Best of LinkedIn: AI in B2B Marketing CW 41/ 42

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

00:00:07: In calendar weeks forty one and forty two.

00:00:10: Frenus is a b to be market research company that supports enterprise marketing teams unlocking full potential of their customer data with help from AI.

00:00:19: So imagine firing like seventy five percent outbound sales reps replacing massive gap strictly engineered data pipeline and then watching your qualified pipeline actually double in just ninety days.

00:00:34: Which sounds completely made up, but it's not?

00:00:36: It is NOT!

00:00:37: Welcome to the deep dive.

00:00:39: If you're sitting there right now trying to figure out how navigate your B-to-B marketing strategy over this next year You are absolutely IN THE RIGHT PLACE.

00:00:47: We are looking at a massive massive shift today.

00:00:51: Yeah A fundamental rewiring really

00:00:53: Exactly So based on our curated stack of sources These are LinkedIn posts from industry leaders Actually operating in the trenches, we are officially moving past the era of prompt experimentation.

00:01:06: It's not about generating a clever image anymore or having an AI write a quick blog post.

00:01:12: We're entering the phase of true operational integration

00:01:15: because the market is demanding actual architecture now.

00:01:18: Not just you know party tricks And for the past two years I mean marketers have basically been treating AI like very fast slightly unpredictable chatbot.

00:01:29: You ask it a question you give it a prompt and it spits out a piece of content.

00:01:32: Yeah,

00:01:33: just task by task

00:01:34: exactly.

00:01:35: But the shift happening in the marketing stack today is moving toward delegation.

00:01:40: It's what the industry is calling agentic marketing.

00:01:43: You weren't asking a tool for a favor anymore?

00:01:47: Your handing a highly specific blueprint to a digital worker essentially And expecting it to execute a continuous process.

00:01:55: And Neri Kankhaya actually put forward a framework that I think really breaks down the mechanics of this shift.

00:02:01: He argues, if you want to give an AI true agency within your organization You have combine three very specific elements skills tools and context

00:02:10: Context being crucial one there.

00:02:12: Right, because most teams stop at the first two.

00:02:14: Like if you have an LLM that can write well...that's a skill!

00:02:17: If you connect it to your CRM or your CMS- That is a tool.

00:02:21: But if you stop right there You essentially have a hypercaffeinated intern just running around your database.

00:02:27: Yeah with absolutely no idea what your company actually sells

00:02:30: Or who your buyers are.

00:02:32: The context So the guardrails and brand voice And historical performance data..That Is What Actually Gives That Agent The Ability To Act On Your behalf Without you know, causing a total disaster.

00:02:43: And we are seeing teams building incredibly complex contextual architectures to manage exactly this.

00:02:50: Johnny Lee and Emma Ridley shared a playbook recently where they deployed the stack of forty different marketing AI agents.

00:02:56: Forty?

00:02:57: Yeah, forty.

00:02:58: but they didn't just dump them all into a single workflow.

00:03:01: They divided them up.

00:03:02: so highly specialized teams have ten across Google ads meta adds SEO an AI search.

00:03:09: But the setup is what really caught my attention there.

00:03:11: They didn't just flip a switch and let forty bots rewrite their ad copy overnight, they built twenty-four specific non-negotiable guardrails into this system before day one even started.

00:03:22: That strict governance has everything?

00:03:24: It really does!

00:03:24: We're talking strict spend caps, mandatory read only access for certain functions... And this immutable change log so human operators can trace every single decision that agent makes.

00:03:35: And when you put those constraints in place, the operational efficiency is just staggering.

00:03:40: Right because it shifts the entire focus!

00:03:42: It's not automation for the sake of speed anymore...

00:03:45: Exactly I mean they deployed one specific agent- Just a single click loss agent and its only job was to monitor their blog In OnePass, just analyzing historical traffic against current anomalies.. ..it found one thousand seven hundred three lost clicks across ten pages.

00:04:03: See that is automation with extreme prejudice regarding metrics.

00:04:07: And Tim Lane over at RTL Equipment shared the underlying structure of his setup and I think it perfectly illustrates this exact discipline, he operates his marketing engine with ten named AI agents.

00:04:20: Wait, named?

00:04:22: Yeah like they have actual personas.

00:04:24: yeah actual personas.

00:04:26: in the system for instance He has an agent name Penny And Penny's entire existence revolves around auditing the ad account against eighty-three specific checks every single week.

00:04:38: Okay, but giving them a name isn't just like a quirky agency culture thing is it?

00:04:43: Not at all!

00:04:44: It forces a structural requirement because if an agent has a name... ...it requires a formal job description.

00:04:51: and once you write a job description for a piece of software You're forced to define the boundaries of what it is explicitly forbidden from doing.

00:04:59: Oh, that makes sense.

00:05:00: Yeah so at RTL equipment none of those ten agents possess publishing rights...none of them!

00:05:06: They can run the eighty-three point analysis they could draft a campaign fixes but everything they generate gets routed into a waiting on Tim column.

00:05:15: So The Human is the bottleneck?

00:05:17: Exactly

00:05:17: By design

00:05:18: Exactly Nothing goes live until A human reviews logic and clicks approve.

00:05:22: it prevents system from just spiraling completely out of control

00:05:25: which brings up Gregoire Chapeleon.

00:05:28: He's running a similar architecture, but he is using sixty cloud code agents specifically to automate ad fatigue and budget pacing.

00:05:35: Sixty Agents just for pacing?

00:05:38: That's intense!

00:05:39: It is...but think about the math.

00:05:40: if campaigns click-through rate drops say twenty percent from its historical peak The agent immediately flags the decay before the cost per acquisition starts climbing and eating into the margin.

00:06:15: job to be done.

00:06:16: He used uberol's UBI agent as an example, it is entirely focused on local marketing monitoring regional listings prioritizing data discrepancies prepping review responses.

00:06:28: so the human marketers role is shifting from execution two management exactly.

00:06:33: you are no longer doing the manual data entry or pulling reports.

00:06:37: your full-time job basically governing output of agents and validating their logic.

00:06:42: okay but let me push back a little bit.

00:06:44: So we're essentially moving from treating AI as a really smart calculator to treating it like a team of junior interns.

00:06:51: But if we don't give them a strict job description, like RTL equipment did aren't they just going to run amok and burn our budgets?

00:06:58: I mean managing sixty agents sounds like an immense amount of engineering overhead for mid market B-to-B Team.

00:07:05: listening to this is the juice really worth the squeeze?

00:07:08: well

00:07:09: If one or those autonomous agents gets confused by data anomaly We are talking about a junior intern formatting a spreadsheet incorrectly.

00:07:17: we're talking about machine potentially bidding ten thousand dollars on completely irrelevant keyword before anyone even notices which is exactly why the architecture shifting toward highly governed task specific nodes

00:07:30: right.

00:07:31: and this introduces in massive strategic pivot.

00:07:34: because if were handing all of our marketing execution an analysis over to these highly structured agents We have to wake up to the reality that B-to-B buyers are doing the exact same thing on their end.

00:07:46: Oh, absolutely!

00:07:47: They're deploying their own bots?

00:07:49: Yeah.

00:07:49: Buyers are handing their vendor research, they're shortlisting... ...they're shopping over to their own AI agents.

00:07:55: So how does a brand actually get found when the buyer isn't human clicking a link but a bot parsing a database

00:08:03: And now brings us directly into this shift from traditional SEO To AEO Answer Engine Optimization and GEO, Generative Engine Optimization.

00:08:14: Oh boy more acronyms to track!

00:08:16: Right but the shift is real.

00:08:18: for the past twenty years we basically optimized websites so Google would rank us as a blue link on a search page.

00:08:24: now We have to optimize our digital footprint So that we become the trusted source material.

00:08:29: an LLM synthesizes when generating an answer for user

00:08:32: And the scale of this transition is really hard to overstate.

00:08:35: Michael Housiak pointed two projections indicating that AI shopping agents could handle ten-to twenty percent... ...of all U.S.

00:08:42: ecommerce by twenty thirty, but here's the fundamental difference.

00:08:47: Bots process structured data.

00:08:50: they do not experience design.

00:08:52: No!

00:08:53: They don't care about your font choices

00:08:54: Not at all.

00:08:55: He brought up a famous Google user test from years ago forty-one shades of blue, where Google's engineers tested exact hex codes to see which specific shade of blue generated the highest click through rate from human users.

00:09:08: Because human beings are psychological buyers we're influenced by visual hierarchy and color theory.

00:09:13: you know...the emotional weight.

00:09:14: a really good hero image

00:09:16: Right, but an AI research agent could not care less about your forty-one shades of blue.

00:09:21: It doesn't feel anything when it reads.

00:09:22: you're emotionally resonant brand manifesto.

00:09:25: When in LLM scans your site to evaluate you as a vendor.

00:09:29: It is looking for hard variables pricing tiers technical specifications compliance certifications and highly structured text.

00:09:36: Yeah if your website relies on heavy visual storytelling But buries the actual data in some unparsable format You simply do not exist to the machine.

00:09:46: You have to spoon feed the bot.

00:09:48: The exact data structures it craves.

00:09:50: It's all about marketing to the machines now

00:09:53: and figuring out how to feed that machine is creating an entirely new technical discipline.

00:09:58: Andre no vice to pala categorizes this as search everywhere optimization.

00:10:04: Connor Gillivan even published a new AI SEO periodic table To track all the exploding acronyms in ranking factors.

00:10:11: Wow, a periodic table for seo now.

00:10:14: yeah

00:10:15: We're talking about optimizing for entity recognition.

00:10:18: You have to ensure the LLM understands exactly what category your software falls into by explicitly linking.

00:10:30: They

00:10:44: absolutely are.

00:10:45: But amidst all these technical frameworks, Vanishkaha Bhargava posted a very sharp warning to the industry that grounds all of this.

00:10:53: She basically said you cannot just use technical hacks To win in an AI-driven search environment

00:10:59: Meaning You can't Just game The system anymore

00:11:02: Right!

00:11:03: You Can add All This Schema Markup You Want And You Could Chunk Your Text Perfectly For An LLM To Read but If You Do Not Possess Genuine Topical Authority Machine Will Just Ignore You.

00:11:13: Generative AI systems have to mathematically calculate who to trust before they assemble an answer.

00:11:19: And she noted that it takes significantly more effort than just, you know, spinning up fifty AI-generated blog posts and calling it a visibility strategy?

00:11:27: You HAVE TO BE AN AUTHORITATIVE ENTITY!

00:11:30: But THAT begs the question – how is that authority actually measured by an

00:11:33: LLM?!

00:11:34: Well, Rand Fishkin & Drew Neiser brought some vital data into this exact conversation….

00:11:39: Right now offsite authority IS THE DOMINANT FACTOR.

00:11:42: AI tools themselves are currently only sending about zero point five percent of actual referral traffic to websites.

00:11:50: Half a percent?

00:11:51: That's almost nothing!

00:11:52: It is almost nothing, that's...

00:11:53: However platforms like Reddit are proving three or four times more valuable for your brand visibility.

00:11:59: Wait really?

00:12:00: Reddit For B-to-B?

00:12:02: Yes and the mechanism behind it is crucial to understand.

00:12:06: LLMs are trained filter out marketing fluff.

00:12:08: They don't just want to read your perfectly optimized landing page.

00:12:12: they wanna know what the market consensus is.

00:12:14: So, they scrape user-generated platforms heavily because that Is where actual humans are discussing their realities of using you product.

00:12:22: Ah!

00:12:23: so LLM weighs third party consensus far heavier than your own self published claims Precisely

00:12:29: Which makes Lee McKenzie's Semrush data incredibly sobering.

00:12:33: The analyze top one hundred brands which consistently mentioned by AI across various industries.

00:12:38: Fewer than half of the brands that the AI mentioned were actually cited as source material for information.

00:12:45: Oh wow!

00:12:45: So, The AI recognizes you as a known entity in this space but it doesn't trust your domain enough to use you as an authoritative source to explain what you do?

00:12:53: Right... It would rather cite third party review site or technical forum to define product.

00:12:58: But if bots just read structured data and ignore all our beautiful branding & emotional resonance Are we destined for a web that is purely utilitarian?

00:13:08: Like are we forgetting the human buyer at the end of the chain.

00:13:11: If bots were doing the searching, parsing structured data instead of experiencing our brand and bots are executing or ad bidding in our content distribution how do actually know this synthetic ecosystem as making the company any real money?

00:13:25: That's

00:13:26: the big question.

00:13:26: Right We can theorize about agentic stacks all day But are these tools driving measurable revenue or we just generating expensive synthetic noise?

00:13:35: Well, the measurement gap in the industry right now is frankly alarming.

00:13:40: Melissa Rosenthal shared data from a recent survey that really puts this into perspective.

00:13:45: over The past year forty seven point eight percent of marketing teams have replaced previously human led tasks with AI.

00:13:53: almost half

00:13:54: but only six point five percent of those teams are actually measuring the ROI of doing so.

00:13:59: Seriously?

00:14:00: Less than ten per cent, So they're outsourcing critical workflow to a machine and can't confidently tell their CFO if it's actually protecting the margin.

00:14:08: Yep

00:14:08: We basically buying high-performance sports car taping over the speedometer just hoping we don't hit wall giving AI free pass on attribution that would never tolerate from traditional media buy or SaaS vendor.

00:14:21: And that lack of scrutiny is incredibly dangerous because the models are far from infallible when it comes to strategic math.

00:14:28: Orrin Greenberg ran an extensive test to prove this, he evaluated eleven different AI models against a hundred and fourteen everyday marketing tasks.

00:14:37: how'd they do

00:14:38: well?

00:14:38: When I came just spotting wasted ad spend or tidying up messy CRM data They performed exceptionally Well.

00:14:44: but then He gave them forty complex budget allocation decisions.

00:14:48: The models produced a forty-two percent error rate.

00:14:52: A forty two percent error on budget allocation?

00:14:55: Yeah,

00:14:56: even the highest performer in the test which was Claude Opus five point five Was functionally at coin toss.

00:15:02: it made the correct allocation decision fifty one percent of time.

00:15:06: Okay.

00:15:06: so if you aren't part that six point five percent measuring the ROI and You hand your budget pacing over to system with a forty two per cent error rate You're just accelerating your own mistakes at an enterprise scale.

00:15:21: Precisely, and Jonathan McKay wrote a brilliant breakdown connecting this exact problem to the wider failure of AISDRs.

00:15:27: you know those autonomous sales development bots that flooded the market recently.

00:15:31: oh yeah the spambots

00:15:32: right.

00:15:33: B-to-B companies are treating large language models like a magic revenue button.

00:15:37: they purchased a tool pointed out a massive scraped list of unverified contacts and just expected to print pipeline.

00:15:45: In reality, it sprays generic cold emails gets the company's domain blacklisted and deeply alienates potential buyers.

00:15:52: It is a complete failure of organizational transformation

00:15:55: Which goes right back to the hook we started with.

00:15:57: Jonathan used Sendoso as the prime counter example of how it integrate AI correctly, they didn't just deploy a spam bot to cover a high volume of leads.

00:16:05: They did the grueling unglamorous work of rebuilding their actual revenue architecture.

00:16:10: Yeah The hard stuff

00:16:11: Exactly!

00:16:12: They cut their human BDR team from fifteen reps down to four But replaced them manual outreach With highly engineered deterministic data pipelines.

00:16:21: They didn't rely on an LLM to guess who was ready to buy.

00:16:24: They built strict rules-based criteria, to filter buying signals score the intent data and route only the absolute highest priority accounts for those four human reps.

00:16:34: And because they fixed a foundation Those four reps double qualified pipeline in ninety days.

00:16:39: It is stark difference between adopting a shiny new tool and adopting fundamental strategy.

00:16:44: Absolutely

00:16:45: Fortunately The analytics side of industry finally building infrastructure To help marketers track this reality.

00:16:52: Tracy Pallon and Matteo Zamban both highlighted a major update to Microsoft Clarity.

00:16:57: The platform is now offering free tracking of bot activity, and AI citations.

00:17:02: Oh that's huge.

00:17:03: It's massive.

00:17:04: Marketers can actually see if an AI agent is finding their business Identify what hidden system prompts are being run against their site And analyze where the friction points are when an AI referred visitor finally lands on page.

00:17:17: Okay so we finally have some visibility into this synthetic traffic.

00:17:20: but When you start measuring this accurately, you inevitably run into our final major hurdle.

00:17:26: If you fix your measurement and realize that generic volume-based AI output simply does not convert B to B buyers what is the actual value of human marketer in this new stack?

00:17:37: That's a big existential question!

00:17:39: Right... because the original promise was that AI would save us time.

00:17:43: But reality on ground is very different.

00:17:46: Justina Perot pointed out that practically every marketer she speaks with is more stressed-out than ever.

00:17:50: Yep, I've heard of the same thing.

00:17:52: The dynamic is so predictable.

00:17:54: Leadership sees a demo of an AI agent.

00:17:57: They assume massive efficiency gains and they immediately mandate headcount cuts.

00:18:01: The remaining team is forced to absorb the excess work Because their drowning in execution.

00:18:06: nobody actually has time To learn how build these sophisticated Agenetic workflows we discussed earlier.

00:18:11: So they just use it as crutch?

00:18:13: Yeah frantic, ineffective ten-minute bursts between meetings just to rewrite an email or summarize a brief.

00:18:22: And it leads straight the commoditization of polished content.

00:18:25: Nicole Leffer had great observation about this regarding AI generated video because you know Claude can now edit videos create motion graphics string together B roll

00:18:35: and The first time you see it looks incredibly impressive.

00:18:38: It does.

00:18:39: But she pointed out that if you watch enough of these outputs they all look identical.

00:18:43: They share the same specific pacing, the same predictable transition cuts... ...they look as she called it

00:18:49: ''clawdy.''.

00:18:50: Clawdy is a perfect descriptor for current state-of-the-art BW content.

00:18:54: I mean we already saw this happen with text.

00:18:56: The market quickly learned to spot the synthetic sameness.

00:19:00: You know?

00:19:00: The infamous overuse of M dash or words like ''del'' and tapestry.

00:19:05: Oh, if I see Delve one more time.

00:19:07: Right?

00:19:07: Now that same synthetic baseline is happening to video and design.

00:19:11: when every single company in your sector has access to the exact same high-level production capability for twenty dollars a month.

00:19:19: That polished capability ceases to be a competitive advantage.

00:19:22: it just becomes The new boring baseline

00:19:26: And Bjorn Radha articulated a philosophy on this that i think Every marketer needs To hear.

00:19:30: he argues that the More AI we integrate into our workflows the more financially valuable being human actually becomes.

00:19:38: When grammatically perfect, highly polished content is infinite and functionally free polish is no longer a differentiator.

00:19:45: It's just table stakes.

00:19:47: Exactly Genuine human perspective original strategic thinking And importantly Human imperfection become your true assets in market.

00:19:56: An AI agent can generate message distributed at scale but only humans create underlying meaning.

00:20:02: Isabel Dantes-Masado echoed this perfectly in her keynote from AICon, she challenged a room full of marketing leaders to fundamentally change how they allocate their resources.

00:20:13: Her mandate was super clear offload the math To AI so humans can fiercely own The magic.

00:20:21: I love that!

00:20:24: Let your agentic workflows handle the bid pacing data structuring AEO formatting The anomaly detection, but the human marketer must own this strategy.

00:20:36: The customer empathy and...the unique point of view.

00:20:39: But

00:20:41: if cloddy perfection is a new baseline for average doesn't that mean our flaws?

00:20:45: Our lived experiences And our eccentricities as human marketers are actually most valuable assets right now.

00:20:52: It requires us to unlearn decades conditioning.

00:20:55: We've been trained to value our worth based on output.

00:20:58: How many blog posts we published, how many campaigns we launched this quarter?

00:21:02: And now that output is commoditized!

00:21:04: Right...we have to value out insight instead.

00:21:08: A machine no matter how well prompted cannot replicate the messy lived experience of solving a highly complex problem for real B-to-B customer.

00:21:18: It really can not.

00:21:19: and in digital landscape flooded with synthetic consensus That authentic human element basically the only thing that will build actual trust with a buyer, which you know leaves us with rather provocative structural question to consider as this technology matures.

00:21:35: Lay it on us?

00:21:36: Well if AI shopping agents are increasingly doing research and buying for our customers And our AI marketing agents are executing selling bidding and optimizing on our end Are we rapidly approaching a closed loop where bots or simply marketing other bots

00:21:53: Wow bought-to-bought commerce.

00:21:55: Right?

00:21:56: And if that synthetic loop becomes the standard, where does human psychology even fit into?

00:22:04: Are we destined to just sit back and arbitrate bought-to-bought negotiations?

00:22:08: That is a wild thought.

00:22:10: We will definitely leave you to chew on that one.

00:22:12: as you look at your own tech stack this week if You enjoyed this episode new episodes drop every two weeks.

00:22:17: also check out our other editions on field marketing channel & partner marketing account based Marketing MarTech go-to market, and social selling.

00:22:27: Thank you so much for joining us on the steep dive And make sure you're subscribed So you catch the next one.