Best of LinkedIn: MarTech Insights CW 20/ 21

00:00:00: This episode is provided by Thomas Allgaier and Frennis based on the most relevant LinkedIn posts about Martek from Calender Weeks, twenty-and-twenty one.

00:00:08: Frennis is a B to be market research company that supports enterprise marketing teams in unlocking full potential of their customer data with help of AI.

00:00:16: you can find more info.

00:00:21: Five million dollars on forty seven different marketing tools.

00:00:25: You think you have this ultimate tech stack,

00:00:27: right?

00:00:27: The dream stack

00:00:28: exactly.

00:00:29: and then your CEO walks into your office and asks This seemingly simple question like hey which specific channel brought in our last one hundred enterprise customers.

00:00:39: Oh, man!

00:00:40: The classic nightmare scenario.

00:00:42: because honestly you have absolutely no idea.

00:00:44: You have no idea?

00:00:44: Pull off three different dashboards and I mean they all give a completely different answer.

00:00:49: Welcome to the deep dive everyone.

00:00:50: Yeah welcome.

00:00:51: And that's scenario is playing out in boardrooms everywhere right now.

00:00:56: Today we're taking a look at the top Martek trends.

00:00:58: We've seen across LinkedIn over the past two weeks.

00:01:01: Yeah, we were cutting right through the theoretical fluff.

00:01:03: You really want to focus on the actual mechanics of what is working for advanced B-to-B marketing teams today

00:01:09: Exactly because the whole technological foundation of our industry Is basically getting rewired?

00:01:15: Today work clustering these insights into A few big themes will cover contextual AI stack rationalization The evolution of marketing ops and Of course CDP strategy.

00:01:25: And that rewiring is really the only place to start, because the ground is literally shifting under our feet right now.

00:01:31: Scott Brinker and Franz Reimersma just released The State of Martek twenty-twenty six report.

00:01:36: Oh yeah!

00:01:37: The numbers they presented at Martik Day paint such a fascinating picture.

00:01:40: They really do.

00:01:41: I mean we are currently sitting at fifteen thousand five hundred five different marketing tools on the landscape

00:01:46: Which, you know sounds like a plateau if you just look at the net growth.

00:01:49: The overall landscape only grew by point seven percent this year.

00:01:52: right and A lot of people saw that flat line And assumed okay?

00:01:55: The martech explosion is finally over.

00:01:57: things are stabilizing but

00:01:58: that flatline Is incredibly deceptive because right beneath that point seven-percent net growth is This massive violent

00:02:07: churn.

00:02:08: Yeah, the turn is wild.

00:02:09: it is last year one thousand four hundred and eighty eight entirely new tools into the market.

00:02:15: And at the exact same time, one thousand three hundred sixty seven tools completely vanished.

00:02:20: Wow!

00:02:21: So we aren't seeing a stabilization at all.

00:02:23: We are seeing like a mass extinction event for a very specific type of software, followed by an immediate re-population.

00:02:30: And

00:02:30: Jen Merkel brought some brilliant clarity to why that specific churn is happening.

00:02:35: She pointed out this fundamental difference between the software thats dying and the software that's surviving.

00:02:40: The tools that vanished were almost entirely built for human workflows.

00:02:43: They're designed around graphical user interfaces

00:02:46: Like dashboards where our human marketer actually logs in and clicks few buttons builds a segment manually and schedules a campaign

00:02:52: Exactly.

00:02:53: But the one thousand four hundred eighty eight new tool that survived and entered the market.

00:02:57: Completely different!

00:02:58: They are being built as headless infrastructure for AI agents.

00:03:01: Yeah, The Interface is no longer designed for eyes.

00:03:04: It's designed to be read and operated by machines.

00:03:07: Which brings us To this concept Brinker & Reimersma call THE GOLDEN CONTEXT.

00:03:12: Oh, This Is A HUGE SHIST!

00:03:14: HUGE!

00:03:15: Historically your MarTech stack was a system of record Right?

00:03:19: It just hoarded static data.

00:03:21: A list of names, an email address maybe a log-of-a-white paper.

00:03:26: someone downloaded six months ago.

00:03:27: But

00:03:27: static records don't generate revenue anymore.

00:03:30: The value is now entirely in fluid real time context

00:03:34: Exactly like what does the customer doing on your pricing page right this second and how Does that relate to your company's current service capacity?

00:03:41: because AI agents require That golden context to make instantaneous autonomous decisions

00:03:46: bought on.

00:03:47: I can hear B to be marketer sweating right now though.

00:03:50: If the new standard requires software built specifically for AI agents to act in real time, does this mean teams have to just rip out their entire legacy stack?

00:03:57: Like let me load a hypothesis here is this transition like traveling to Europe where you have to throw away all your expensive electronics because of the wall?

00:04:06: outlets are totally different?

00:04:08: or Is there way to just buy universal adapter For systems we already rely on?

00:04:13: Fortunately We're looking at a Universal Adapter scenario.

00:04:17: Daryl Alfonso published a fantastic breakdown of how this actually works mechanically.

00:04:21: Okay, good!

00:04:22: The adapter is called MCP the Model Context Protocol.

00:04:27: In the previous era if you wanted an AI application to read data in your CRM Your engineering team had build custom point-to-point API integration

00:04:37: which was expensive, brittle and broke every single time a platform updated.

00:04:41: Exactly!

00:04:42: You have one custom pipeline for Salesforce totally different custom pipeline from Arcado another for your content management system just to maintenance nightmare.

00:04:49: So how does MCP fix

00:04:51: that?

00:04:51: It acts as an open universal standard.

00:04:53: A legacy tool only needs support MCP once.

00:04:57: From that moment on any AI assistant that speaks MCP can securely plug right into it.

00:05:01: Oh wow.

00:05:02: So an AI agent can read your existing CRM data, analyze past campaign performance in your analytics suite and draft a hyper-personalized email without anyone building a custom bridge between those three islands?

00:05:15: Exactly!

00:05:16: And we are seeing the practical ROI of that architecture already.

00:05:19: Jay Sanderson just cataloged fifty different use cases where teams are utilizing a RAG-driven AI intelligence layer.

00:05:25: Right, and RAG stands for Retrieval Augmented Generation?

00:05:28: Correct!

00:05:28: Instead of an AI just guessing the answer based on like generic data it was trained on, RAG acts as secure binder to your specific company's data.

00:05:37: You hand that binders to the AI before it acts.

00:05:39: And Sanderson showed how this layer sits at top of all existing supposedly outdated stacks.

00:05:43: so you keep Salesforce or HubSpot.

00:05:45: Exactly.

00:05:46: But you use ARIC and MCP connections so a sales rep can instantly query your massive disorganized content library, pull the exact technical PDF of prospect needs to close.

00:06:07: Yeah, but we have to be careful here.

00:06:08: Plugging a universal AI adapter into a disjointed bloated stack doesn't automatically solve your go-to market problems right.

00:06:16: in most cases it just scales your existing chaos at the speed of light

00:06:20: which introduces a massive consequence because if you're underlying data structure is a mess The AI just executes bad decisions

00:06:28: faster.

00:06:29: Exactly, and Anthony Sog laid out the grim reality of this.

00:06:33: he accurately described most enterprise marketing stacks as a graveyard.

00:06:39: Ouch,

00:06:39: but fair.

00:06:39: Very Fair!

00:06:40: You have CMOs overseeing forty-seven different tools spending millions in licensing fees and they are still suffering from the scenario we mentioned at the very start They cannot attribute revenue to specific channels.

00:06:51: Alex Gluz actually dug into mechanics of why these stacks become graveyards.

00:06:56: He noted that most successful AI driven marketing leaders aren't stacking more tools to solve

00:07:01: problems Right...they're radically reducing their tool set

00:07:03: Yes, and more importantly they are assigning exactly one explicit job to each platform.

00:07:09: So research happens exclusively in tool A. content generation happens in tool B

00:07:14: because when you allow the roles of your software To drift like When your CRM tries to be an email sender?

00:07:20: And your email center tries to being analytics platform Your stack becomes The bottleneck.

00:07:27: Let me challenge that slightly though.

00:07:28: Aren't we just describing standard marketing operations here like keeping the database clean, making sure tools are integrated?

00:07:35: Why are suddenly rebranding this as architecture discipline?

00:07:38: The sheer scale of failure demands a new classification.

00:07:42: Bray Brockbank reframed this perfectly.

00:07:45: He pointed out that we are no longer just drawing pretty architecture diagrams for a slide deck,

00:07:50: right?

00:07:50: Those massive visual maps with the hundred logos on them

00:07:53: exactly?

00:07:54: Enterprise teams or suffering from systemic coordination failure because The technology is expanding infinitely faster than human governance.

00:08:02: marketing architecture has to evolve into a rigorous operational discipline.

00:08:06: It's about intentionally engineering the systems that govern revenue.

00:08:10: Yes, That distinction makes a lot of sense.

00:08:12: it's basically The difference between buying a pile of expensive bricks and actually knowing how to engineer A load bearing wall?

00:08:18: That's a great way To put it And the current waste is staggering.

00:08:23: Kumar Babu Vanapali shared a statistic that should terrify any CFO.

00:08:27: I saw this Enterprise teams utilize on average just thirty three percent Of their Martek stacks capabilities.

00:08:33: Two-thirds of the functionality they pay for sits completely dormant.

00:08:37: So, layering an AI agent over a stack where you only understand one third of the underlying mechanics creates this massive orchestration gap?

00:08:45: Exactly!

00:08:45: You're attempting to move from a system that merely executes scheduled tasks... ...to a systems that autonomously thinks and reacts.

00:08:53: That orchestration is fatal

00:08:55: And Vanipali also noted that shift fundamentally changes what holds value in company.

00:09:00: We are moving toward a prompt stack.

00:09:02: Yes, the specific instructions and context you feed the AI

00:09:05: right those prompts actually become highly guarded corporate intellectual property just like your source code or your customer list.

00:09:10: but if your architecture is brittle Those prompts fail.

00:09:14: he highlighted that seventy eight percent of AI project failure stemmed from poor instructions And chaotic data environments not from flaws in the underlying AI models themselves.

00:09:24: Wow.

00:09:24: So if architectural governance is now the most critical mandate for marketing team, we have to ask who actually owns The Blueprints?

00:09:31: Which brings us

00:09:39: Yeah, Ashley Langford posted what might be the most painfully accurate description of mom ops I have ever heard.

00:09:46: Oh!

00:09:46: The oldest daughter analogy?

00:09:47: Yes

00:09:48: she compared working in marketing ops to being the corporate version of the oldest daughter.

00:09:53: you carry this massive invisible load across every team.

00:09:57: oh man it is so true.

00:09:58: You remember all that data lives...you fix broken workflows sales created and clean up strategic messes left by leadership.

00:10:06: She also described the department as twelve highly capable raccoons hiding inside a trench coat trying to keep the go-to market engine running.

00:10:14: Which is hilarious!

00:10:15: It is, but the humor masks very serious operational vulnerability.

00:10:21: mobs has traditionally been treated You know, the people you submit a ticket to when you need a campaign schedule or data cable plugged in.

00:10:28: Right!

00:10:29: But Tara P highlighted how AI is aggressively changing that dynamic.

00:10:33: She noted that implementing AI is currently acting as world's most expensive QA test

00:10:38: Because it's stress testing every duct-taped integration and exposing every single manual workaround.

00:10:43: Exactly!

00:10:44: MOPs professionals can no longer just be platform administrators.

00:10:48: They must be strategic adults in their room defining where automated guardrails actually go.

00:10:54: Let me play devil's advocate here for a second, though.

00:10:56: If mox is currently drowning in that oldest daughter syndrome-like spending eighty percent of their week manually deduplicating contacts pulling lists and fixing broken API sinks how on earth do they suddenly have the bandwidth to step back an act as high level systems architects?

00:11:13: That Is The Exact Paradox Teams Are Facing Right Now And Mike Rizzo Provided The Mechanical Solution.

00:11:18: The Relief Comes From Deploying Conversion Agents.

00:11:21: Okay

00:11:21: what are those?

00:11:22: They are specialized AI agents tasked specifically with taking over the recurring execution floor.

00:11:28: Think about the mundane high volume tasks, running sync checks between your webinar platform and sales force routing leads based on territory rules, standardizing job titles in your database.

00:11:40: So

00:11:40: I see so the agents are absorbing all of operational busy work?

00:11:43: Precisely!

00:11:44: By offloading that high-friction low strategy execution to conversion agents... The human mops team finally regains their capacity.

00:11:53: They stop running the factory floor and actually have time to engineer go-to market architecture.

00:11:58: And Premad M took this trajectory into its logical conclusion.

00:12:01: He argued that this transition permanently shifts mobs out of back office cost center category, they become business value engineers.

00:12:09: I

00:12:09: love that title.

00:12:10: It's

00:12:10: accurate too.

00:12:11: When you are a team governing data that feeds AI You hold absolute boardroom currency

00:12:20: Because you're no longer just optimizing a newsletter send.

00:12:23: You are actively engineering the mechanics of how that company generates revenue.

00:12:28: However, and this is big however, mom-ups can design most elegant architecture in world And conversion agents can execute those designs flawlessly.

00:12:37: But entire system still collapses if underlying customer data is siloed

00:12:42: Which brings us to engine room powering agent shift the customer data platform, or CDP.

00:12:48: Right?

00:12:49: The CDP category has faced a lot of skepticism lately.

00:12:52: there's this vocal segment on the market claiming CDPs are dead arguing that modern data warehouses or composable technologies have basically rendered them obsolete.

00:13:01: but Harsha Khorasala manned today highly technical defense.

00:13:07: Yeah, his argument cuts right to the core of performance marketing mechanics.

00:13:11: He pointed out that data warehouses in standard attribution tools critically lack real-time live instrumentation and directs KPI flows.

00:13:18: And KPI stands for conversions API which is how your systems talk directly to ad network

00:13:23: servers.

00:13:24: Right if you were a brand burning millions of dollars a month on meta and Google those platforms are highly incentivized To report a phenomenal return on ad spend.

00:13:32: Oh, for sure.

00:13:33: They will happily take credit from a prospect who organically visited your pricing page six times but happened to scroll past one retargeting ad right before they finally clicked purchase.

00:13:43: Exactly the problem!

00:13:45: Harsha argues that robust CDP equipped with live SDK and unified identity graph provides actual directional truth

00:13:54: because it tracks the user's behavior across every touch point in real time.

00:13:58: Yes, so you can prove your true incremental ROAS.

00:14:02: It separates customers who are converted by ad from those who will buy anyway.

00:14:07: A data warehouse which typically relies on batch uploads simply cannot provide that real-time behavioral verification.

00:14:13: But the architecture

00:14:14: of CDP itself is fracturing too Elad.

00:14:17: Simon raised a critical warning about debate between packaged and composable CDs.

00:14:22: Mechanically, these are two entirely different philosophies.

00:14:24: Very Different.

00:14:25: A packaged CDP hoards data.

00:14:28: It requires you to extract data from all your systems and copy it into the CDPs centralized hub.

00:14:33: But a composable CDP does opposite

00:14:35: Right!

00:14:35: It assumes that your data stays where it already lives like in a snowflake warehouse And the CDP simply queries and activates them From there.

00:14:42: Both are valid engineering approaches but The danger lies in misunderstanding how they interact.

00:14:49: Simon warned against the trend of building a Frankenstein architecture.

00:14:52: Oh, where you have companies buying a legacy package CDP Welding a basic warehouse connector onto it and claiming they now have a composable system Yes

00:15:01: Which causes massive latency?

00:15:04: You are taking a system designed to hoard and copy data And forcing it to act as a system that simply points to data.

00:15:10: And that architectural clash creates duplication errors and slows the entire activation process to an absolute crawl.

00:15:26: Oh, I read this...

00:15:34: But Kumar points out the mechanical flaw!

00:15:37: If your data ingestion pipelines and your identity stitching processes take twenty-four hours to sink across a Frankenstein architecture, you're automated.

00:15:46: triggers are acting on yesterday's news.

00:15:48: It is like having a world class sniper rifle but your spotter using binoculars that show what happened twenty four hours ago.

00:15:55: That exactly it.

00:15:56: If your foundational data layer relies on legacy batch logic, the incredibly expensive orchestration engine you put on top of it is useless for true in-the moment personalization.

00:16:07: You just have to kill the batch mindset

00:16:09: entirely.".

00:16:10: And to achieve that real time capability we have to rethink.

00:16:17: Desmond Fua introduced a brilliant framework for this called SEMA and TIC.

00:16:21: Right, he highlighted that when AI personalization fails it's rarely the fault to the AI model itself.

00:16:27: It fails because the underlying data lacks structured content models and clear entity identification.

00:16:32: Let's break down the mechanics of that like what does lack of structured content actually look inside marketing department?

00:16:38: Well, it looks like a shared drive full of unstructured PDFs loosely named webpages and image files with ZO tags.

00:16:46: An AI agent scanning that drive has no inherent understanding what is actually looking at.

00:16:50: So

00:16:50: to fix this you must apply rigorous metadata clear schemas.

00:16:55: in AI-ready indexing You have to tag the technical specs The intended audience And product lines To every single asset

00:17:03: Because without that structured foundation, the AI cannot retrieve the correct information to activate the right message.

00:17:09: Food

00:17:09: as whole point is that structuring your data is no longer just a tactic for better SEO.

00:17:14: it's a mandatory foundational capability if you want your AI marketing function at all.

00:17:19: So synthesizing everything we've covered.

00:17:20: today We are looking at a landscape That has violently churned out tools built For humans replacing them with infrastructure build for AI agents Right?

00:17:32: agents to our data.

00:17:33: We are seeing a desperate need for actual architecture discipline, to prevent our tech stacks from becoming bloated graveyards.

00:17:40: we have marketing ops evolving from back office raccoons into strategic business value engineers leveraging conversion agents to buyback their time.

00:17:49: and finally we have the CDP providing the directional truth and structured real-time data required to make any of this actually work

00:17:58: which perfectly brings us necessary shift in perspective.

00:18:03: Prit and Roy analyzed all of this movement surrounding the state-of-martic report, an offer to thought that just completely reframes the conversation.

00:18:11: Yeah!

00:18:11: This was a great post.

00:18:12: Right

00:18:12: now every marketing leader is running around asking you the exact same question Which AI tools are top performing companies buying?

00:18:20: Roy points out it's entirely wrong

00:18:22: because the tool itself isn't differentiator.

00:18:25: The real question you must ask yourself what did your team build before they arrived?

00:18:29: The companies that are currently dominating with AI integration did not simply purchase a magical piece of software last month.

00:18:37: They spent the past three years doing the tedious, highly unglamorous work of cleaning their data infrastructure, structuring content and strictly defining architectural governance.

00:18:49: They fixed plumbing deep inside walls before they ever tried to install smart faucets.

00:18:54: If you enjoyed this episode new episodes drop every two weeks!

00:18:58: Also check out our other editions on account-based marketing, field marketing channel marketing AI and B to be marketing.

00:19:04: go to market.

00:19:05: And social selling.

00:19:06: Thank you so much for joining us On this deep dive.

00:19:09: make sure to subscribe So you don't miss the next analysis.

00:19:12: as You evaluate your own marketing stack This week ask yourself if you are looking at a clean governed blueprint or If you were wandering through chaotic mansion built by forty seven different contractors.

00:19:22: The agents are moving in and they require A Clean map.

00:19:25: see you Next time.