Best of LinkedIn: AI in B2B Marketing CW 15/ 16

00:00:00: This episode is provided by Thomas Allgaier and Frennus, based on the most relevant LinkedIn posts about AI in B-to-B marketing in calendar weeks fifteen and sixteen.

00:00:09: Frenness is a b to be market research company that supports enterprise marketing teams

00:00:20: And we are diving into something today that is just, well honestly a bit terrifying for a lot of marketing teams.

00:00:26: Right I mean imagine spending fifty thousand dollars to write the ultimate definitive buyer's guide And it is perfectly optimized.

00:00:37: An AI model reads, learns absolutely everything about your product category from it and then you know when a user asks for recommendation that same AI explicitly tells the buyer to go purchase from your biggest competitor which

00:00:49: just

00:00:50: wild.

00:00:51: It's not a hypothetical scenario either...it happening right now.

00:00:53: on the craziest part Is at traditional analytics dashboard won't even show drop in traffic

00:01:05: You know, playing around in the sandbox treating it like a parlor trick to write quick email drafts.

00:01:09: Yeah

00:01:10: exactly.

00:01:10: But looking at the insights curated for today's deep dive that era is entirely over The top.

00:01:16: trends circulating on LinkedIn right now show a massive shift.

00:01:19: We are moving from basic experimentation into serious governed operational execution

00:01:26: And the margin for error is shrinking rapidly.

00:01:29: So to map this out for you today, we are going unpack three major shifts happening in the BDB marketing space right now.

00:01:35: Yep

00:01:36: lots to cover.

00:01:37: first We need a look at discovery This new battleground of AI visibility and why traditional search metrics are breaking.

00:01:43: Then, you know if we assume.

00:01:44: You actually get found will explore agent to go-to-market workflows

00:01:48: right which is basically what happens when AI starts executing your outreach autonomously

00:01:53: exactly.

00:01:54: and finally We'll look at the fallout of all this automation Which is driving a massive return?

00:01:58: To human led brand in content strategy.

00:02:01: Let's start right at the top of the funnel with discovery because I mean If the AI models don't know who you are or worse if they don't trust you None of the downstream marketing even matters.

00:02:12: Totally

00:02:13: Alex Groberman highlighted a massive shift recently when Google's John Mueller essentially confirmed that AI search visibility needs to be treated with the same severity as traditional SEO.

00:02:23: Let's pause on that because for a decade, SEO has been very straightforward transaction.

00:02:28: You target keyword you rank get click and measure website traffic Right.

00:02:32: it was super linear.

00:02:33: But Nancy Pizzacar and Laura Cashin both pointed out that in the world of generative engine optimization, or GEO.

00:02:42: That transaction is completely broken.

00:02:44: Success isn't a click anymore it's measured by retrieval rates.

00:02:47: Yeah...that's huge distinction.

00:02:49: I want to dig into mechanics.

00:02:51: What exactly does a retrieval rate mean for marketing team looking at their reporting?

00:02:55: Well, a retrieval rate measures whether an AI system actually pulls your brand's entity into its generated answer at all.

00:03:03: In the past Google gave the user ten blue links and the user chose but now the AI model synthesizes an answer directly.

00:03:11: so if a buyer prompts in AI with you know what are the best enterprise CRM tools?

00:03:17: And the AI writes a comprehensive paragraph that doesn't mention

00:03:21: Your retrieval rate for that query is zero.

00:03:23: Exactly, zero!

00:03:24: You don't get a chance for a click because you weren't even included in the synthesized reality that AI just presented to the buyer.

00:03:30: Okay

00:03:30: let's unpack this.

00:03:31: If traditional SEO was like trying to build the brightest tallest billboard on the highway Is AI visibility more like... Trying to be invited into right exclusive dinner parties?

00:03:43: How do actually get the AI talk about you behind closed doors?

00:03:47: That is phenomenal way of putting it And someone, Yosef broke down exactly how you build the structural trust to get that invite.

00:03:53: Okay?

00:03:53: How so

00:03:54: he pointed out that AI models don't read web pages The way humans do.

00:03:58: they look for structured data specifically schema markup and knowledge graphs.

00:04:02: Wait

00:04:02: can You clarify how That actually changes the model's behavior because Schema mark up sounds like a back-end Web development task not A strategic marketing initiative.

00:04:10: it definitely Sounds technical but It is deeply strategic.

00:04:14: if You just have plain text on a page the large language model has to probabilistically guess what your content is about.

00:04:21: It has to infer context,

00:04:23: right?

00:04:23: But schema markup gives the AI hard structured labels.

00:04:27: It explicitly says this string of text is a product, This number is price, This bullet point Is feature.

00:04:34: Ah so it removes the AIs need to guess?

00:04:36: Exactly and when you remove the guesswork You increase the mathematical confidence The model has in your data.

00:04:42: You transition from being a random String Of Text on the internet To becoming a verified entity That the AI structurally understands And trusts.

00:04:51: So you're effectively predigesting the information for the LLM, so it doesn't have to spend computational power figuring you out.

00:04:57: That's a great summary!

00:04:59: And Maja Vojta recently shared a highly practical six-step AEO playbook that builds on this.

00:05:04: She noted that semantic headings and direct answers are critical.

00:05:07: Yeah...that makes sense.

00:05:08: But the most fascinating data point she shared was about content freshness.

00:05:13: According her research, eighty three percent of AI citations come from pages updated within last twelve months.

00:05:19: Wait, eighty three percent.

00:05:20: So if my company's definitive buyers guide or get a competitive comparison page was published eighteen months ago and hasn't been touched.

00:05:29: The AI essentially treats it as obsolete.

00:05:31: Yes, exactly because these models are constantly being updated via retrieval augmented generation or our group to pull in the most current context Wow if you're a competitor update of their page last week?

00:05:43: The AI considers there data more reflective of the current reality and You know It's not just your own web properties that they I is pulling from

00:05:50: right.

00:05:51: Antoine Navette shared some remarkable data showing a massive surge in LLMs citing social platforms.

00:05:56: Between March and April, citations of Reddit jumped from under three hundred to over eleven hundreds.

00:06:01: That has huge jump!

00:06:03: Yeah...and LinkedIn's citation surged significantly as well

00:06:06: Which honestly makes a lot of sense when you think about buyer behavior.

00:06:09: Buyers go to Reddit and LinkedIn because they want unvarnished, human vetted natural language peer reviews

00:06:16: Exactly!

00:06:16: They don't want corporate marketing gloss.

00:06:19: The AI is simply mirroring that demand for authentic consensus

00:06:24: which brings us to the dark side of this new search landscape and the scenario you brought up at the very beginning of The Deep Dive.

00:06:31: Johnny Naster has been tracking a phenomenon he calls ghost rankings,

00:06:35: oh right!

00:06:36: And it is a massive threat to content marketing ROI.

00:06:40: Yeah...the fifty thousand dollar buyer's guide that sells the competitor product Right.

00:06:44: Why does actually happen?

00:06:46: How can an AI read my content cite as source but then recommend someone else.

00:06:52: While Nassar used a brilliant example of tracking twenty-five direct to consumer brands, he found that a tool like Proplexity will cite a brand like Traeger specifically their Beginner's Guide To Grilling...to formulate its answer on how to grill.

00:07:05: Okay!

00:07:05: That makes sense for the informational part

00:07:07: Right But when it comes to transactional recommendation what to buy?

00:07:11: It recommends Weber.

00:07:13: This happens because the LLM categorizes your content as an informational entity, but categorizes you competitor a superior transactional entity.

00:07:22: That feels incredibly parasitic!

00:07:24: If an AI engine is just going to scrape my top of funnel educational content use it to train its own answers and then route actual purchase intent into my competitor.

00:07:34: shouldn't B-to-B marketers block AI crawlers entirely?

00:07:38: I mean...I get that reaction.

00:07:39: Why even play game where house steals chips?

00:07:42: It's a valid point, but blocking the crawlers is essentially career suicide in this new environment.

00:07:47: If you block the bots—you aren't even the informational source anymore!

00:07:50: You've vanished

00:07:51: completely.".

00:07:52: So what's the play then?

00:07:53: The strategic play isn't to hide your data — it's to ensure your off-site reputation.

00:07:58: those Reddit and LinkedIn mentions Antoine Navet talked about aligns with on site schema so that AI trusts for both information and transaction.

00:08:07: Okay, so let's say we get our schema fixed.

00:08:09: We update our content every six months and the AI models finally start serving us up in their synthesized answers.

00:08:15: The discovery problem is solved right.

00:08:18: what happens next?

00:08:19: How do we capture that demand without just throwing massive headcount at it?

00:08:22: yeah That leads us into the second major theme.

00:08:25: were seeing a gentic go-to market execution.

00:08:28: This is where things move from passive visibility to active outreach, and the speed of innovation here is just staggering.

00:08:35: Hans Decker recently broke down instantly .ai's new sales agent.

00:08:40: Ooh!

00:08:40: The workflow for that is wild It really

00:08:42: is.

00:08:43: You drop in a target website URL And within thirty seconds this autonomous agent reads the site identifies your ideal customer profile enriches lead data writes personalized messaging handles follow-up cadence attempts book meeting

00:08:56: A thirty second turnaround for an end-to-end sales motion.

00:09:00: Here's where it gets really interesting, if everyone can buy these tools for twenty dollars a month are we just buying over glorified send buttons?

00:09:07: Yeah If all have the same AI Where is actual competitive advantage?

00:09:11: That Is Exactly The Hidden Danger and Keith Salons shared a cautionary tale that proves your point perfectly.

00:09:17: He deployed an AISDR that sent out forty-two hundred supposedly highly personalized emails over a six day period.

00:09:25: Wow!

00:09:26: That's a lot of volume On

00:09:27: the surface, The Metrics looked amazing.

00:09:30: Let me guess...the deliverability crashed

00:09:32: Spectacularly.

00:09:34: It was complete failure And the reason why it failed is crucial to understand.

00:09:39: The underlying data that AI was feeding on, was fundamentally flawed.

00:09:43: Hell so!

00:09:43: Well

00:09:44: the agent was instructed to pull personalization from the prospects LinkedIn activity but they didn't distinguish between a prospect's original post and a simple repost.

00:09:54: Oh wow So you have prospects getting emails saying hey loved your deep insights in the future of supply chain?

00:09:59: And the prospect sitting there thinking I didn't write that, i just clicked repost on a company update three weeks ago.

00:10:05: Precisely!

00:10:06: It instantly destroys trust.

00:10:08: and because the AI was using the same underlying sentence structures and syntactical cadences across thousands of emails gmail's algorithms caught on almost immediately.

00:10:16: Of course

00:10:17: they did.

00:10:18: Yeah.

00:10:18: by day three The campaign was severely throttled.

00:10:21: By Day four it banished to promotional and spam folders.

00:10:24: This highlights really critical flaw in how teams are viewing AI.

00:10:29: If everyone has access to the exact same large language models, The AI itself is essentially a commodity.

00:10:35: So where's the actual competitive advantage?

00:10:37: if I'm an enterprise B-to-B marketer what is my moat?

00:10:41: What's fascinating here Is that the advantage isn't the AI tool but data architecture underneath it.

00:10:47: Carolyn Healy and Oren Greenberg both hammered this point.

00:10:50: home Agents do not just read data, they take autonomous action across your systems.

00:10:55: If you have a fragmented CRM outdated lead scoring or broken UTM parameters deploying an Autonomous Agent doesn't fix the process.

00:11:03: It

00:11:03: probably makes it worse right?

00:11:04: Exactly!

00:11:05: Its like hooking up high-speed automated assembly line to warehouse that is completely unorganized.

00:11:11: You aren't creating efficiency...you are producing garbage at ten times speed.

00:11:17: Greenberg noted that ninety four percent of go-to market teams claim to be using AI, but a massive portion of them are seeing zero commercial return because their underlying data is a mess.

00:11:29: That makes Gabe Rogel's observation incredibly timely.

00:11:33: He pointed out that platforms like demand-based AI are pivoting away from just building more generative text

00:11:39: tools.

00:11:39: Right, they're focusing almost entirely on unifying go to market data.

00:11:43: Exactly!

00:11:44: They were trying to build an intelligence layer.

00:11:46: so before the agent takes action it actually has accurate historical context

00:11:51: And once you have that context You change how you interact with model.

00:11:55: Ryan Staley made a brilliant point about how top tier teams are operating.

00:11:59: Wait,

00:12:00: what do you mean?

00:12:00: they aren't typing Cromsets?

00:12:02: You know how use the tools.

00:12:03: Not if you want operational leverage.

00:12:05: Staley argues that If you are typing the same instructions into an interface more than three times your process is broken.

00:12:12: Oh I see

00:12:13: The most advanced teams Are building permanent skills directly Into platforms.

00:12:17: like Claude They're taking the complex multi-step logic of their best performing sales rep How to analyze a transcript how they identify a pain point, how they map it to specific feature and baking that logic into the system's architecture.

00:12:33: So you do heavy strategic lifting once.

00:12:35: then the system executes automatically at scale?

00:12:38: Exactly!

00:12:39: But there is massive trapdoor hiding underneath all this automated execution.

00:12:43: we have talk about tokenomics.

00:12:45: Ricky Salanke and Liza Adams brought up an issue that is flying completely under the radar for most marketing & finance teams.

00:12:51: Oh, definitely!

00:12:52: When you buy traditional SaaS software You are used to a flat predictable monthly subscription.

00:12:57: Yeah But eGenic AI workflows introduce highly variable consumption-based costs

00:13:03: because every single time The AI has to think or reason through a problem Or run background step to navigate file system it consumes tokens.

00:13:10: Exactly.

00:13:11: It's not like buying an unlimited Metro pass for the month, it is much more like paying a taxi.

00:13:16: but the meter runs faster every single time that driver has to stop pull over and look at map.

00:13:23: figure out next turn.

00:13:24: That's great analogy!

00:13:25: Glankey shared mind blowing example about Uber.

00:13:28: They gave five thousand of their engineers access AI coding tools in December.

00:13:33: By April just four months later Their entire annual budget was completely incinerated.

00:13:39: Wow not because the tools failed, but because they worked so well that usage skyrocketed.

00:13:45: And Liza Adams added some crucial context to that.

00:13:47: depending on this specific model you use whether it's GPT-IV, cloud three opus or an open source model The token cost for exact same task can swing by a factor of ten to thirty times.

00:13:58: Ten to thirty time?

00:13:59: That is massive!

00:14:00: Yeah

00:14:00: If you have autonomous agent iterating on lead enrichment in background while your asleep and gets caught in logic loop It absolutely blow your quarterly budget out water.

00:14:09: You have to track unit economics at a microscopic level now.

00:14:12: Which brings us to a fascinating crossroads.

00:14:15: and our third, final theme We've established that autonomous agents can burn through your budget in the weekend And they completely ruin your center reputation if you're data is bad.

00:14:26: All of this high volume AI output Is COMPLETELY useless If core message is hollow!

00:14:32: And we are seeing massive human-led rebellion against blind automation.

00:14:38: It's a necessary correction, honestly.

00:14:40: CellConca really nailed this when observing the current state of AI-generated advertising.

00:14:45: Yeah what do you say?

00:14:46: He pointed out that these Ai ads often look incredible The imagery is polished, the copy's grammatically flawless.

00:14:52: But they are failing to convert at a fundamental level because they lack real customer

00:14:56: insight.".

00:14:57: The hooks might technically follow a copyrighting formula but don't actually say anything meaningful.

00:15:01: Exactly!

00:15:01: They lack strategic depth that makes buyers stop scrolling.

00:15:04: I have case study which perfectly illustrates the antidote for this generic polished garbage.

00:15:09: Jonathan MK shared massive success story from a company called Thrive Stack.

00:15:14: Like everyone else...they tried fully automated AISDR outbound approach

00:15:19: And predictably, their open rates tanked right?

00:15:22: Yep.

00:15:22: Their deliverability suffered and it was a total waste of money.

00:15:26: but instead trying to optimize the prompts to make this spam slightly better they killed the program entirely.

00:15:32: That takes guts.

00:15:33: I mean walking away from the promise.

00:15:35: infinite scale is hard.

00:15:37: what did they pivot too?

00:15:38: They pivoted into education...they built an AI-assisted newsletter.

00:15:43: But instead of using AI to generate cold pitches, they used it to scale the creation of incredibly high-value deep dive content.

00:15:50: Okay that's smart!

00:15:51: Yeah They put prospects on an eight week nurturing track for two solid months...they didn't ask for a meeting..They just educated broke down industry problems and provided frameworks.

00:16:00: Wow by the time a prospect finally talked to sales?

00:16:03: They were already fully educated on the methodology.

00:16:06: The result was a staggering five hundred fifty-one percent ROI.

00:16:11: They stopped yelling at cold audiences and started building the relationship?

00:16:14: If we connect this to the bigger picture, well this raises an important question... Is the craft of marketing actually more critical now than ever?

00:16:22: It sounds like AI is a giant megaphone but if your core positioning is weak you're just broadcasting your own irrelevance at

00:16:29: scale!

00:16:30: Exactly You are just making sure everyone in your industry knows how little value you offer faster than ever

00:16:36: before.

00:16:36: Yeah, precisely.

00:16:38: Louis Cho shared a story that proves the value of human context perfectly.

00:16:42: He was pitching a new AI marketing workflow to his CMO who is deeply skeptical.

00:16:48: The CMO believed their brand's tone and technical complexity were too nuanced for large language models.

00:16:55: How did you overcome that?

00:16:56: Because, That is a very common objection in Enterprise B to B. He

00:16:59: proved the CMO wrong In exactly eleven minutes.

00:17:02: But The secret wasn't some magic prompt.

00:17:04: The Secret was the context.

00:17:06: he fed the model

00:17:07: Right...the data architecture.

00:17:08: again.

00:17:09: Cho didn't just open Clawd and say, write a campaign brief for complex software product.

00:17:14: He uploaded their highly specific brand guidelines.

00:17:17: he uploaded the performance data of there three best historical campaigns.

00:17:22: .He Uploaded The raw transcripts of actual customer interviews And Their internal tone-of-voice documents.

00:17:28: so he essentially built A temporary knowledge graph For the model.

00:17:32: only then did he ask for the brief.

00:17:35: When the CMO read it eleven minutes later?

00:17:37: He was stunned.

00:17:38: He said, this sounds exactly like us.

00:17:42: So what does all mean?

00:17:43: It's

00:17:43: the perfect illustration of what Kirstie Nunes has been arguing.

00:17:46: She pointed out that generative AI operates purely on probability not truth.

00:17:50: Right it doesn't actually understand.

00:17:51: your brand is simply predicting The next most likely word based on its training data.

00:17:56: Therefore human intent and clarity of strategic thought And hyper specific context you provide are only things matter.

00:18:03: If you feed a generic input it mathematically have to give you generic output

00:18:07: And the major platforms are actively enforcing this reality now.

00:18:11: Sophie Carr observed that Google's massive core update in March specifically targeted and penalized mass-produced, unreviewed AI content.

00:18:21: Yeah they just wiped it out of search results.

00:18:23: But at the exact same time They heavily rewarded content That demonstrated human validated expertise Original research & real world experience.

00:18:33: The algorithms are getting significantly better at filtering out the noise.

00:18:37: Which

00:18:37: brings us to a really grounding insight from Franz Reimersma, he noted that in the history of marketing technology and data have never actually been the core problem...the problem has always been human ability to extract true customer's voice from sea-of-data.

00:18:52: That is ultimate differentiator.

00:18:54: Yeah an AI can synthesize one thousand transcripts but it cannot feel the emotional friction of buyer experiences and it can not make the strategic trade-offs required to build a distinct brand position.

00:19:06: It is incredibly validating here that human element isn't becoming obsolete, its actually become the premium layer on entire technology stack.

00:19:13: Absolutely You use AI for heavy lifting data synthesis schema structuring rapid iteration code yeah positioning empathy decision what say?

00:19:25: That is where the marketer actually earns their keep.

00:19:27: The winning model isn't AI replacing the marketing department, it's human-led AI... ...where systems scale execution under very rigid Human commercial direction.

00:19:37: We've covered an immense amount of ground today.

00:19:39: we moved from the shifting mechanics of AI discovery and schema to The hitting risks in token costs of autonomous agents all the way To protecting the human craft of the message itself.

00:19:49: it's a lot to process It is

00:19:51: but before we wrap up I want to leave you with one final thought to mull over.

00:19:55: let's hear.

00:19:56: I want to circle back to Antoine Navet's data about how AI models are increasingly pulling their citations from platforms like LinkedIn, Reddit and community forums.

00:20:06: If these models using unvarnished peer conversations as a primary source of truth well what is your company offsite footprint currently teaching the AI about you brand when aren't in that room?

00:20:17: If a model scrapes Reddit right now to learn about your product, what answer is it going?