Best of LinkedIn: MarTech Insights CW 36/ 37

00:00:00: Provided by Thomas Allguyer and Frennus, based on the most relevant LinkedIn posts about MarTech in calendar weeks thirty six and thirty seven.

00:00:08: 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.

00:00:20: Yeah, and in this deep dive.

00:00:21: We're really just cutting through all the noise from recent LinkedIn discussions to bring you The top Martek trends exactly.

00:00:28: we're focusing heavily on what strategic b-to-b marketing leaders Actually need to know right now To stay ahead

00:00:35: right covering everything From You Know AI agent workflows to the actual underlying architecture of your tech

00:00:41: stack.

00:00:41: So Just imagine for a second firing up Your Marketing Platform On A Monday Morning But instead of clicking through endless drop-down menus or dragging and dropping those workflow nodes.

00:00:50: Oh, the endless nodes!

00:00:52: Or writing all that complex Boolean logic... Instead you just type into a prompt build campaign to win back our churned enterprise accounts in the logistics sector.

00:01:00: And this system

00:01:02: does it?

00:01:02: It pulls data, drafts messaging sets timing and launches Which

00:01:08: is wild.

00:01:10: Today we're exploring why traditional marketing campaigns like as we've known for twenty years is effectively dead.

00:01:17: And what has actually taken its place?

00:01:19: Yeah, I mean it's a massive structural shift and if you are marketing leader listening to this You've probably felt the friction of that old model for years now.

00:01:29: Oh absolutely

00:01:30: To really grasp why this happening we should look at an argument surfaced recently by Swatantra Kumar.

00:01:37: The core idea here was traditional campaign.

00:01:41: It was always just a compromise.

00:01:43: A compromise because of human limitations?

00:01:45: Exactly,

00:01:45: think about it no B to B marketer ever actually woke up wanting to talk to quote-unquote segment.

00:01:50: Right you don't want to blast the generic message to five thousand midmarket CIOs.

00:01:54: No!

00:01:54: You wanna have five thousand highly specific one-to-one conversations.

00:01:59: But Because we didn't have that human bandwidth To hold thousands simultaneous personalized conversation We had build these approximations

00:02:07: Right and built static segments.

00:02:09: Yeah, static segments.

00:02:10: instead of talking to individuals.

00:02:12: We built rigid predefined email journeys Instead of having an actual dynamic dialogue based on what the buyer needs in that exact moment

00:02:20: and The entire martech industry basically industrialized That compromise.

00:02:24: we

00:02:24: all spent years getting certified into these massive software platforms just To learn how to you know operate the machinery at that compromise.

00:02:31: yes

00:02:32: And that is precisely the bottleneck that's breaking open right now.

00:02:36: With AI, That clunky interface between marketers intent and software execution is just dissolving.

00:02:42: So goal-based agents are fundamentally changing the unit of work?

00:02:46: Exactly!

00:02:46: You no longer build a campaign step by step.

00:02:49: instead you give an autonomous agent objective budget strict guardrails.

00:02:54: Then the agent decides audience parameters?

00:02:57: Yeah it tailors individual message dictates timing per person self-optimizes based on live signals.

00:03:04: The goal becomes the unit of work, not.

00:03:30: Okay, for those who might not be deep in the weeds on API protocols what does MCP actually enable like...in plain English?

00:03:39: It essentially allows natural language control over complex software APIs.

00:03:47: Headless meaning no front-end user interface.

00:03:49: Exactly, you're completely separating the backend database from the UI.

00:03:53: so instead of logging into Salesforce and hunting through a clunky UI to find the right filters The AI becomes the new head.

00:04:00: So just brief Claude in plain English

00:04:02: Right You ask it to write specific SQL query To find a cohort of buyers And the AI executes that command directly Into the platform's back end.

00:04:10: That shifts the required skill set for marketing teams.

00:04:13: I mean premium is longer on software certification?

00:04:16: Definitely

00:04:17: not

00:04:17: And software vendors are already reacting to this.

00:04:20: Jason Tabbert highlighted how Optimizely used their recent Opticon event, to launch virtual teammates.

00:04:26: Oh I saw that!

00:04:28: Are those just fancy chatbots?

00:04:29: No

00:04:30: these aren't your standard frustrating website bots That forget who you're every time you refresh.

00:04:36: These are role specific AI coworkers.

00:04:39: You can literally deploy a virtual CRO manager, a conversion rate optimization specialist directly into your workflow.

00:04:46: And the critical differentiator there has to be persistent memory right?

00:04:50: Exactly!

00:04:50: This Virtual teammate retains your specific organization's context-your brand guidelines and past performance data

00:04:57: So you don't have to rebrief it every single session.

00:04:59: Right It works alongside a lean human team handling all of heavy lifting of data analysis which frees up humans on strategy.

00:05:07: But wait, I have to push back a little here.

00:05:09: If the traditional UI goes away and we're all just chatting with an AI to execute our marketing doesn't the platform itself become a dumb pipe?

00:05:16: That's what you mean!

00:05:17: Like if i'm telling Claude to run my marketing... ...and my competitor is telling Claud to run their marketing where is competitive mode?

00:05:25: If everyone is renting the exact same base-level AI intelligence The software isn't a differentiator anymore.

00:05:32: that is the existential question for software vendors right now.

00:05:35: Bill Hobbib actually explored this dynamic.

00:05:38: He noted that AI is actively eroding traditional Martek economics,

00:05:42: because the value used to be in operating the complex machinery

00:05:45: exactly for The last decade.

00:05:48: the company that could afford the most sophisticated platform and hire the Most expensive specialist run it usually one

00:05:54: but when everyone has access To abundant cheap a I

00:05:57: Operating the tools becomes commoditized.

00:05:59: So, the moat has to shift away from tool itself and toward like proprietary data in strategy we feed into it.

00:06:06: The winner is one who provides AI with unique high-quality fuel

00:06:10: Precisely Your unique business context your nuance understanding of buyers first party data that become the mote Which

00:06:17: honestly creates some natural friction.

00:06:19: Handing over keys on an autonomous AI agent is absolutely terrifying if that agent looking at bad unstructured data.

00:06:27: Oh, it's a nightmare scenario.

00:06:29: if the AI is going to act autonomously The data layer has to evolve from just storing information To providing actual rich context

00:06:38: which means the mandate for data teams Is fundamentally changing?

00:06:41: The era of the traditional customer data platform CDP is evolving so rapidly.

00:06:46: we're seeing this rise Of a new discipline called Context Engineering.

00:06:50: Yeah, Adeel Abbas recently shared a pretty provocative quote from Unifor I's CMO about this.

00:06:55: They flat out declared that the CDP era is over.

00:06:58: Wow!

00:06:59: That's massive claim especially given how much money enterprises have poured into CDPs in those last five years.

00:07:04: It sounds aggressive but reasoning holds up.

00:07:06: Think what CDPs were built to do?

00:07:08: They solved historical problem Data fragmentation.

00:07:11: Right

00:07:11: they are designed for storage and unification Pulling marketing sales support data into one profile

00:07:17: Which was huge step forward.

00:07:19: But today, simply storing unified data isn't enough.

00:07:23: The real bottleneck now is intelligence and decisioning.

00:07:26: Right!

00:07:27: Connecting that unified data to the wider stack so an AI agent can instantly understand and act on it without human interpretation?

00:07:35: Exactly... By the way just jump in here if you want catch future deep dyes as we track how these massive tech shifts unfold.

00:07:41: make sure subscribe so don't miss another edition because this landscape literally changes weekly.

00:07:47: It really is.

00:07:49: So bringing back to this shift from data, to intelligence Rupert Steffner argues strongly that context engineering needs to be treated as a completely distinct discipline From traditional data engineering.

00:08:00: To put into perspective Data Engineering Is essentially high level plumbing.

00:08:03: It's about extracting data, moving it safely structuring so the database doesn't

00:08:07: crash.

00:08:08: Yeah ensures that data is reliable but Context Engineering synthesizes raw data to answer actual business questions.

00:08:15: Steffner calls engineering a customer heartbeat.

00:08:18: I love this phrase.

00:08:19: The difference between just extracting a data point like knowing a prospect downloaded white paper and synthesizing meaningful intense signal.

00:08:26: Right

00:08:27: telling AI agent that this prospect actively in buying cycle needs immediate engagement.

00:08:33: It has the translation of raw data into business meaning.

00:08:36: John Miller had an insight that really brings this down to earth, he pointed out.

00:08:49: Think about a seasoned human marketing specialist logging into HubSpot or Marketo.

00:08:54: They rely on a massive amount of unspoken rules in their head.

00:08:57: Yes, the human knows that you always clone the Q-three webinar template V four not v three.

00:09:03: they know the strict naming conventions.

00:09:05: exactly which executive accounts absolutely must be suppressed

00:09:08: right so you don't accidentally send a ten percent discount code to

00:09:14: An AI.

00:09:15: agent doesn't inherently know any of that.

00:09:17: If you want an agent to build a campaign, You have take all the deep operational context and engineer it into what Miller calls AI skills.

00:09:24: But I could see a seasoned data engineer rolling their eyes at this.

00:09:28: They might argue like we've been building data models in enrichment prep lines for decade Why rebranding as Context Engineering?

00:09:36: The crucial difference is domain knowledge.

00:09:39: You simply cannot engineer meaningful marketing context without deeply understanding B-to-B Marketing Strategy.

00:09:45: Pradik Padra highlighted this when discussing autonomous agents in B to D catalogs.

00:09:49: Agents don't just want a raw list of SKUs from the database?

00:09:52: Exactly!

00:09:53: A raw SKU means nothing to a generative model, The AI needs contextual enrichment pipelines.

00:09:58: It need's know why and how specific product solves a specific buyer problem

00:10:03: So it can actually recommend.

00:10:04: naturally into conversation.

00:10:06: Right

00:10:06: A data engineer ensures the SKU table sinks at midnight.

00:10:09: A context engineer insures that AI understands that SKU-A is the absolute best solution for a midmarket compliance officer struggling with GDPR, and requires marketing

00:10:19: acumen.".

00:10:20: And if we pull back to look at this entire organization you realize very quickly Which brings us to the actual architecture of The Martek

00:10:34: Stack itself.

00:10:35: For a few long companies have looked at Martek as just a collection Of disjointed software subscriptions,

00:10:40: we have to start treating it As the literal architecture of your marketing operating model.

00:10:44: Stephen Getze made A brilliant observation about this.

00:10:47: he pointed out that if you look closely At a company's martech stack diagram It reveals their entire operational culture.

00:10:55: You don't just see a bunch of vendor logos?

00:10:57: No, you see the organizational silos made visible through software licenses.

00:11:02: You see duplicated capabilities because teams refuse to share tools.

00:11:07: You can literally trace the flow of data until instantly whether sales marketing and IT actually collaborate

00:11:13: or if they operate like entirely separate companies.

00:11:16: it's so true.

00:11:17: A broken stack is just a mirror reflecting a broken organization.

00:11:22: Bray Brockbank added a powerful perspective to this, noting that the recurring marketing problems we all face rising cack campaigns that take three weeks to launch sales misalignment.

00:11:32: They are rarely campaign problem

00:11:34: right?

00:11:34: We always try to fix The symptom.

00:11:35: we tweak the email subject line or just the ad spend.

00:11:39: but Brock bank argues these Are fundamentally architecture problems.

00:11:42: if your data has To be manually exported from a marketing platform cleaned in Excel and Manually uploaded into the CRM before sales can see it you don't have a campaign problem.

00:11:52: You have an architecture problem

00:11:53: and you absolutely cannot optimize your way out of a broken architecture, no amount of AI copy generation will fix the broken data pipeline

00:12:01: And there's hard empirical data backing this up now.

00:12:04: Frans Ramersma and Donovan Neil May recently analyzed nine hundred and eighty eight real world marketing stacks for The Apex Martek Report.

00:12:12: Wow

00:12:12: almost a thousand stacks.

00:12:14: Yeah they wanted to figure what top performing companies actually do differently And they found that the outperformers do not just go out and buy tools with most features.

00:12:23: They don't suffer from shiny object syndrome.

00:12:26: So how to approach procurement differently?

00:12:28: Their technology investments strictly follow their industry & business strategy.

00:12:33: If a company's growth strategy relies heavily on product-led growth, their stack is going overindexed.

00:12:39: heavily on Product Telemetry Tools and in-app messaging Makes sense.

00:12:43: But if they are focused on enterprise account based marketing Their architecture prioritizes deep intent data integration and direct mail routing.

00:12:53: The stack is a deliberate engineered reflection of how they intend to win in their specific market,

00:12:59: not just a Frankenstein monster or whatever software had the best sales pitch that quarter right?

00:13:05: I can hear skeptics pushing back on this though like it sounds like we're treating Martek like complex urban city planning instead.

00:13:14: Sometimes a team just needs to send out a webinar and invite quickly.

00:13:17: Aren't we over complicating things?

00:13:19: It's a fair concern, but to lean into your urban planning analogy imagine you spend a billion dollars building a massive state-of the art stadium which is essentially what buying a massive enterprise marketing platform is like.

00:13:32: sure But to save time, you don't build any roads to it.

00:13:35: You don't the integrations of pipelines and governance rules.

00:13:39: The stadium is completely useless because nobody can get there.

00:13:42: I see

00:13:43: Juan Mendoza shared some alarming research showing that forty percent of enterprises take what he calls a faith-based approach.

00:13:52: That is a terrifying phrase for a CFO to hear.

00:13:55: It really is, it means forty percent of organizations literally cannot demonstrate the commercial value of their marketing technology in actual financial terms.

00:14:05: they're spending millions just hoping it works on faith.

00:14:08: and that happens when you ignore The Architectural Foundation

00:14:11: which leads us to the final an arguably most critical piece of this puzzle.

00:14:15: let's say You've Done the Hard Work.

00:14:17: You've fixed the architecture, engineered the rich context.

00:14:20: You have autonomous agents ready to execute.

00:14:23: how do you ensure these agents don't hallucinate and destroy your brand?

00:14:26: Or

00:14:27: blow through your entire quarterly ad budget while you sleep

00:14:30: exactly as AI takes on actual execution?

00:14:33: The biggest hidden cost in risk for B to be marketers right now is this sheer lack of governance an observability.

00:14:40: This is the unglamorous reality of AI that vendors definitely don't put on their landing pages.

00:14:45: Greg Kilstrom highlighted this beautifully, teams are rushing to buy AI for speed but skipping the governance paperwork and it results in a massive correction tax.

00:14:54: What does that correction tax actually look like?

00:14:56: In a day-to-day

00:14:57: workflow It looks like human marketers spending six point four hours every single week just fixing bad AI output.

00:15:04: It's spending three hours rewriting an AI-generated email because the model hallucinated a product feature.

00:15:10: Ugh,

00:15:10: painful!

00:15:11: Kielstrom even cited a chilling example of a retailer that lost over sixty two thousand dollars in margin on a single automated e-mail send.

00:15:19: Sixty

00:15:19: Two Grand just from one E-mail?

00:15:21: Yes

00:15:22: Because... An autonomous agent decided to optimize conversion rates by pulling a stale deep discount rule from an old database.

00:15:32: Nobody had ever written down the explicit governance rules of what the agent was and wasn't allowed to

00:15:37: do.".

00:15:38: That

00:15:38: kind of mistake is a career-ending risk for a CMO, it proves that when your stack begins making autonomous decisions traditional application monitoring falls completely short.

00:15:47: knowing if the server is online

00:15:52: introduced a conceptual solution to this.

00:15:54: He argues that marketing teams now need an AI observability control tower,

00:15:58: A Control Tower?

00:15:58: I like that distinction!

00:15:59: Traditional

00:16:00: monitoring watches the uptime.

00:16:01: A control tower watches the decisions.

00:16:03: it acts as an oversight layer actively monitoring campaign agents.

00:16:07: they select audiences and watching optimization agents tweak bid prices.

00:16:11: So it slags anomalies?

00:16:13: Yes To prevent runaway compute costs or explicitly stop agent from making an out of bounds decision Imagine a dashboard that flags you and says Agent A is attempting to increase keyword bidding by four hundred percent, approve or deny.

00:16:27: And we are finally starting.

00:16:29: see software companies build this kind of operational oversight directly into their platforms.

00:16:34: Thomas Scott Adams give great example.

00:16:37: regarding Adobe Workfront They recently made their AI collaborators generally available.

00:16:41: These

00:16:41: are the agents designed to handle the operational overhead, right?

00:16:44: Status updates, task routing...

00:16:46: Exactly!

00:16:47: The endless administrative busy work that burns teams out.

00:16:50: But the key differentiator is these agents operate entirely inside a governed visible and auditable system.

00:16:56: So

00:16:56: it's not a black box

00:16:58: No!

00:16:58: Human managers can look at the system logs And see exactly what the AI did when they make this change And logic behind why it routed tasks in certain ways.

00:17:06: Wait, I have to challenge this a bit.

00:17:08: Thinking about the agility we talked earlier.

00:17:10: if we demand strict audit trails on a control tower before an AI agent can even update as simple project status aren't we just reintroducing that exact same IT bureaucracy?

00:17:21: We are trying to escape.

00:17:22: It definitely feels like a paradox but Phil Crum made a brilliant counterpoint when discussing webflow source.

00:17:29: He argues that counterintuitively guardrails actually give you agility.

00:17:33: Really?

00:17:34: How

00:17:34: so?".

00:17:35: In a complex enterprise, if you don't have guardrails everyone is terrified of the blast radius.

00:17:47: They are going to move incredibly slowly.

00:17:51: That makes sense, but when you have a heavily governed system where everyone knows the AI physically cannot exceed certain boundaries that fear is completely removed.

00:17:59: The guardrails turn the paralyzing fear of a massive mistake into the confidence to move.

00:18:04: fast.

00:18:05: structure actually creates the freedom.

00:18:06: I love that

00:18:07: exactly.

00:18:09: Now we spent this whole deep dive talking about how to build these incredibly intelligent agentic systems engineering deep context, fixing architecture establishing governance all to perfectly predict and automate customer behavior.

00:18:23: But I want to leave you with a provocative thought to mull over building on an insight from Yeogita Wadhwa.

00:18:30: what if the ultimate evolution of personalization in B to be marketing isn't about better prediction at?

00:18:36: What if it's not about AI agents getting smarter at guessing what a buyer wants to read next?

00:18:58: That's a fascinating question.

00:18:59: to end on.

00:18:59: If you enjoy this episode, new episodes drop every two weeks!

00:19:02: Also check out our other editions of Field Marketing, Channel Marketing and Partner Ecosystems, AI & B-to-B, GoToMarket, ABM & Social Selling.

00:19:10: Thank You so much for joining us in this deep dive And remember to subscribe.