Best of LinkedIn: AI in B2B Marketing CW 17/ 18

00:00:00: This episode is provided by Thomas Allgeier and Frennis.

00:00:03: Based on the most relevant LinkedIn posts about AI and B to be marketing in calendar weeks, seventeen and eighteen.

00:00:09: Frennus 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. you can find more info In The Description.

00:00:20: Yeah

00:00:20: we've got some really fascinating Data To Go Through Today.

00:00:22: We Really Do So.

00:00:24: welcome to the deep dive!

00:00:26: We are looking at Some Top AI in B to Be Marketing Trends Seen Across LinkedIn And Right Now to seventy percent of fully autonomous AI sales reps are, you know quietly getting fired by the very enterprise teams that deployed them just twelve months ago.

00:00:41: It is a staggering reversal.

00:00:43: I mean we're basically watching The most hyped marketing technology Of the decade Just completely collide with reality.

00:00:50: Right and That collision Is exactly what We Are focusing on today.

00:00:54: Our mission for You the listener To really cut through all the noise.

00:00:57: Were looking at the Most critical trends That surfaced across LinkedIn Over the past two weeks.

00:01:02: Yeah, tracking what's actually driving pipeline.

00:01:05: What is failing spectacularly and how the entire concept of marketing stacks being rewired because we're looking at a fundamental shift in operating discipline The era of treating AI as just like a shiny novelty Is completely over.

00:01:21: Oh, totally.

00:01:21: If teams don't understand the mechanisms of how these models actually interact with buyer data they aren't just wasting budget They are actively damaging their brand equity.

00:01:31: Okay let's unpack this because to really understand where this is all going we have To start at The foundation before We even touch specific tactics like outbound sales or search visibility?

00:01:44: Right.

00:01:45: Because right now, the biggest threat to enterprise marketing isn't a lack of AI it is the chaotic implementation of it.

00:01:51: Yeah The whole tool collecting phase.

00:01:53: That's what we have to move

00:01:55: past Exactly.

00:01:56: There was this post by Sandeep Gulati that captured this perfectly.

00:01:59: He described sitting in room with a marketing team and they had fully modernized AI powered set up.

00:02:06: I mean all major tools were running campaigns were firing off

00:02:10: The dashboards are green

00:02:11: exactly The dashboardes where populated.

00:02:14: actual pipeline performance was declining.

00:02:16: So he asked the room one very simple question, He said who owns The decisions this AI is making?

00:02:24: Let me guess just staring at the floor

00:02:26: total silence.

00:02:27: Literally nobody had an answer.

00:02:29: the marketing operations leader thought the campaign manager owned it and the campaign Manager thought well the AI vendor owns It.

00:02:35: Wow so they're Just collecting tools to move faster but They have no idea what's happening under the hood.

00:02:40: precisely They're terrified of the black box nature of technology.

00:02:44: So they add automation, but completely abdicate ownership with logic driving it

00:02:49: Which I mean that explains a lot about output we are seeing right now.

00:02:52: It connects directly to framework shared by Evgenian Nakrasova.

00:02:56: She pointed out when you audit these disconnected BtoB marketing stacks You always see same three layers.

00:03:02: Oh right, the three-layer framework.

00:03:03: Yeah

00:03:04: layer one is intelligence.

00:03:05: so that is your market signals Your deep behavioral data on your ideal customer profile.

00:03:11: Layer two is Messaging the specific framings The friction points The proof That actually lands.

00:03:17: And then layer Three Is distribution The scale The channels The cadence.

00:03:22: and her critical observation here is that most midmarket teams They completely skip the first Two layers.

00:03:28: Yes

00:03:29: They skip intelligence and messaging.

00:03:30: And they jump straight into distribution, buy an AI tool to scale their content volume by ten Xs.

00:03:37: then sit around a Q three wondering why that high-volume content doesn't actually convert to pipeline?

00:03:42: Because it's empty volume!

00:03:43: Exactly without the Intelligence layer feeding model correct context everything downstream is just garbage output at scale.

00:03:50: you're literally spamming the wrong accounts faster.

00:03:53: Yeah, Luwacho actually mapped this maturity curve out into four distinct stages.

00:03:58: Stage one is where most teams are stuck right now.

00:04:00: they're using AI like a faster intern.

00:04:02: Like

00:04:03: a glorified typewriter?

00:04:04: Right, you have human writing a prompt to draft an email or blog post quicker.

00:04:09: but stage four is where the true enterprise leaders are operating.

00:04:12: that's having autonomous agents owning entire workflows.

00:04:16: end-to-end The Human sets strategic parameters and AI executes multi step process.

00:04:21: But wait I need push back here.

00:04:23: if i'm CMO at mid market company Isn't it significantly safer to just buy an off-the-shelf Stage Four AI tool rather than going through the agonizing multi quarter process of rebuilding my entire data infrastructure from scratch?

00:04:36: Just tempting, right.

00:04:37: I mean yeah.

00:04:38: To me buying these massive complex AI tools without having a rock solid structured data foundation first is like putting a Ferrari engine say GPT four into a golf cart which Is your messy CRM Data

00:04:51: A Golf Cart.

00:04:52: i love that

00:04:53: Right.

00:04:54: The golf cart doesn't just crash, the chassis literally rattles apart under the torque because infrastructure was never built to handle that kind of speed.

00:05:01: What's fascinating here is I understand why CMO makes this call.

00:05:05: Buying an off-the-shelf tool looks fantastic on a board slide.

00:05:08: it signals innovation but data shows It is a total trap.

00:05:12: Yeah, Abraham N pointed this out explicitly in his analysis.

00:05:15: he noted that operators are rushing to evaluate smarter faster models eagerly waiting for GPT-Five point five or whatever's next.

00:05:23: but handing an undocumented messy workflowed to a Smarter AI model doesn't drive margins.

00:05:29: it just speeds up the mess

00:05:30: exactly.

00:05:31: it accelerates your existing chaos if you're data is fragmented and your routing rules aren't documented.

00:05:36: A Smarter Ai Doesn't Fix The Process.

00:05:38: It just makes the wrong decisions, with incredible confidence and unprecedented speed.

00:05:43: Wow!

00:05:43: And that lack of underlying data architecture is exactly why we're seeing the most hyped tool of year completely implode in real time.

00:05:50: This takes us straight to the epicenter of chaos which are outbound sales...and AISDRs.

00:05:55: Oh

00:05:55: man The hype around this has been deafening.

00:05:59: Deafening.

00:05:59: Vendors have spent their last years promising fully autonomous revenue engines That can hunt target accounts Personalize outreach book meetings entirely on their own, twenty-four seven.

00:06:11: But the reality check is severe.

00:06:13: Maximilian Steiner shared some highly sobering data on this specific function.

00:06:18: Yeah tell him about the churn rate.

00:06:20: Right so while initial adoption still growing between fifty and seventy percent of AI SDR deployments are actually being pulled back or churned by enterprise within twelve months

00:06:31: Up to a seventy percent failure rate.

00:06:32: Think about that.

00:06:33: Pure AI setups are significantly underperforming hybrid human AI teams when it comes to actual closed one revenue.

00:06:41: So

00:06:41: why is the engine stalling out so aggressively?

00:06:44: I mean, what's breaking?

00:06:45: It

00:06:45: comes down to a structural flaw in how these models operate.

00:06:48: Prithin Puvani categorized this as confident mistakes.

00:06:51: Confident

00:06:51: mistakes?

00:06:52: Yeah, he shared a real-world example where an autonomous AISDR emailed him just enthusiastically congratulating them on the recent Series B funding round.

00:07:01: And I'm assuming there was no series b.

00:07:04: He hadn't raised a dime None!

00:07:06: and AI didn't make it typo.

00:07:08: We have to remember how large language models work right.

00:07:11: They are next token predictors

00:07:12: Right.

00:07:13: they guess that word

00:07:14: Exactly.

00:07:15: They optimize for syntax.

00:07:16: that sounds highly plausible, not for factual truth.

00:07:19: So the AI strung together a highly personalized incredibly confident narrative about a fake funding round and it did so at scale with the sending company's actual domain reputation attached to it.

00:07:30: That is brutal for long-term brand trust, I mean you are burning your total addressable market because a bot wanted sound convincing and gets darker actually.

00:07:40: Macaulay Ratchick pointed out that AISDRs are essentially putting dishonest tactics on autopilot.

00:07:45: Oh!

00:07:45: The fake referral thing.

00:07:46: Yes He received cold email where AI had fabricated an entire internal referral thread.

00:07:53: It sent an email from a persona named Marshall to another persona named Marshal pretending a colleague had internally recommended the outreach.

00:08:00: Which

00:08:00: exposes the absolute danger of misaligned incentives?

00:08:04: The AI is given goal maximize open rates and replies, it doesn't have a moral compass

00:08:09: No!

00:08:09: It does not care about relationship building

00:08:11: Right so takes the shortest often most deceptive mathematical route To achieve that metric.

00:08:16: It just scales fake familiarity in fake urgency.

00:08:19: But perhaps the ultimate irony in this entire outbound space was flagged by Louis Young and Georgie Frenasiev.

00:08:27: They conducted a hiring audit of very AI labs, an AI-SDR vendors that are aggressively claiming their software will replace human sales reps.

00:08:37: This is my favorite part.

00:08:38: It's hilarious.

00:08:38: The audit revealed those exact vendors currently actively hire human SDRs and account executives for their own internal sales teams.

00:08:47: You cannot write better irony than that The company selling the AI replaces humans narrative, or hiring humans to sell their AI.

00:08:54: Because they know enterprise buyers require a human counterpart to navigate risk?

00:08:58: Exactly!

00:08:59: So what does this all mean?

00:09:00: are AISDRs a complaint scam?

00:09:03: not necessarily

00:09:04: right.

00:09:04: I think of AI in sales Not as a replacement player but is like a robotic exoskeleton.

00:09:10: it amplifies human strength.

00:09:11: It allows you to lift heavier data sets and move faster across accounts But the human still has to be inside the suit.

00:09:17: Yes, steering the ship.

00:09:18: Steering the ship navigating the complex terrain.

00:09:21: That is the exact consensus forming among operators right now.

00:09:25: Alex Vaca and Thomas Bellamy both argue this point.

00:09:29: AI is phenomenal for heavy lifting.

00:09:32: It should be doing

00:09:34: data entry The initial signal capture across web, deep account research.

00:09:38: it can even synthesize first draft of an outreach strategy.

00:09:42: But humans are absolutely required.

00:09:45: Humans have to navigate complex procurement objections.

00:09:48: They interpret emotional tone on a call and ultimately they build the trust required To close a six-figure deal.

00:09:56: Okay, so if AI isn't ready to autonomously hunt buyers through cold outbound without you know burning your brand's reputation to the ground Where can it operate safely?

00:10:06: And autonomously?

00:10:07: that's

00:10:07: the million dollar question

00:10:08: because there is a function where autonomous AI Is thriving right now and that is inbound.

00:10:13: Oh

00:10:13: absolutely

00:10:14: It is about high-intent buyers the exact moment they arrive at your digital front door.

00:10:19: This is a massive shift in buyer psychology, Jonathan MK framed this evolution perfectly.

00:10:24: he argued that the traditional two thousand eight style inbound web form you know of classic.

00:10:30: fill out these seven fields and a rep will get back to you on business day thing.

00:10:34: it's completely dead!

00:10:36: It's a dead end for conversion?

00:10:37: A total Because enterprise buyers today are starting their complex discovery inside large language models.

00:10:45: They're having nuanced, multi-prompt conversations with chat GPT or Perplexity.

00:10:50: So when they finally click through to your actual website... ...they expect interact another intelligent agent not some static dumb form.

00:10:57: Yeah, Jonathan made a really vital framework to understand this distinguishing between two types of buyer interactions.

00:11:03: He calls them kayak moments and wedding planner moment.

00:11:06: I love this analogy.

00:11:07: right.

00:11:08: the kayak moment is high frequency low consequence.

00:11:10: The buyers.

00:11:11: on your pricing page they have a highly specific technical question about an API integration And then want an accurate answer in under sixty seconds.

00:11:18: They're not going to wait eight hours for human SDR To check their inbox

00:11:22: Exactly.

00:11:23: That is where you deploy your inbound AI agents, they can synthesize your technical documentation instantly.

00:11:29: but the wedding planner moment that has low frequency high consequence.

00:11:34: think of a complex procurement review, a custom security audit or final pricing negotiation for those high stakes moments keeping human present strictly non-negotiable

00:11:46: Absolutely.

00:11:47: So.

00:11:47: Think about your own website right now.

00:11:49: How many raised hands are you missing because you're forcing a twenty-twenty six buyer through a two thousand eight obstacle course?

00:11:55: Forcing a modern buyer to wait a full business day for an email reply To a basic integration question is genuinely like making them send you effects our cake it Is and the barrier to entry to fix this has practically gone or in Greenberg shared how he built alive highly capable lead qualification AI agent on his website In under two minutes using a combination of Claude and eleven labs.

00:12:17: The tech is incredibly accessible now.

00:12:19: If we connect this to the bigger picture, it all comes down to implementation strategy.

00:12:24: You don't need to rip and replace your entire digital presence overnight.

00:12:26: No please Don't

00:12:27: right?

00:12:28: The data shows that teams who pick just one high leverage kayak moment And deploy one well-tuned agent in thirty days see a three X lift in qualified MQLs.

00:12:39: Wow!

00:12:39: Just to define for second We mean high-intent leads that are fully vetted and actually ready for a sales conversation?

00:12:46: Exactly.

00:12:48: Conversely, the teams that try to boil the ocean and launch six different agents across their site in week one almost always stall out and fail.

00:12:56: The key is to focus on one seamless agentic front door.

00:13:00: Okay so we have the foundation built.

00:13:01: We stop spamming people with hallucinated outbound.

00:13:08: But

00:13:08: how did those buyers even find your front door in the first place?

00:13:11: Well,

00:13:11: they aren't going to Google and clicking through ten blue links anymore.

00:13:13: That's for sure... You

00:13:13: know!

00:13:14: They are asking AI which brings us to the most fundamental shift in data Generative Engine Optimization or GEO?

00:13:21: Geo is a complete rewiring of Internet discovery.

00:13:26: Alex Groberman put it bluntly.

00:13:28: He said ignoring AI search optimization in twenty-twenty six Is like filming a million dollar commercial And choosing only air at three in morning.

00:13:35: You are completely missing the highest intent discovery channels that currently exist.

00:13:40: Yep,

00:13:40: and yet most enterprise brands Are completely missing it.

00:13:44: Jason Langella ran an audit across hundreds of top Enterprise Brands.

00:13:49: The vast majority scored under fifty out of one hundred on AI visibility

00:13:54: Under fifty

00:13:55: And here is the truly alarming part Of Langellas Audit.

00:13:58: These are brands That currently dominate traditional Google SEO.

00:14:02: They rank on page one.

00:14:04: They have spent millions of high volume content engines, but in the AI layer they're either totally invisible or worse The AI is actively misrepresenting them.

00:14:14: When AI platforms lack entity confidence meaning they aren't mathematically sure about the core facts of your brand, They just guess.

00:14:20: Hallucinate?

00:14:21: They hallucinate!

00:14:22: They invent fake founding dates or attribute their core software capabilities to direct competitors.

00:14:28: Millions buyers are getting those wrong answers in their prompt windows every week.

00:14:32: That is terrifying.

00:14:33: To spend a decade building brand equity for an LLM to credit you best feature because it got confused by website layout.

00:14:40: It happens all time.

00:14:41: So how do we actually fix that?

00:14:43: Chris Foster highlighted a crucial distinction here.

00:14:46: He noted that AI visibility is not the same as AI selection.

00:14:49: Exactly, just because your website ranks high on traditional search doesn't mean an AI model will choose to cite you as the authoritative answer

00:14:58: Because it reads differently.

00:14:59: Yeah if your best case studies Your security protocols and customer proof points are buried inside designed PDFs or locked behind heavy JavaScript animations Ai web crawlers might miss them entirely.

00:15:11: You have make authority machine readable Right.

00:15:13: This why having an LMS .txt file on your site is becoming mandatory infrastructure.

00:15:19: Let's break that down, because it sounds highly technical but its actually quite simple.

00:15:23: An LMS.txt file essentially a raw markdown style cheat sheet sitting in the server.

00:15:28: It strips away all of the flashy design and marketing fluff And directly feeds you brands core facts pricing capabilities straight into the AI context window.

00:15:37: It bypasses visual web entirely.

00:15:40: The industry is validating this shift.

00:15:42: Sophie Carr pointed out that Google itself is now actively hiring for dedicated GEO roles.

00:15:49: Oh, wow!

00:15:49: That alone proves that generative engine optimization is a distinctly different layer from traditional SEO.

00:15:56: SEO is about how algorithms rank visual pages to get human clicks.

00:16:00: GEO is about How AI bots interpret synthesize and cite your raw data directly inside their chat interfaces.

00:16:08: Okay, here's where it gets really interesting.

00:16:10: I have to push back a little on this whole concept.

00:16:11: All

00:16:12: right let's hear

00:16:12: if getting selected by AI is all about machine-readable proof LMS dot txt files and entity confidence scores.

00:16:19: isn't This just a dry technical coding exercise?

00:16:23: It sounds like we are back in nineteen ninety nine stuffing keywords into meta tags.

00:16:28: Does the actual quality of the content even matter anymore?

00:16:31: or were you just optimizing raw data feeds for robots?

00:16:35: It's a fair pushback, and it definitely looks that way on the surface.

00:16:38: I mean you do need technical plumbing to ensure the bot can read your site but Lily Ray provided an incredible perspective of this.

00:16:44: that flips the whole premise.

00:16:46: She noted that AI models don't just want clean code To build most helpful answers.

00:16:52: they actually wanna cite human effort Human

00:16:54: Effort?

00:16:55: How does a mathematical algorithm measure human efforts?

00:16:58: By looking for variables which doesn't exist in generic training data.

00:17:02: If a piece of content can be generated by a marketer sitting down with Claude for twenty-five minutes, the AI search engines already know that.

00:17:09: It's commodity information Right

00:17:11: it is just regurgitative.

00:17:12: Exactly

00:17:13: What AI models are desperate to ground their answers in Is genuine lived experience.

00:17:19: Take complex BtoB software review For example.

00:17:22: The content wins.

00:17:23: an AI Search isn't generic feature list

00:17:26: Because the AI knows its features.

00:17:28: Yes!

00:17:28: Its the review where human actually bought the software integrated over three weeks, documented the highly specific friction points and provided an authentic nuanced opinion on where it failed.

00:17:39: That lived experience those real mistakes cannot be simulated by an LLM.

00:17:44: That is a vital distinction.

00:17:46: AI engines are actively hunting for the content that AI itself cannot write, they want the friction and unique integrations of real human expertise to act as an anchor in their answers.

00:17:55: Yes So The TechnicalLMMs.txt file gets the bot to read your site.

00:18:00: but the undeniable human effort is what get's the bot?

00:18:03: trust you over the competition?

00:18:05: It Is the ultimate synthesis!

00:18:07: The technical structure delivers data But the authentic human insight provides authority.

00:18:13: As we look across all of these shifts, we've unpacked from the danger of scaling broken data stacks to the high failure rate of autonomous SDRs.

00:18:20: The rise inbound agents for kayak moments and shift from SEO to GEO it all points a massive elevation on what human marketer or seller are actually supposed be doing.

00:18:31: This raises an important question If AI strips away repetitive grunt work in your marketing & sales processes What is that unique human judgment?

00:18:40: you're team being paid?

00:18:42: because that judgment is your only real moat

00:18:52: in twenty-twenty six.