Best of LinkedIn: AI in B2B Marketing CW 27/ 28

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

00:00:07: In calendar weeks twenty seven and twenty eight.

00:00:10: Frennis 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 A I

00:00:19: yeah.

00:00:19: And uh i'm really glad to be digging into this material with you today

00:00:22: right because if you're listening To this You've probably been navigating your btb marketing With like The exact same GPS for like, twenty years now.

00:00:30: Pretty

00:00:30: much.

00:00:31: and then overnight sixty percent of your Google traffic just vanishes into an AI engine.

00:00:36: the map went totally blank

00:00:37: completely gone exactly.

00:00:39: so today we're figuring out a rebuild.

00:00:41: it weird dissecting how brands actually get discovered.

00:00:44: Now this search engines are basically turning in to answer engines plus The real mechanics of scaling AIS DRs without Just burning Your entire territory To the ground which

00:00:54: is So easy to do right now.

00:00:55: Oh hundred percent.

00:00:56: And we're also gonna look at the hidden data traps waiting for you and why human taste is actually becoming the ultimate competitive mode.

00:01:04: That vanishing traffic thing, it's probably the perfect place to start honestly because a transition from traditional SEO to AEO answer engine optimization or GU generative engine optimization.

00:01:16: It's forcing a complete tear down of the in-mound playbook.

00:01:20: I mean, ranking number one on the search page just doesn't guarantee a user is going to visit your website anymore.

00:01:25: Right and the metric that really grounds that reality comes from Kyle Atwater-Morley.

00:01:31: he noted that roughly sixty percent of Google searches now end with absolutely no click to any web site.

00:01:37: Sixty percent?

00:01:38: That's just wild!

00:01:39: It's huge.

00:01:40: The user intent is entirely intercepted.

00:01:42: they ask a question.

00:01:45: Well, zero reason to actually click through the source material.

00:01:50: Which fundamentally breaks traditional funnel right?

00:01:53: A user reads that synthesized answer They trust brands recommended in text and they make a buying decision based solely on summary.

00:02:01: So you can technically win keyword battle.

00:02:04: You can secure top organic spot And still watch your pipeline completely dry up because AI caught users before even reaching domain.

00:02:13: But there's... There's a secondary later to this that is actually quite concerning for marketers and it has to do with attribution.

00:02:20: Oh, you're talking about the ghost citation problem?

00:02:22: Yes.

00:02:23: Yeah, Mohamed Kamar Shafiq ran an analysis on this And then numbers are bit grim.

00:02:27: honestly He found that sixty two percent of AI citations or what he calls Ghost Citations.

00:02:34: Let us explain the mechanism there.

00:02:35: How exactly is he defining a ghost citation

00:02:38: So essentially means the AI engine scrapes your site, right?

00:02:42: It uses your proprietary research to construct its answer.

00:02:45: But it completely fails to name your brand to the user.

00:02:49: you're basically an uncredited ghostwriter.

00:02:51: like

00:02:51: in one specific study he highlighted medium was used as the foundational source material sixteen times for a set of queries but the platform was named The User exactly zero times.

00:03:03: That creates a massive attribution black hole, and you know we can't treat these AI models as monolith either because they handle citations very differently based on their specific architecture.

00:03:14: right he pointed out that Gemini for example will only provide a clickable hyperlink to sources about twenty one point four percent of the time.

00:03:23: okay.

00:03:23: however it explicitly mentions the brand name in the generated text.

00:03:28: eighty three points seven percent the time.

00:03:31: Oh interesting, so Gemini acts more like a traditional PR mention?

00:03:35: Like you get the brand awareness even if we don't get that direct backlink traffic.

00:03:38: Precisely.

00:03:39: but then you look at chat GPT and it does the exact opposite really.

00:03:41: yeah It has relatively high frequency of providing those citation links But a staggeringly low rate actually mentioning the brand name in the conversational text itself.

00:03:51: So if a buyer isn't you know Hovering over those tiny footnote numbers in ChatGPT.

00:03:55: They have absolutely no idea whose expertise they're reading.

00:03:59: That is, I mean if we're doing all the heavy lifting and models are just stripping our names off of final product.

00:04:05: The obvious question how do you force a model's hand?

00:04:08: We have to navigate a lot noise here too.

00:04:12: issued a fairly blunt warning about this.

00:04:15: He was calling out agencies that are just repackaging their standard legacy SEO services, slapping an AEO sticker on the deck and selling it as a brand new

00:04:23: strategy.".

00:04:24: Oh you see it constantly right now?

00:04:25: All

00:04:26: of time!

00:04:26: They're optimizing H-two headers maybe adding some FAQ drop downs.

00:04:30: pretending somehow satisfies large language models.

00:04:33: but he notes real AI visibility operates in completely different temporal rhythm.

00:04:38: He points out there is a brutal seven-day recency window for certain queries.

00:04:43: Wait, meaning what?

00:04:44: Structurally because like if we publish a massive pillar page doesn't that hold its weight?

00:04:51: Not in an AI ecosystem, not when it's constantly re-indexing for the most current context.

00:04:57: If you publish a batch of technical articles and just walk away your AI visibility can degrade in a matter of days.

00:05:04: The models prioritize newer fresher signals to formulate their answers.

00:05:09: Man that is exhausting But I do want to look at a tangible tactic to combat this, because Gaitan and Nino DiNardi laid out a strategy that directly addresses how LLMs parse information.

00:05:20: He argues the absolute key is comparison content—the classic versus pages.

00:05:26: It makes complete technical sense when you think about these models actually synthesize data.

00:05:30: They map relationships between concepts using semantic vectors.

00:05:33: So a versus page natively structures your product into contract Nodes, Pros, Cons feature comparisons.

00:05:40: You're handing the AI the exact structure format it needs to weigh two options against each other.

00:05:44: and if you refuse to build that comparison content yourself like out of some fear of mentioning a competitor... The AI doesn't just give up!

00:05:53: No.. It finds somewhere else

00:05:54: Right….

00:05:55: It relies on outdated competitor pages or maybe some two-year old Gartner review to explain what your product does, you completely surrender the narrative.

00:06:03: You have to force the AI to read your updated positioning by giving it that contrast its looking for?

00:06:09: Exactly and Edan Aramajic actually quantified this approach which I love.

00:06:13: he worked with a FinTech company that had absolutely zero presence in AI overviews Nothing.

00:06:19: Wow,

00:06:20: okay.

00:06:20: and by strategically publishing fifteen comparison articles He drove their AI visibility score on comparison queries all the way up to ninety point five percent.

00:06:30: I have to imagine.

00:06:31: the pipeline impact of a ninety percent visibility score is fairly immediate.

00:06:35: Oh yeah, on day nineteen in the campaign.

00:06:37: they secured direct inbound lead sourced entirely from an AI recommendation.

00:06:41: That's

00:06:42: incredible

00:06:42: Zero ad spend no sales outreach A buyer researched inside and AI saw the brand definitively recommended over the competitor And just initiated contact.

00:06:52: That's the dream.

00:06:53: Right,

00:06:54: and by the way to you listening if you want to stay ahead of rapid structural shifts like that seven-day recency window we just talked about Just hit subscribe To catch future editions of these deep dives.

00:07:06: We are tracking all these changes as they happen so you don't have too.

00:07:09: it's a critical shift And You know let's connect the mechanics here for a second.

00:07:13: If inbound discovery is drying up because AI's intercepting the search intent, that naturally pushes marketing and revenue teams to panic a bit.

00:07:21: Oh definitely!

00:07:22: They try to compensate for lost inbound by just cranking their outbound volume... And they're using AI to do it.

00:07:28: which brings us this sudden explosion of AISDRs.

00:07:32: Right,

00:07:32: because every sales leader right now is currently being pitched some AISPR platform.

00:07:38: the value proposition Is literally always the same lower your headcount increase Your outbound volume by a factor of ten and just book meetings on autopilot.

00:07:47: But Diego Mangabera offered a really necessary reality check here.

00:07:51: He argues that when AISD ours fail it's rarely a failure.

00:07:54: The actual language model.

00:07:56: It's almost always the result Of a fuzzy ideal customer profile.

00:08:00: Well, yeah because AI is essentially just an execution multiplier.

00:08:03: Yes It just amplifies whatever strategy you feed it.

00:08:06: Exactly so if your targeting is vague and your core messaging is weak Bolting in a tool onto that process doesn't fix the core issue.

00:08:16: It just allows you to waste five hundred emails a day burning through your total addressable market much Much faster than a human team ever could.

00:08:26: It's a fundamental clared problem masquerading as the technology problems.

00:08:31: And you know, listening to this might be thinking your CRM is clean enough to run one of these autonomous agents.

00:08:38: but Nicholas Rabanis made a brilliant point about that gap between demo environment and reality.

00:08:44: when an AISDR vendor pitches Their demo is running on perfectly sanitized pristine sample data.

00:08:51: Quit up, BCRMs are never pristine!

00:08:53: Never.

00:08:54: they're usually graveyards of outdated contacts.

00:08:56: Yeah people who left the company three years ago.

00:08:58: Exactly if you point a high-speed AI agent at a decayed list where half the context have changed jobs You aren't doing outbound.

00:09:05: your just annoying the wrong people at unprecedented scale.

00:09:08: He argues that AISDRs absolutely require scored total addressable market data so highly tiered accounts layered with live real

00:09:17: life signals.

00:09:18: Right, we're talking about API feeds tracking recent funding rounds executive job changes or tech stack deployments.

00:09:25: without that live signal layer the AI just has no context for its outreach

00:09:29: which makes sense.

00:09:30: and Elric Liglor broke down how Snowflake navigated this exact infrastructure challenge.

00:09:36: they actually managed to get a massive fifteen X lift in their outbound reply rates, jumping from zero point five percent up to seven point six percent.

00:09:44: That's

00:09:44: a massive jump for an org that size huge.

00:09:48: But they didn't achieve that by simply buying a new AI tool and just unleashing it on their three hundred SDRs.

00:09:53: They had to rebuild the architecture underneath the reps, didn't they?

00:09:56: Yes exactly!

00:09:57: The completely tore down there old lead scoring models And rebuilt the intense signal routing so that the AI was only drafting messages based on highly qualified contextual triggers.

00:10:07: So the AI is really just the final thin execution layer On top of this massive data restructuring.

00:10:13: Right Exactly.

00:10:13: Let me push back slightly on the execution side of this though.

00:10:17: Okay, even if you have perfect data right and perfect intense signals aren't enterprise buyers just going to develop an immediate aversion to robot callers?

00:10:27: I mean we can all hear that unnatural cadence That weird half-second latency pause when AI voice agent responds... That

00:10:35: is the common assumption And i certainly feel like i can spot in ai immediately.

00:10:39: Right But Gabe Larson shared an experience.

00:10:43: he received a cold call from an AISDR.

00:10:46: He realized what it was within seconds and was fully prepared to hang up

00:10:50: as most of us would

00:10:51: be exactly, but he didn't.

00:10:53: And they actually admitted the AI outperformed The vast majority.

00:10:56: if human SDR is he interacts with wait

00:10:58: outperform them.

00:10:59: on What criteria though

00:11:00: brevity inefficiency you got straight-to-the point?

00:11:02: It didn't stumble over a script.

00:11:04: it clearly Didn't understand.

00:11:05: it knew that specific context Of why I was calling and crucially it didn't fake rapport.

00:11:11: Oh, the fake rapport is the worst.

00:11:13: Right didn't pretend to ask how his weekend was.

00:11:16: he noted that The bar for human outbound has gotten so embarrassingly low That a concise context-aware AI Is actually a relief to a busy executive

00:11:25: which points To a larger structural shift and go-to market strategies.

00:11:28: honestly Alan rafting noted that buying isolated ai tools like one vendor for your AI voice caller, a different tool for your intent data.

00:11:39: Another for email enrichment it just creates disconnected spam machines.

00:11:43: none of those systems share state.

00:11:45: yeah

00:11:45: And the winners are building connected GTM systems, like you pointed to platforms like Alta where inbound signals outbound messaging and intent data operate a real-time loop.

00:11:54: Right So if Prospect hits high value pricing page The system instantly identifies similar accounts on CRM and economically launches a highly tailored outdown sequence into those secondary accounts within minutes.

00:12:05: See that connected loops sounds incredibly powerful But then entire vision completely collapses.

00:12:10: If foundation it sits is cracked We keep circling back this data architecture, it is the single variable that dictates if any of this autonomous AI actually works in practice.

00:12:23: And Justin Hardy quantified just how cracked that foundation is.

00:12:27: for most companies.

00:12:28: It's bad isn't?

00:12:29: Its really bad.

00:12:30: He shared a metric indicating that forty five percent standard marketing data Is dirty and B to b data decays at an aggressive rate of thirty percent per year.

00:12:40: And that decay rate is compounding too.

00:12:42: Companies merge, domains change, buyers get promoted or just leave the industry.

00:12:47: So if you integrate an autonomous AI agent with a decaying CRM or un-permission third party data You aren't achieving scale...you're basically deploying hallucinated decisions across your entire customer base.

00:13:01: took this concept of shared state a step further.

00:13:04: He pointed out that agentic AI where the system is taking independent action on behalf of the business, it only creates value if it has complete real time context across the whole enterprise.

00:13:14: right and I agent can't just be trained on your marketing PDFs and call it today.

00:13:18: no It needs to query or live inventory databases.

00:13:22: yeah you're fulfillment logistics Your historical order State.

00:13:26: If your AI marketing agent autonomously spins up an aggressive promotional offer for a key account, but it doesn't have the permissions to see that you're supply chain is delayed and

00:13:43: It's like we're building a state of the art high speed AI bullet train, but were stubbornly trying to run it on rusty broken tracks from twenty fifteen.

00:13:51: That

00:13:51: is the perfect way to visualize it.

00:13:53: Yeah And Jurg Storm framed The economic reality Of this beautifully.

00:13:57: he stated

00:14:01: Wait, the one that almost finishes unpacked it for a second.

00:14:03: Well basic reasoning and conversational abilities are rapidly becoming cheap commodities.

00:14:07: now Right Having an AI who can just chat with user and recommend white paper?

00:14:11: That's table stakes.

00:14:13: The real defensible business value lies in completing the regulated complex workflow...the last mile of execution.

00:14:20: He used the example of Anvil introducing AISchema mapping.

00:14:23: Okay so how does this work in practice?

00:14:26: Instead of human downloading a complex unstructured PDF invoice reading it and manually typing the data into a CRM.

00:14:33: The AI reads the document, maps the specific fields perfectly to the company's proprietary data architecture And turns it in to structured workflow-ready data.

00:14:43: Wow Yeah No human intervention required.

00:14:47: It finishes the task

00:14:48: completely.

00:14:49: But I have to ask a structural question here.

00:14:51: Sure If we're spending all our resources rigorously cleaning our data and structurally mapping it via API so that autonomous agents can process it without errors.

00:15:02: I mean, who are we actually formatting this for?

00:15:04: Are we even marketing to humans

00:15:06: anymore?".

00:15:06: I don't think there's a real consensus on that yet but Arvin Seychon highlighted the paradigm shift you absolutely have to take seriously.

00:15:13: He argues marketers must begin optimizing

00:15:16: machine preference.

00:15:18: As agentic commerce matures AI agents increasingly acting as actual buyers.

00:15:23: A human executive tells their AI.

00:15:25: find me three enterprise software solutions security compliance standards and cost under a hundred grand.

00:15:32: The AI goes out, does the research compares the features...and builds the

00:15:36: shortlist.".

00:15:37: So if the AI is building this short list it doesn't care whether your website has a beautiful parallax scroll or if you're brand video uses like sweeping emotional storytelling?

00:15:46: Not at all!

00:15:47: The AI agent cannot be persuaded by emotional marketing.

00:15:50: It's looking for structured data, API documentation clear pricing tables and machine legible signals that definitively prove your product capabilities.

00:15:59: Wow if you're data is locked inside a beautiful but unreadable brand video the AI simply moves on to a competitor with structured specs.

00:16:07: That Is A Staggering Pivot Because If Machines Are Essentially Selling To Machines.

00:16:12: You Know Our AI SPR Pitching Their AI procurement agent, and we are optimizing our collateral for machine legibility.

00:16:20: What does the human marketing leader actually do all day?

00:16:23: It forces

00:16:23: a total reevaluation of our value.

00:16:26: And this brings us to the final reality We have to confront.

00:16:30: Human judgment is becoming The ultimate differentiator

00:16:32: which makes sense.

00:16:33: I mean we're already seeing the consequences Of removing human judgement.

00:16:37: Sarastella Latanzio shared A study revealing that forty percent of long-form posts on linkedin Are already fully ai generated.

00:16:44: forty percent

00:16:45: Yep, the platforms are just flooded with this generic perfectly structured AI slop.

00:16:50: Well when the cost of producing content drops to zero volume ceases to be a competitive advantage.

00:16:56: it just becomes a commodity.

00:16:57: I'd argue its even worse than that.

00:16:59: pumping out indistinguishable high-volume AI content isn't just ineffective It is actively damaging brand equity.

00:17:07: Oh absolutely

00:17:08: buyers immediately recognize this synthetic tone and they mentally downgrade your authority.

00:17:13: Michelangelo Simonte shared some research on how agencies are handling this.

00:17:17: Eighty-seven percent of agencies using generative AI right now, it's ubiquitous.

00:17:22: but the

00:17:22: top performing agencies not use to blindly pump out more generic blog posts.

00:17:27: they're deploying AI strictly to eliminate tedious low value work.

00:17:31: routine analytics reporting basic data formatting initial keyword clustering All

00:17:36: right, so by automating the drudgery.

00:17:38: they're deliberately keeping humans in the driver's seat for creative decisions strategic taste and brand voice.

00:17:44: Taste is basically becoming a new defensive moat

00:17:47: And applying that tastes requires deep technical understanding of your tools.

00:17:52: Luicho offered some highly tactical advice on matching AI models to the specific difficulty of the task.

00:17:58: Okay

00:17:58: He pointed out a glaring inefficiency.

00:18:00: right now Marketers are getting access to high-end frontier models like Claude fable five and they're using it To write basic social media captions,

00:18:09: which is like using a Super Bowl ad budget to run A localized retargeting banner.

00:18:13: Yeah just a massive misallocation Of resources

00:18:15: exactly.

00:18:16: he explains that fable Five cost roughly five times more per API called than a mid-tier model like Opus.

00:18:22: Five times, yeah.

00:18:24: Fable five is not designed to be faster typewriter.

00:18:26: it's strategist engineered for long horizon complex reasoning tasks.

00:18:30: I see

00:18:31: if you need an AI to analyze a fifty page customer journey brief and identify structural gaps in the funnel You deploy fable.

00:18:40: If you just need to A B test some email subject lines use that cheaper fast or model actual intent.

00:18:49: So we're transitioning from being the manual laborers in the marketing engine to being the strategist who allocate the compute power?

00:18:55: Precisely,

00:18:56: but Eddie Reynolds offered a vital reframe on how he actually apply that strategy

00:19:00: right because Right now teams are running around with this shiny new technology asking what can AI do help us?

00:19:07: It's a solution in search of a problem.

00:19:08: Can it summarize our Zoom calls?

00:19:10: Yes, can it draft blog posts?

00:19:12: yes but summarizing meeting doesn't necessarily move the needle on pipeline generation

00:19:16: exactly.

00:19:17: he argues.

00:19:18: we need to invert the question entirely.

00:19:20: stop asking what they I can do.

00:19:23: Start by looking at your business fundamentals and asking where is our pipeline broken?

00:19:27: Where's our biggest opportunity to capture revenue?

00:19:30: if you're inbound conversion Is highly efficient, but your outbound response rate is abysmal then you allocate Your AI engineering resources to fix Outbound.

00:19:39: human judgment dictates.

00:19:40: We're to point the machine.

00:19:41: The AI provides the velocity But the human determines the vector.

00:19:45: I love that.

00:19:47: So as we wrap up this analysis what does the core concept are?

00:19:50: listeners should be dissecting with their teams this week.

00:19:52: I'd say look at the trajectory of every post we analyze today.

00:19:55: AI intercepting discovery, AI autonomous SBRs, AI agents acting as procurement researchers... Yeah We are rapidly approaching a business environment where AI agents will predominantly interact with other AI agents and when that entire middle layer of execution is automated The brands that secure the deal won't be the ones with the cleverest chat GPT prompts The winners will be the brand backed by the most undeniable real-world human credibility.

00:20:21: Authentic subject matter expertise is the only signal an AI cannot easily

00:20:43: replicate.