LinkedIn: Martech Insights CW 42/ 43

Provided by Thomas Allgaier and Franis, based on the most relevant LinkedIn posts about MarTech in calendar weeks. forty-two and forty three. Franus is a B to B market research company that supports enterprise marketing teams and unlocking full potential of their customer data with help from AI. You can find more info in description.

So if I told you That the more efficiently you automate your marketing this year The more software vendors gonna financially penalize for it? I mean i'd say you're crazy. but That is exactly what we're seeing right now. Right, today We are really looking at a fundamental rewiring of the Martek industry.

Yeah and were taking the top martech trends buzzing across linkedin specifically for you Strategic B-to-B marketing professionals out there. It's no fluff deep dive. Exactly Our mission to this Deep Dive is look past buzzwords. You know Were exploring massive industry consolidation The literal relocation of customer data back into warehouse Which it huge Huge Plus the rather shocking financial realities of agentic AI pricing models.

And finally, The human toll of cognitive sprawl that falls directly on your shoulders. Well let's just start with the biggest move dominating the feeds right now because to understand where Martek is going we really have look at money Right? Always follow the money. Yeah.

So Salesforce announced they are requiring braze for a staggering five point six billion dollars which is just a massive earthquake for the customer engagement platform landscape. It really is. Jens Christian Jensen shared this brilliant analysis on it that contextualizes the scale here. he pointed out, That This Is Actually Salesforce's eighth acquisition in twenty-twenty six alone.

Wait eight Just this year? When you look at that pace they aren't just buying market share. They've bought companies handling commerce conversation data content billing service and now engagement. right Now engagement.

They're mapping the entire life cycle. Which really brings up that old sweet versus composable debate? For years, you know marketers wanted a composable stack piecing together independent tools so he weren't locked into one ecosystem. And Braza was practically poster child for that independent approach Exactly.

But Jensen makes a crucial distinction there. He argues that composable architecture isn't dying. The composability is just moving inside the platform. So Salesforce's trying to offer fully composable stacks but under one master's subscription.

Yeah, and one unified data model? I mean i hear this strategic vision there... But let's inject a little practical reality for the people actually managing these systems day-to-day. Oh!

The integration nightmare Total Nightmare. Mateusz Dąbrowski pointed out A very painful truth For Salesforce customers Once This braze deal Actually closes Salesforce is going to have four completely different marketing automation platforms operating Under its umbrella. Four, that's just wild. Yeah you've got marketing cloud engagement account engagement which let us be real most of it is still called Pardot Marketing Cloud Next and now Brace.

That is a massive amount overlapping functionality to govern. I tell ya what she actually joked about this pointing out the irony for buyers like if your company spent last two years migrating from Salesforce to Brace to get fresh start. well welcome back Right back where you started. You're just waiting on internal roadmaps to see how your tool integrates with the mothership.

It makes me wonder if buyers are getting trapped in these endless, internal migration loops. The friction there is very real. But Chris Parker he elevated this conversation With a really sharp insight that looks past the immediate integration headaches. Oh yeah What did he say?

He argues this acquisition isn't just about Salesforce hoarding more execution channels. It's a multi-billion dollar bet on who owns the AI decision layer The Decision Layer? Okay, let's break down the mechanics of that. Right because historically software was an execution engine.

A marketer writes a rule If they download white paper or wait two days send them to email. The platform didn't think it followed flow chart Exactly But AI flips that entirely. The battle isn't about execution anymore, any cheap tool can send an email. So true the real Battle is about which platform gets to ingest all the signals and decide what message to Send through Which channel at?

What exact moment or if it should send anything at All. so when Salesforce buys braids they're buying a massive piece of That autonomous decision-making engine. They want to be the brain. If the battle is over who gets to be the brain?

that immediately raises a massive architectural question. What data? Is feeding that brain? Oh, for sure because an AI making autonomous decisions on bad or outdated data It's just gonna make bad decisions infinitely faster than human could exactly.

The intelligence layer is completely useless without a flawless data foundation and four years We've forced marketers into really flawed setup. Aaron Bird highlighted this perfectly. He noted that legacy marketing automation platforms are essentially sealed databases, right? They usually only sync with your CRM.

Yeah But you're CRM doesn't hold the most valuable buying signals The real gold. like product usage telemetry building history complex support tickets That data sits fifty feet away in the enterprise data warehouse. So marketers are stuck trying to personalize campaigns using just like a first name and a job title. Well, the actual behavioral signals that tell you if an account is about to churn are completely out of reach which Is why teams try to copy all that behavioral data Out-of-the-ware house and stuff it into The marketing platform.

we can within A few months the sink breaks do copies don't agree And nobody knows Which database holds? The actual truth It's a mess! Which brings us To why Ferdinand van butzler's recent update is so timely. he discussed Databricks customer lake entering beta.

Oh, this is a fundamental paradigm shift. It really is. it puts the customer data platform The CDP inside the lake house. just to clarify for anyone listening who might not be a data engineer A traditional data warehouse is highly structured and kind of siloed while a data Lake holds fast amounts Of raw data cheaply right?

And a lake House combines the two cheap storage but strict governance Exactly. So, for an end saying we're taking the CDP and dropping it natively into that environment? That's zero copy architecture. in a nutshell.

today you take data out move to your cdp match again covered at second time customer lake removes that second copy entirely. so just build audience segments directly on table sitting in central data catalog. yes The data never leaves building. This connects so perfectly to a framework Cheval Kulstresla shared.

He calls it the five-piece framework for trusted activation Purpose, person place payload and proof. that's one. His central argument is we need to stop moving The underlying customer record And only move the decision. That payload concept is critical there.

Yeah I like to think about It Like A nightclub. in the old setup Handing a profile to marketing platform was like handing the bouncer of customers entire life biography. Transaction histories, support tickets everything. Just know if they can commit Right Which is massive security risk and incredibly inefficient.

But with Chivalre's zero copy approach you keep all that heavy data locked down in the lake house. You just hand the Bouncer VIP list. The media platform needs who to target And what message. it doesn't need their whole-life story.

It completely eliminates the need To create yet another conflicting copy. You've just passed lightweight instructions. By the way, if tracing these architectural shifts from AI decision layer down to zero copy lake house is helping you connect dots on where your tech stack is heading make sure hit subscribe! Yeah we really love breaking down mechanics of these changes for ya.

so...you don't want miss future deep dives. Definitely So. looking at were are Data secured in Lake House Platforms fighting to own the AI Decision Layer?

But how does actual work get done? On a random Tuesday without completely losing control of what the machine is sending to your buyers. Because you can't just plug a large language model into your CRM, give it budget and hope for the best? No!

It'll hallucinate a discount code and email your entire database at three AM Exactly...it requires a completely new operating model. Lohet Barua laid out five necessary architecture layers for what he calls an agentic go-to market system. The AI needs a harness.

Barua breaks down into context skills agents approvals & reporting. Walk us through how that harness actually functions in the software. Well, first you have context. this is The data that AI agent needs to ground its decisions like those buying signals from the lake house we talked about.

okay I got it. then You have skills and instead of giving me a free reign? You give it discrete technical capabilities. Like one skill might be Enriching a contact record another is writing a draft email so you actively restrict what it's physically capable of doing.

Yes. And then you groove those skills into distinct agents with clear roles, a research agent an outreach agent. and what about the approvals layer? That is The most critical part.

that's where You define the boundary between autonomous action in human judgment. routine data enrichment might happen instantly In the background but any customer facing message gets flagged for a human to review. This represents a massive shift in how marketing ops teams actually build campaigns. John Miller's recent work with the new platform called fave highlights this perfectly.

Oh that shift away from flowcharts, right? We're so used to building those giant sprawling. if this then that flow charts That take three weeks to map out on whiteboard and Miller points. Out that Those are being replaced by AI driven playlists.

Yeah, let's explore the mechanical difference there because I have to admit when i first read about playlists. I was skeptical. if I don't Have my giant flowchart How do I guarantee? The AI doesn't send this same e-book To My top prospect three days in a row.

you feel like You need that absolute control over the path exactly. but That controls an illusion isn't it? buyers rarely follow the exact path you mapped out. they never Do.

and thats what the playlist solves. Instead of hard coding every single path, the team builds a library of content. Then you set negative constraints like never send the same asset twice or Never email an account with an open critical support ticket within those guardrails. The AI evaluates real-time signals and decides dynamically who gets what piece of content.

today It's an incredibly elegant solution for scale. it completely removes the need For those fragile engagement streams that break the second. a prospect skips this step? But, and this is a big bot.

This brings us to hidden and frankly shocking reality about the financial cost of automation. Oh man! Carter McKay's deep dive into HubSpot new credit-based pricing model should be massive wake up call. If you are managing a MarTech budget right now listen closely to these numbers.

My jaw was on floor reading this... McKay points out that HubSpOT has promised Wall Street thirty percent operating margin by twenty thirty. To hit those aggressive targets. they're moving towards a credit based system Now.

charging credits for AI agents makes sense, right? Running large language models requires compute power which costs real money. Sure but HubSpot is applying this new credit meter to core traditional non-AI features too. Right.

under this model it cost fifteen credits To send a single marketing email. Fifteen credits Yeah. and It costs three credits For every single workflow action. If a prospect accepts A quote that's one hundred Credits.

Let's do the math on that. If you build a standard workflow that checks five conditions, updates the CRM property and triggers one email. You are burning through credits at an astonishing rate just for basic automation. McKay calculated that a totally normal month from mid-tier customer would blow right past their included credit limit triggering hundreds of dollars in automatic overages.

it completely flips the incentive structure for how we built systems For the last decade. if you were good marketing ops manager your goal was to automate everything make team more efficient. Now the better you automate your system, the more of a vendor taxes for using it. You aren't just managing marketing performance anymore - you're managing AI fin-ups!

…you literally have to look at a campaign and calculate microeconomics in the tech stack… Like if an agent decides that buyers really need three follow up emails to convert but their monthly budget can only afford one who wins? It introduces a massive friction point And this highlights the ultimate bottleneck. We have fifteen thousand plus martech tools flooding the market. Data architecture is moving to The Lake House, AI agents making autonomous decisions and now financial meters ticking on every single workflow action...

the entire burden of orchestrating this chaos falls squarely on the human operator! And that burden has a name. Nero Vias introduced a concept that perfectly captures the daily reality of being a strategic BDB marketer right now. Cognitive Sprawl.

Yes The analogy he uses for cognitive sprawl is just fantastic. He compares a highly composable modern tech stack to high-end restaurant kitchen. Okay, I love this one. you have all these amazing specialist stations the grill the saute station the pastry chef.

But in this digital kitchen there was no pass the past being the central counter where the head chef checks every single plate before it goes out To the dining room. exactly Because there is no unified workspace. in a sprawling Mar-Tex stack, the human campaign manager becomes the pass. Viya says that the marketer literally become's a Human ETL Pipeline which is wild.

extract transform load. it's data engineering term for what software supposed to do automatically. But in our reality, the human is doing it. You're manually extracting an audience list from your Segmentation tab, transforming a copy into an AI Assistant Tab loading images of the Digital Asset Manager and tracking approvals on separate spreadsheets.

You are manually carrying contacts across fourteen different browser tabs just to launch one campaign. It's exhausting! It slows down cycle times then leads to massive burnout... And you know, this cognitive sprawl doesn't just happen by accident.

It happens because we fundamentally misunderstand our tools. Carrie does expand it on us brilliantly with her Martek Explorer concept didn't she? She did! She argues that stacks become sprawling unmanageable messes Because teams constantly confuse the tool's capability With its responsibility.

This is such a critical distinction. Okay let's ground in real example Capability vs Responsibility. So capability is simply what a tool is technically able to do. Responsibility, as with the business has strategically decided that tool was for right?

Okay let's say you have HubSpot outreach and intercom in your stack. Technically all three of those platforms have the capability To send an outbound email. but if you haven't assigned strict responsibility If you haven' drawn A hard architectural line saying Intercom Is only For in-app support Outreach is only for sales and HubSpot owns marketing email, then the tools start treading on each other. And as a result your buyer gets three conflicting emails from different systems in the exact same day.

Because every team uses their capability directly in front of them. You get overlapping functionality Conflicting data and zero accountability Which leads to our final point from Ashley Langford. because architecture is constantly shifting and platform responsibilities are muddy The marketing ops professional ultimately becomes the memory of a company. Oh, they hold all the undocumented tribal knowledge?

They're only ones who know why specific lead routes to enterprise teams set up commercial Or nobody ever sends campaign on last business day or quarter. They know which executive demanded that one useless stash for three years ago that is still burning compute credits today. Exactly Langford points out that this critical logic isn't documented in a wiki anywhere. It just lives on the head of whoever has been, in the system...

the longest." It's an incredible superpower to possess that institutional memory? You become invaluable! But from an architectural standpoint it is a terrifying single point-of-failure.

Yeah if that person leaves for new job or cognitive sprawl simply burns them out. The entire operational logic of marketing engine walks out door with him. So if you are a strategic B-to-B marketing professional listening to this, You're navigating a minefield right now. You're dealing with massive vendor consolidation where platforms trying buy up the brain of your operations?

Your'e trying push data down into lake house so that you don't have fifty conflicting copies decaying in different silos. You try and build harnesses for AI agents. they can run playlists without bankrupting department on workflow credits. And you're trying orchestrate all these!

without turning your team into exhausted, fourteen-tab human data pipelines. Surviving this really requires a fundamental shift in how we view marketing technology. We have to stop looking at MarTech as the loose collection of apps to be managed? We have start treating it is unified operating system that require strict governance ruthless financial oversight and very clear human boundaries regarding capability versus responsibility which leads perfectly to a final thought for everyone to chew on as these tectonic shifts continue.

Yeah, I want to leave you with this. As we move rapidly toward the world where your company's AI marketing agents are interacting directly With your buyers AI procurement agents literally machines negotiating with machines To evaluate software. will traditional be-to-be marketing collateral even matter? Well We still need PDFs and webinars.

or well The future of martech simply Be about optimizing raw perfectly governed data feeds for machines to read and evaluate on behalf of the human buyer. That is a wild thought to leave on, are we just formatting data or other algorithms? Well if you enjoyed this episode new episodes drop every two weeks. also check out our other editions on field marketing channel marketing in partner ecosystem AI & BtoB go-to market ABM and social selling.

thanks for listening. thank you so much for joining us on This Deep Dive And don't forget to subscribe To stay ahead Of The Shifts. See You Next Time.