Best of LinkedIn: AI in B2B Marketing CW 39/ 40

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

00:00:06: In calendar weeks, thirty nine and forty.

00:00:09: Frenious is a b to be market research company that supports enterprise marketing teams in unlocking full potential of their customer data with help from AI.

00:00:17: Yeah we've got really packed deep dive for you today.

00:00:21: We do So.

00:00:23: if you think about it For last two decades building A B-To-B Marketing Strategy was Essentially like designing a massive beautiful storefront on the absolute busiest street in the world, right?

00:00:34: Google.

00:00:35: Right you optimized your window displays.

00:00:37: You know you fought tooth and nail for the best real estate on the block.

00:00:40: You made sure your sign was the biggest one there And then you just sort of stood back and watched The foot traffic rollin

00:00:47: exactly.

00:00:47: and the rules are that real estate were incredibly predictable.

00:00:50: I mean even if the competition was fierce you knew Exactly the path that the buyers were walking down to find you.

00:00:56: But looking at the massive stack of insights we're diving into today, it really feels like that main street is emptying out.

00:01:02: Yeah rapidly Like

00:01:03: buyers just aren't window shopping anymore.

00:01:06: It seems they've all hired personal shoppers who bypassed the streets entirely go straight to their warehouse and then bring back a curated shortlist directly on the buyer's desk

00:01:15: Which is terrifying thought if your entire budget tied up in Main Street real estate.

00:01:19: Totally So.

00:01:20: our mission for the steep dive is to unpack this fundamental shift.

00:01:24: We are looking at how visibility is, frankly violently pivoting from search rankings to AI recommendations

00:01:32: and we'll get into how AI agents Are actually becoming the operating layer of marketing teams

00:01:37: right?

00:01:37: And what all these latest benchmarks in market moves Actually mean for your strategy going into twenty-twenty seven?

00:01:44: because the speed of this transition Is what's catching even highly competent veteran marketing teams totally off guard.

00:01:50: They are still optimizing for a street that buyers aren't actively abandoning.

00:01:55: So let's start right there with our first theme, because the data on organic clicks taking a hit is just harsh wake-up call.

00:02:02: Oh it's brutal!

00:02:03: Lance Black shared some recent Ahrefs research and shows that presence of an AI overview correlates to a fifty eight percent lower click through rate than number one organic result.

00:02:14: That is massive.

00:02:15: Just put in perspective You could execute a flawless SEO strategy.

00:02:20: Win the top spot you've been chasing for like three years, and you still lose more than half your traffic.

00:02:26: instantly?

00:02:26: Instantly

00:02:27: because The Nature of the Battle has completely changed.

00:02:30: from where do I rank to?

00:02:32: am i one Of the answers right?

00:02:34: Tatiana Priebuszinskiya calls this recommendation visibility And she highlighted This brutal reality For legacy brands Like you can possess massive domain authority.

00:02:45: Oh yeah, all the backlinks in the world.

00:02:46: Exactly thousands of high quality back links incredible historical traffic.

00:02:51: But when a buyer opens an AI interface and asks what fenders should I consider for enterprise data compliance?

00:02:58: You might be completely invisible.

00:03:00: Wow Yeah your competitors populate the chat And you're brand simply doesn't exist in that context

00:03:06: which is i mean?

00:03:07: That's terrifying When you factor in the buyer behavior data that thomas ross brought up.

00:03:12: Currently, ninety-four percent of BWB buying groups are locking in their vendor shortlist before they ever even speak to a sales rep.

00:03:18: Ninety four percent?

00:03:20: Yeah!

00:03:20: So if you miss that initial AI generated recommendation You aren't just losing a click...you're totally disqualified from the revenue Before your new deal was on the table.

00:03:30: Exactly so we have look at mechanics.

00:03:32: how these engines actually decide who gets on this short list.

00:03:36: Right

00:03:36: How do they pick?

00:03:37: Well, Alex Groberman analyzed fifty-seven point two million AI citations and the behavioral gap is just staggering.

00:03:46: If a buyer asks an AI about your brand by name like a branded query Your site has cited seventy seven point six percent of time.

00:03:54: Okay that makes sense.

00:03:56: But if they run in unbranded category query Like best marketing automation tools You're citation rate plummets to two point two percent.

00:04:05: Wait I need to wrap my head around A fourteen X drop?

00:04:10: Yeah.

00:04:11: Why is the engine punishing unbranded queries so severely like, Is it just hallucinating or is it actively biased against niche brands?

00:04:20: It's not bias...it basically comes down to difference between information retrieval and consensus synthesis.

00:04:28: Okay unpack that.

00:04:29: So when a buyer uses your brand name The AI architecture triggers direct retrieval task.

00:04:35: It looks for your own properties because obviously you are the definitive source on yourself,

00:04:40: right?

00:04:41: Nobody knows my brand better than my website

00:04:43: exactly.

00:04:44: but On a category query.

00:04:46: The AI switches modes entirely.

00:04:49: it's now trying to synthesize A consensus from the broader internet.

00:04:53: Oh I see.

00:04:54: Yeah, it looks with the most authoritative Frequently validated answers that can assemble from third-party industry sources.

00:05:01: It isn't looking for your opinion of yourself anymore It's calculating the industry opinion of you.

00:05:06: Okay, so if everyone is bleeding organic clicks and The AI filters heavily based on this synthesized consensus What are the actual mechanics to win these recommendations?

00:05:17: I'm assuming we can't just You know stuff keywords into a blog post anymore.

00:05:21: No absolutely not.

00:05:22: the entire discipline has shifted to answer engine optimization or AEO.

00:05:26: okay

00:05:27: And the mechanics are actually highly structural.

00:05:30: Rob Hoffman shared this highly specific tactical approach.

00:05:34: that illustrates this perfectly.

00:05:36: What's the tactic?

00:05:37: He advocates for overhauling your about page into an eight-section format, but The crucial mechanic here is embedding a literal Wikipedia style HTML keyfax table

00:05:48: like just a raw data table

00:05:50: Yeah Rose specifically delineating founder pricing competitors services headquarters all that.

00:05:57: But wait, that sounds incredibly rigid like almost robotic.

00:06:00: Why would I strip away my carefully crafted brand narrative for a boring data table?

00:06:05: Because you are formatting data for a parser now not human Right

00:06:09: right.

00:06:10: large language models were heavily trained on Wikipedia's architecture.

00:06:14: They rely on that info box structure to categorize entities in their vector space.

00:06:18: So it is literally about speaking the machine's native language

00:06:21: Exactly When an AI crawler hits your site.

00:06:24: if it has to parse through this flowery narrative marketing copy.

00:06:28: Just to figure out your pricing tier, it might just skip you entirely but if finds a cleanly tagged HTML table It can instantly extract and categorize those facts with high confidence which makes You much more likely be cited as a definitive answer.

00:06:42: That is so wild And that structural simplicity Is totally echoed by Lily Grozeva's recent conference data.

00:06:48: Oh

00:06:49: what did she say?

00:06:49: She was saying that the tactics winning A.I.

00:06:51: citations aren't these exotic technical rebuilds, it's doing simple things like putting The Current Year in your title utilizing clear tables of contents and specifically phrasing Your H-II tags as the exact questions a buyer would ask.

00:07:05: Right!

00:07:05: It is about reducing the cognitive load on the machine.

00:07:08: Exactly

00:07:09: Structure is just baseline though...the real friction for B to D marketing team will be endurance.

00:07:16: Davis McCain shared some research from Perfound that completely shatters our traditional SEO assumptions.

00:07:22: How so?

00:07:23: Well, in AI search a citation has half-life of just eleven days.

00:07:27: Wait...eleven

00:07:28: days?!

00:07:28: Eleven Days!

00:07:30: In Traditional SEO if I earn top three ranking for high intent keyword That is a fortress like i can defend the traffic for years.

00:07:37: Why would an AI Citation vanish under two weeks?

00:07:41: The data shows seventy-eight percent of these cited pages fall to half their peak share within two weeks.

00:07:47: And Matt Hines framed this takeaway perfectly, authority in AI expires rather than accumulates.

00:07:54: Expires rather then accumulates?

00:07:56: That's a huge mindset shift!

00:07:58: Totally...

00:07:59: A traditional search engine uses human clicks over time to reinforce the value of static link.

00:08:03: right but answer engines operate on dynamic vector databases that continuously ingest new database

00:08:10: So they just want whatever is

00:08:12: newest.

00:08:13: They prioritize freshness and recency to ensure that aren't giving outdated answers.

00:08:17: You get cited in a burst because your content the newest relevant node.

00:08:22: But as newer data enters model, you're weight decays!

00:08:26: You cannot run an AI visibility campaign for quarter walk away.

00:08:31: It requires always on programs.

00:08:33: Okay so does this mean strategy becoming a content mill?

00:08:38: spam the engine with daily listicles just to feed their recency bias?

00:08:42: Actually, no.

00:08:43: That will completely backfire based on where architecture is heading.

00:08:47: Oh

00:08:47: thank goodness!

00:08:48: Rand Fishkin predicts that AI systems are going to rapidly exhaust The value of the internet's training text.

00:08:55: There simply isn't enough high-quality texts left to ingest.

00:08:58: They're running out of Internet.

00:08:59: Exactly So.

00:09:00: to combat this... ...the models soon start waiting real human usage metrics.

00:09:05: Did actual humans read, scroll and share this information?

00:09:09: Which connects perfectly to the fifty percent cliff.

00:09:12: Right!

00:09:12: The fifty-percent cliff.

00:09:13: Shirley McBeth introduced this concept at a summit And it fundamentally changes how you allocate budget.

00:09:19: .The premise is basically This If more than half of data feeding an AI's understanding Of your brand originates from your own website ,the answer engines stop believing You.

00:09:28: It just cuts you off.

00:09:29: Yeah Your visibility falls right off a cliff

00:09:32: Because models are programmed To seek third party validation to prevent manipulation.

00:09:37: They need to see independent analysts, industry influencers and unowned media confirming what you're saying right?

00:09:43: If you claim to be the premier enterprise solution but literally no one else in the internet is saying it The math just stops supporting your claim.

00:09:50: It's a fascinating paradox like To conquer the machines.

00:09:53: Your PR earned content teams actually have to prove that You are highly valued by humans Exactly Which is wild.

00:10:01: Hey, real quick.

00:10:02: If you're finding these insights valuable make sure to subscribe so you catch future editions where we track these exact shifts.

00:10:08: So um We've established how we need to restructure externally so buyers can find us.

00:10:13: Yep Let's pivot to our second theme here internal operations because AI isn't just changing the storefront It's completely rewriting How the marketing team actually operates in The warehouse

00:10:25: right?

00:10:25: Oh absolutely We are moving way past the era of just using AI as a basic chatbot to generate email subject lines.

00:10:33: we are officially entering The Era of Integrated Agenetic Systems.

00:10:37: Agenetic systems?

00:10:39: Yeah,

00:10:39: Blake Imperial shared a workflow that perfectly illustrates this maturity.

00:10:43: What did he do?

00:10:44: He used Claude code To build a full paid landing page in under an hour But he didn't stop at a mock-up!

00:10:51: He pushed it straight to Google Ads via a Webflow MCP connection.

00:10:55: Wow.

00:10:55: And the key here, wasn't that the AI could just write HTML?

00:11:00: The key was the context through that connection.

00:11:03: They had direct real time access to the team's hot jar heat maps their GA for analytics and they're internal Slack discussions.

00:11:11: That is massive.

00:11:12: That context is the differentiator

00:11:14: totally.

00:11:16: Jonathan Martinez noted that For advanced teams, the stack has actually shrinking down to just four main cloud connectors.

00:11:23: really

00:11:23: Just four.

00:11:24: yeah.

00:11:25: Fathom for ingesting meeting transcripts, Notion for your core brand and strategy guidelines, Appify for scraping social listening data.

00:11:33: And Gmail for historical communication context.

00:11:36: that covers basically everything

00:11:38: right.

00:11:39: when an agent can pull from the actual nervous system of your business like that The output stops being generic.

00:11:45: it becomes highly strategic

00:11:46: When people are already doing this at scale.

00:11:48: Ira Bodner brought us up They have ten automated cloud code workflows running right now.

00:11:54: Ten Yeah,

00:11:55: and these aren't just drafting blog posts.

00:11:57: These bots are doing comprehensive Google and Meta ads audits managing budget pacing detecting statistical anomalies.

00:12:04: they find wasted spend And just email the human manager The same day.

00:12:09: that's incredible.

00:12:10: But you know to share a volume of execution?

00:12:12: These agents can handle creates a whole new kind of risk.

00:12:15: Oh for sure.

00:12:16: Brendan Hufford warned about this.

00:12:18: he calls it clawed slop

00:12:19: clawed sloth.

00:12:20: It sits with the perfect word.

00:12:21: so great work.

00:12:22: It's this scenario that is likely happening in enterprise teams right now.

00:12:26: A marketer uses Claude to generate a nineteen page campaign report they don't actually read.

00:12:31: Oh

00:12:31: no

00:12:31: They send it to the CMO who then use his Claude summarize and generate feedback, you just end up with an endless loop of unread AI reports passing between machines.

00:12:43: Ok here where gets really interesting though.

00:12:47: If agents take over eighty percent of the actual execution, writing and auditing what happens to human marketer?

00:12:54: Do we just review Claude's law all day?

00:12:58: because I don't want my career as an editor for a robot.

00:13:02: To prevent that management framework has to evolve.

00:13:05: Kate Payne actually mapped this transition beautifully into the forty year old SLII leadership model

00:13:11: The situational leadership one

00:13:12: Exactly which was originally designed for managing human employees.

00:13:16: Her analogy is that giving agents autonomy, Is just like that.

00:13:19: You have to stage it based on risk and complexity.

00:13:23: So you move from directing To coaching to supporting to finally delegating.

00:13:28: But how do you actually apply That to a machine?

00:13:30: Am I like having A one-on-one coaching session with Claude?

00:13:33: Operationally It's about guardrails.

00:13:35: Okay

00:13:36: In the directing phase you give The agent strict parameters And manually approve Every single headline it generates.

00:13:41: okay very hands-on

00:13:42: Right.

00:13:43: And then as the model proves it understands your brand voice, you move to coaching.

00:13:47: maybe It builds the whole ad set and you only tweak the targeting parameters.

00:13:51: Make sense?

00:13:52: And eventually for low risk tasks You reach delegating where The agent is just allowed To allocate a daily budget autonomously.

00:13:59: But wait who takes the fall?

00:14:00: when the agent hallucinates and spends ten thousand dollars on the wrong keyword like accountability?

00:14:05: Is a huge bottleneck there

00:14:07: Exactly!

00:14:08: And Carolyn Healy addressed this.

00:14:10: She argues that every single production agent needs to have an Agent ownership card.

00:14:14: What's that?

00:14:15: It's a literal rule named human owner attached to that specific workflow.

00:14:19: because as she points out Judgment doesn't scale with compute.

00:14:23: ooh, that's good line

00:14:24: right.

00:14:25: AI removes the production bottleneck.

00:14:27: You can generate a thousand assets in minute.

00:14:29: But strategic context brand safety That still operates at humans beat.

00:14:35: if an agent damages the brand You can't just look your board in the eye and say, well... The model did it.

00:14:41: No

00:14:41: definitely not!

00:14:42: Pascal Bornet calls this threat Scope Collapse.

00:14:46: Scope collapse?

00:14:47: Yeah on paper handing all execution to agents looks like a massive productivity boom but for the human marketer if you're entire job shrinks down to reviewing machine output It feels like a demotion.

00:14:58: Oh completely.. Your no longer creative architect you are maintaining factory floor

00:15:02: So the only way survive that scope collapse is to redefine your professional value.

00:15:09: Bornet suggests that future job titles will actually describe what we own rather than what we personally produce.

00:15:15: What do you mean?

00:15:16: You'll own the system architecture, The strategic intent, the safety guardrails and Lomit Patel echoed this exactly.

00:15:24: He stated that Future CMOs Will manage systems And leverage not just people!

00:15:29: You aren't carving wood by hand anymore... ...you are calibrating robotic arms.

00:15:34: That requires an entirely new infrastructure though, which actually leads us to our final theme today.

00:15:40: Product launches funding and market moves.

00:15:43: the market is aggressively building the infrastructure for this new reality.

00:15:46: Yeah we finally have a way to measure This beyond just vendor hype.

00:15:50: The launch of optimized leaves mark bench version one point five shared via shop cut.

00:15:54: Islam in Alexander Atzberger Is a huge deal.

00:15:58: Mark Bench like sweet bench first software engineers

00:16:01: exactly.

00:16:02: It objectively proves if a model can do the job using nearly three hundred scenarios based on real complex marketing tasks.

00:16:11: Okay, so what did the benchmark actually show?

00:16:13: The most profound data point proves that an AI harness Can cut marketing task costs by eighty eight percent.

00:16:20: Eighty-eight

00:16:20: percent

00:16:21: Yes from forty dollars and fifty two cents for a raw model down to just four dollars ninety one cents while Actually scoring higher on quality.

00:16:30: Wait, how does a harness drop the token cost that much?

00:16:33: I thought token costs were just static.

00:16:36: A raw model can drift right.

00:16:38: it might hallucinate or waste computational cycles.

00:16:42: overthinking is simple prompt and AI harness uses techniques like prompt chaining And caching to force them on into highly constrained efficient pathways.

00:16:52: Oh so it stops at from rambling.

00:16:53: exactly okay It stops the LLM for wasting tokens.

00:16:58: So you drastically reduce the compute cost and You get higher quality because it's hyper focused.

00:17:03: so The value isn't just buying the API access It's the architecture around it And the underlying models are getting terrifyingly capable on their own.

00:17:12: Justin Hardy gave an update on Gemini for Argonne.

00:17:15: Oh, the token window update?

00:17:17: Yeah

00:17:17: They've expanded the maximum output tokens to one million.

00:17:21: its built from multi-step long horizon workflows

00:17:24: Which totally changes what is possible for a marketing team With a one million token window, you can instruct the model to say ingest your entire CRM history for the past two years.

00:17:35: Analyze six months of raw sales called transcripts and diagnose exact funnel

00:17:39: leaks.".

00:17:40: Right it can just hold all that disparate data in its active memory simultaneously.

00:17:45: but as we get more embedded We also have to deal with regulations.

00:17:49: Always.

00:17:49: Chris Long brought up a really crucial regulatory update.

00:17:52: OpenAI is rolling out hidden watermarks It's called text grain to chat GPT outputs, but they're doing it in the EU to comply with the EU AI

00:18:01: Act.

00:18:02: Which adds a huge layer of technical friction for global teams... But The Ultimate Friction is where this entire ecosystem is heading next.

00:18:10: Sandy Carter introduced a concept that redefines everything we've talked about today.

00:18:15: Agentic commerce

00:18:16: right?

00:18:17: So what does all mean by the actual transaction if I have AI tools building my marketing and AI tools reading

00:18:24: Carter envisions a near future where personal AI agents like Meta Muse or OpenAI Dots will negotiate directly with brand agents.

00:18:33: So my marketing agent isn't trying to persuade the human buyer anymore?

00:18:36: Nope, it's pitching the buyers' Personal AI Agent.

00:18:40: The next battle in commerce happens entirely in the background before the human customer ever even sees the digital shelf.

00:18:47: Wow It just requires a complete teardown of how we view influence.

00:18:52: A machine doesn't care about a slick networking dinner or beautiful PDF, it just analyzes your data points perfectly

00:18:59: Which leads us with the final thought for you to mull over.

00:19:02: If were entering an era where your brand's AI agent has to pitch a buyer personal AI agent How do you market into a machine that does not feel emotion?

00:19:11: Right

00:19:12: What is Brand Affinity even mean when the buyer isn t algorithm?

00:19:16: That Is The Question Of Decade.

00:19:18: Well, if you enjoyed this episode new episodes drop every two weeks.

00:19:22: Also check out our other editions on field marketing channel and partner marketing account based marketing MarTech.

00:19:27: go to market and social selling.

00:19:29: Thank you so much for joining us.

00:19:31: hit subscribe And we will see you next time.