Best of LinkedIn: MarTech Insights CW 30/ 31

00:00:00: provided by Thomas Allgaier and Frennis based on the most relevant LinkedIn posts about MarTech in calendar weeks, thirty-and-thirty one.

00:00:07: Frennis is a BDB market research company that supports enterprise marketing teams in unlocking the full potential of their customer data with the help of AI.

00:00:16: you can find more info in the description

00:00:18: right.

00:00:18: so well actually imagine cutting a check for like nine million dollars every single year for the privilege of copying your own data from one server to another.

00:00:29: I mean, it sounds completely absurd right?

00:00:31: But that's not a hypothetical.

00:00:33: no It's very real frustration right now from a fortune-one thousand CTO.

00:00:36: Yeah And um honestly perfectly encapsulates why The MarTech stack as we know is just like buckling under its own weight.

00:00:43: yeah We are looking at a fundamental rewiring Of how enterprise marketing operates.

00:00:47: the tolerance for you know buying software Just to say have but it's essentially at zero Right Now

00:00:52: exactly.

00:00:53: So today in this deep dive, we're dissecting three massive shifts from the latest industry discussions.

00:01:00: We've got a brutal C-suite crackdown on stack strategy A total architectural rethinking of customer data platforms and then The very sobering mechanics of getting agentic AI to actually function

00:01:13: without completely breaking your brand Which is the key part

00:01:16: right.

00:01:16: so let's start with that crack down for a long time.

00:01:18: The prevailing strategy was just you know accumulation by the new tool, plug it in and figure out that use case later.

00:01:25: Yeah...the growth at all costs era.

00:01:27: but based on research we're seeing finances finally stepping in.

00:01:30: They are demanding actual accountability now

00:01:33: and the scrutiny is just unprecedented.

00:01:35: I mean, Vrog Arderall recently highlighted this McKinsey survey of uh five hundred twenty one global marketers And it visualizes this problem perfectly.

00:01:43: Oh yeah The data on how success has measured It's a glaring disconnect.

00:01:47: So they found that roughly seventy percent of CEOs Are judging marketing based On hard revenue and margin Right?

00:01:52: The actual money?

00:01:52: Exactly.

00:01:53: but only about thirty-five percent Of CMOs are actually tracking those same metrics for themselves.

00:01:57: Wait, so if the person signing your checks is measuring you on revenue how are only a third of marketing leaders even tracking it?

00:02:05: It's

00:02:05: wild

00:02:06: isn't it!

00:02:07: Okay.

00:02:07: It's like buying a Formula One car for your team but the CEO is judging you on your lap times while you're only tracking.

00:02:14: You're completely misaligned on the physics of the race.

00:02:20: I love that analogy and that misalignment, it's exactly the mechanism causing these massive AI initiatives to just stall out.

00:02:29: they get stuck at this basic surface level adoption

00:02:31: because you can't prove The ROI

00:02:33: right if you approach the CFO to fund an AI expansion But your whole justification is built around like campaign velocity or content output?

00:02:41: The conversation Just dies in the room

00:02:43: cuz they don't care about output They care about margin.

00:02:45: yes If you can't prove the payback in CFO's language-meaning margin expansion or net new revenue, You're just never gonna get capital required to turn AI into a real structural advantage.

00:02:57: Which honestly explains the incredible mental gymnastics happening behind closed doors right now To protect existing budgets.

00:03:05: I mean there is budgeting model being discussed by Udkarsh Vikram Singh That categorizes every Martek dollar Into three buckets.

00:03:14: Yeah Survive, thrive or jive?

00:03:16: Yes.

00:03:17: It's a harsh reality check.

00:03:19: So survive is your basic infrastructure.

00:03:21: You literally cannot operate without it

00:03:23: Right like you're.

00:03:23: CRM deliverability tools consent management Exactly.

00:03:27: and then thrive Is the strategic layer that compounds in value over time.

00:03:31: Think first-party data architecture unified profiles but The Jive bucket

00:03:35: That's where they really problem lies.

00:03:36: Oh absolutely!

00:03:37: Its expensive experiments.

00:03:39: the AI pilot that never actually shipped to production or you know, that massive dashboard rebuild that took six months and literally zero people log into.

00:03:47: Right

00:03:47: but here is the organizational psychology at play with that jive spending rarely gets cut outright

00:03:53: because people hide it

00:03:54: exactly.

00:03:55: marketing leaders quietly relabeled as thrive on this spreadsheet.

00:03:59: they basically camouflaged the experimental waste as strategic infrastructure just to survive a quarterly budget review

00:04:06: And eventually that camouflage collapses, because the maintenance costs just consume everything.

00:04:13: Mateen Shake makes a really compelling argument here.

00:04:16: he says strategically removing tools is a much more powerful lever right now than adding them.

00:04:22: Because of the hidden tax, we tend to look at software costs as just a licensing fee.

00:04:26: Wait!

00:04:26: The upfront cost?

00:04:27: But under-the-hood every single new platform introduces a massive hidden tax.

00:04:32: you've got API integration maintenance data reconciliation pipelines compliance auditing...

00:04:37: The organizational learning curves alone.

00:04:39: Exactly.

00:04:40: Eventually, the sheer weight of maintaining all those integrations just outweighs whatever marginal capability that Tool brought in first place.

00:04:47: So you end up with a stack that is incredibly capable on paper but completely paralyzed in practice?

00:04:52: Precisely!

00:04:53: Lisa Morrell actually pointed to the new Apex Martech matrix – this was from CMO Council and Martech Drive…to illustrate

00:05:00: Oh, the maturity mapping one.

00:05:01: Yeah

00:05:02: they evaluate a marketing org not just on the technical capabilities They purchased but on an organizational Maturity required to actually turn those tools into business value.

00:05:12: then map them Into four profiles underbuilt over built precision or powerhouse

00:05:18: and Overbuild is The fascinating One here

00:05:20: because it's A direct result of that jive spending Buying advanced capabilities without having the internal data maturity or even a talent to execute on them.

00:05:30: Okay, I get that i understand The instinct to blame this drawing software stack But is it possible?

00:05:36: We're diagnosing the wrong disease here like if an organization Is overbuilt.

00:05:41: isn't that just a symptom of a missing underlying strategy?

00:05:44: That's

00:05:44: very fair pushback.

00:05:45: yeah, I mean there's a metaphor floating around from you've a Raj Agarwal about the Odyssey that captures this perfectly.

00:05:51: Odysseus had a really crude map, but it got him where he needed to go.

00:05:55: if A Smarter Tool arrives say like a hyper-advanced AI routing app and you throw away the old map That's smarter tool just reroutes You based on micro efficiencies.

00:06:05: It saves you four minutes today But it cost you ten years of wandering If you don't actually input your final destination?

00:06:11: You

00:06:11: Just end up going in circles

00:06:12: right.

00:06:13: Are we just letting the algorithms optimize us into a circle because we lost the destination?

00:06:16: It's critical distinction.

00:06:18: The tool is rarely primary failure point, it's missing foundational architecture.

00:06:23: If your overarching destination to truly understand and react to customer no amount of application layer software will fix fundamentally broken data foundation.

00:06:34: You can't orchestrate journey if instruments are lying

00:06:37: because the data itself is siloed or delayed, or duplicated.

00:06:41: Which brings us right back to that staggering nine million dollar quote about copying data.

00:06:46: Oh yeah!

00:06:47: The architectural warfare happening right now over customer-data

00:06:50: platforms.".

00:06:51: The frustration at the enterprise level has just reached a boiling point.

00:06:56: Anthony Rocio shared insights from Fortune.

00:06:58: One Thousand CTOs and the resentment toward traditional packaged CDPs palpable.

00:07:05: Yeah, because when you have a massive enterprise data warehouse like BigQuery all your proprietary customer data already lives there.

00:07:12: Exactly and traditional CDPs require you to copy that data out of your warehouse and ingest it into their proprietary black box just to build an audience

00:07:22: which is insane!

00:07:23: That CTO spending nine million dollars a year wasn't paying for advanced marketing.

00:07:27: they were just paying attacks on data duplication

00:07:30: basically a ransom

00:07:33: For nine million dollars a year, that platform better be predicting the future and physically making my morning coffee.

00:07:39: Seriously!

00:07:40: By the way if you want to make sure your always staying ahead of million dollar architectural shifts like this just take a quick second to subscribe.

00:07:45: so don't miss our future deep dives.

00:07:47: Definitely So...this frustration is driving massive rise in warehouse native & composable CDPs Like growth loop

00:07:54: And mechanical shift.

00:07:56: there is profound right?

00:07:57: Oh, completely.

00:07:58: Instead of copying the data into a new platform, A Composable CDP sits directly on top your existing cloud data warehouse.

00:08:06: It queries it right where it lives—it's zero-copy architecture.

00:08:10: The

00:08:10: competition to own that serving layer is getting intense.

00:08:14: Rapali Krishna highlighted that Databricks is aggressively entering the CDP market now.

00:08:19: Yeah they're building identity resolution segmentation and activation directly in their lake house environment.

00:08:26: They are going head-to-head with Snowflake to become the definitive singular serving layer for customer data.

00:08:32: Okay, I see the appeal of centralizing everything and cutting out the duplication But let's talk about the physics of actually querying a massive data warehouse.

00:08:39: The latency yes

00:08:40: if we were running sequel queries against the petabyte scale lake house Aren't we sacrificing?

00:08:47: Live web personalization.

00:08:49: It feels like having a brilliantly smart Michelin star chef in a kitchen that is simply too far away from the dining room.

00:08:55: it takes Too long to get the food out.

00:08:57: exactly The meal's perfectly customized, but by the time the waiter walks it out to the table The customer has already paid the check and left the building.

00:09:04: That

00:09:04: is a great way to put it And that latency issue is the Achilles heel of a purely composable approach.

00:09:10: Oscar Lopez Cuesta defines this exact problem as the stale attribute payload gap.

00:09:17: The stale attribute gap?

00:09:19: Yeah, if you architect your personalization engines solely on batch loads from a centralized warehouse.

00:09:25: the data processing takes time.

00:09:27: You might inject a thirty minute old attribute into a live user session

00:09:32: and mechanically A thirty-minute lag means you are offering a customer a discount On an item they literally just checked out

00:09:38: with.

00:09:38: Or you're recommending a white paper they just finished reading.

00:09:40: It doesn't look uncoordinated, it breaks trust!

00:09:43: So how do you actually reconcile that?

00:09:45: The need for centralized zero copy data but with the requirement of sub-second edge level activation...

00:09:52: ...the consensus is pointing heavily toward hybrid CDP architecture like what Telium offers For example.

00:09:58: How does this work mechanically?

00:09:59: A hybrid approach bifurcates workload.

00:10:02: It handles the heavy immutable batch data in the warehouse natively, but it simultaneously maintains a lightweight real-time event stream at the edge.

00:10:11: Got it!

00:10:12: So it reconciles the historical deep dive analytics with the sub second profile freshness you need when user clicks button on your website?

00:10:20: Exactly...it gives you best of both worlds.

00:10:22: But architecture only solves speed and storage problem.

00:10:27: What happens when the tracking mechanisms themselves are actively fighting each other?

00:10:31: Ah,

00:10:32: The identity resolution nightmare.

00:10:33: Yes Prachi Jane raised this fascinating technical conflict she calls the hidden Identity problem.

00:10:40: say you have the adobe web SDK and marquardo munchkin running side-by-side on the exact same webpage Which

00:10:46: happens all the time in enterprise stack

00:10:48: all the Time under the hood.

00:10:49: They're issuing entirely separate IDs and dropping parallel cookies into the same user's browser.

00:10:54: And that creates a bifurcated identity graph!

00:10:57: To your analytics platform, one single anonymous human looks like two completely distinct visitors acting simultaneously….

00:11:03: …and when those separate IDs eventually try to merge in the back end... It just scrambles the attribution models.

00:11:10: Your customer journey analytics become mathematically useless because the fundamental premise of who did what is corrupted at that very moment in collection?

00:11:19: It's

00:11:20: two incredibly smart systems creating total chaos, because nobody enforced a universal translation layer Right.

00:11:27: So how do you fix it?

00:11:28: Well fixing this requires an entirely different approach to data modeling.

00:11:32: Devastrita Dutta argues we have to structurally shift from a data-first CD key design, To an audience first design.

00:11:40: Meaning

00:11:40: what exactly?

00:11:41: In the data first approach engineers just pipe every available field into the cdp and try to make sense of it later.

00:11:47: But in an Audience First Approach before A single schema is finalized you define The exact audience You need to activate.

00:11:53: Oh

00:11:53: I see start with the destination.

00:11:55: It maps perfectly back to the Odyssey metaphor From earlier

00:11:58: Exactly.

00:11:59: You trace those specific required attributes backward to their source systems, you define mechanically what fields must be real-time versus what can be processed in batch... ...you build the data model exclusively around the requirements of activation rather than just blind ingestion!

00:12:13: You have to know who want talk before building a plumbing to reach them.

00:12:18: And this pristine sub second data environment is no longer nice for better email targeting.

00:12:24: it's the absolute prerequisite for the final shift we need to dissect today, which is agentic AI and orchestration.

00:12:32: Because if you feed scrambled identity graphs into autonomous AI agents...

00:12:37: You aren't just sending a bad email!

00:12:38: You are orchestrating a real-time brand disaster.

00:12:41: A hundred percent The shift from execution to orchestration.

00:12:48: We are moving away from building linear channel specific batch campaigns and moving toward unified ecosystems that respond instantly to customer intent.

00:12:56: Yeah, I actually saw Nuri Duff point out a very revealing new job title emerging in the market head of Customer Experience Orchestration.

00:13:04: That title alone tells you everything it proves.

00:13:07: The industry realizes that static campaigns or dead The future is defining parameters and letting algorithms orchestrate the next best action across every touch point in real time.

00:13:17: But letting agents orchestrate experience requires absolute microscopic context.

00:13:24: I mean AI's an intuitive right?

00:13:26: It is ruthlessly literal.

00:13:28: Chris Treviter shared a story that should honestly terrify anyone rushing into this blindly.

00:13:33: A HubSpot customer was paying for fifty six thousand monthly breeze credits.

00:13:38: They spun up an AI prospecting agent to go out and find leads.

00:13:41: And let me guess it didn't find any.

00:13:43: It was sourcing exactly zero contacts per day.

00:13:45: a total failure to execute

00:13:47: Exactly, but the reason it failed is what matters here?

00:13:50: Yeah wasn't a software bug.

00:13:51: The company's persona data was actually incredibly rich But the literal fields for their ideal customer profile like the ICP job titles were left entirely blank in the configuration.

00:14:01: Oh

00:14:01: wow yeah

00:14:03: We are so used to software giving us allowed red error code when something is broken.

00:14:08: AI agents don't fail loudly, they fail quietly.

00:14:11: They act on empty configuration fields.

00:14:13: you forgot about six months ago and will confidently do nothing while burning through your API credits.

00:14:18: That quiet failure is the defining risk of agent marketing.

00:14:22: An agent doesn't possess common sense, it only possesses the context window you provide.

00:14:28: And because of a rogue agent is so high we are seeing vendor platforms physically build governance directly into user interface to restrict what AI can mechanically access.

00:14:38: Okay but how does that actually work in practice?

00:14:41: How do you leash an autonomous agent?

00:14:42: You restrict its API permissions at server level.

00:14:46: Andrea Veggiani noted that platforms like Brace and Showpad are now explicitly publishing what their remote model context protocol or MCP servers do not expose to AI agents.

00:14:57: So it's security by limitation?

00:14:58: Exactly!

00:14:59: For example, Bloom Reach restricts its agents so they can only pull from existing pre-approved voucher pools.

00:15:06: The agent mechanically lacks the API endpoint required to hallucinate and generate a ninety percent discount code on the fly.

00:15:12: That smart Because always it might just give away the farm to close a deal.

00:15:16: Right,

00:15:17: and Lokanath Charterie ran extensive tests on Workfront's new content reviewer agent.

00:15:22: It acts strictly as an advisory layer... ...it will catch obvious brand slips in tone issues but The system architecture refuses to let the Agent act As the final approval gate.. ..It cannot physically publish the asset!

00:15:34: The human remains the unavoidable bottleneck.

00:15:37: Yes,

00:15:37: deliberately so.

00:15:38: Okay I understand limiting A single agent inside of Single Platform But the whole promise of orchestration is cross-platform fluidity.

00:15:47: Say an agent in your CRM hands off a task to an Agent In Your Content system, and that agent hands it into third agent on your ad platform.

00:15:54: If something goes completely rogue at two AM on day three across three different vendor ecosystems How On Earth do you know which human authorized back on Day One?

00:16:03: Now you're hitting on the exact technical hurdle that industry is scrambling to solve right now.

00:16:07: Kenny Rajan provided some brilliant insight here by pointing a specific protocol, it's RFC.

00:16:13: eight six nine three.

00:16:15: Okay unpack the mechanics of this for us.

00:16:17: So This Is The Oath.

00:16:18: two point o token exchange standard.

00:16:21: Historically, if a system needed to do something it might use the master API key.

00:16:26: But in an agent chain you can't just pass a Master Key down that line.

00:16:30: Too dangerous.

00:16:31: Way too dangerous!

00:16:32: With Token Exchange when Agent A asks Agent B to perform a task, Agent B doesn't use Agent A's broad credentials.

00:16:39: instead securely trades its current credential with authorization server for newly minted highly restricted downscope.

00:16:47: token specifically one narrow test.

00:16:50: It's like a chain of signed permission flips.

00:16:52: Instead of handing the AI, the master keys to the building it has to go to security desk at every single door.

00:16:58: hand in previous slip and get new one that only opens next specific room.

00:17:02: That is perfect analogy.

00:17:04: And this security desk records exactly who requested original entry.

00:17:07: That exact mechanism creates an immutable traceable delegation chain.

00:17:14: If an agent hallucinates a campaign across three platforms, you don't just see that machine did it.

00:17:19: You can trace the cryptographic permissions back through every single hop to the exact human user who originally initiated workflow.

00:17:26: It solves accountability problem by baking identity directly into API calls

00:17:30: Exactly.

00:17:31: But implementing this level of technical governance completely shatters traditional marketing org chart.

00:17:37: if software is doing executing what are humans even do?

00:17:41: Managing the agents, John Miller shared an incredible case study about DAXCO's CMO Wendy White.

00:17:47: She deployed over eighty distinct AI agents across her marketing organization.

00:17:52: Eighty?

00:17:52: That is a massive fleet!

00:17:53: It

00:17:53: IS.

00:17:54: and to make operation of that scale function without imploding she had pull a VP into dedicated AI center excellence.

00:18:01: but more importantly completely restructured her entire team dismantling old specialist silos

00:18:07: like email person ad person content person gone

00:18:11: Because when the agents handle the channel-specific execution, specialists silos become obsolete.

00:18:17: The bottlenecks in her organization were no longer than manual tasks—they are handoffs and communication gaps between human workers managing their

00:18:25: agents."

00:18:27: Exactly!

00:18:28: So she had to move people into generalist vertically integrated roles where a single human manages a pod of agents from strategy all the way.

00:18:39: Real Longacre actually framed this perfectly.

00:18:42: He calls us the inevitable rise of The Marketing Engineer.

00:18:45: A

00:18:45: marketing engineer like that?

00:18:46: Yeah, we are moving past the era Of the campaign manager.

00:18:50: tomorrow's competitive advantage isn't writing clever copy right.

00:18:53: it's designing the data pipelines architecting the token exchanges and Calibrating the decision engines to actually execute work autonomously.

00:19:00: We've covered a staggering amount of technical ground today From the C-suite fundamentally rejecting experimental jive spending, to intense architectural battles over zero copy versus real time CDPs and stark reality that AI agents require pristine context and cryptographic governance just to function safely.

00:19:21: It's

00:19:22: a lot of process!

00:19:23: But if there is one final forward looking thought I want you walk away with it comes from Scott Brinker.

00:19:28: Okay who's here?

00:19:29: We've spent this entire deep dive talking about the plumbing, data architectures and agents.

00:19:35: And as these AI systems increasingly do actual web browsing, vendor researching or procurement for us The immediate assumption is that traditional visual website...is dead.

00:19:45: Right

00:19:45: like why build a beautiful home page if machine reading

00:19:49: it Exactly?

00:19:50: But Brinker argues the reality will be the exact opposite.

00:19:53: We are going to see a rise of parallel web experiences, yes there But alongside that, because the machines are handling the mundane tasks.

00:20:09: The moments when humans do choose to show up and interact will demand a highly visceral incredibly high fidelity emotional experience.

00:20:19: You aren't just designing for eyeballs anymore.

00:20:21: you're going to have to design digital experiences For cold literal algorithms And deeply emotional human simultaneously.

00:20:28: It's a completely new paradigm for digital discovery.

00:20:32: It always comes back to the human element, doesn't it?

00:20:34: You can build the most advanced autonomous Formula One car in the world with a zero-copy data foundation reading the track in milliseconds.

00:20:42: But at the end of day there is still a human driver who needs actually feel the road.

00:20:48: If you enjoyed this episode new episodes drop every two weeks.

00:20:51: also check out our other editions on field marketing channel and partner marketing AI & BDB.

00:20:56: go to market ABM and social selling.

00:21:00: Thank you so much for joining us on this deep dive.

00:21:02: Don't forget to check your own configurations, make sure that destination is set and we will catch in the next one!