Best of LinkedIn: AI in B2B Marketing CW 05/ 06

00:00:00: This episode is provided by Thomas Allgeier and Frennus based on the most relevant LinkedIn posts about AI in B to B marketing, calendar weeks five and six.

00:00:09: Frenness is a B-to-B market research company helping enterprise marketing teams sharpen their strategies and outreach with customer segmentation ideal customer profiles deep dives customer needs analysis buying center insights.

00:00:23: Yeah, it's good to be back.

00:00:24: We're looking at early February twenty-twenty six and you can just feel the mood has shifted.

00:00:31: It

00:00:31: really has.

00:00:32: I mean The hype cycle.

00:00:33: It's not just settling down, is it?

00:00:35: Feels like its completely crashed into a wall of well reality.

00:00:39: A Wall Of Reality Is Good Way To Put It.

00:00:41: The Flashy Demos Are Out.

00:00:42: We're Now Talking About Operating Models And You Know Actual Sales Numbers

00:00:46: Exactly.

00:00:46: Its Moved From Look At This Cool Thing to Okay.

00:00:49: this is messy and expensive.

00:00:50: How Does This Actually Work In The Real World?

00:00:52: And Your Seeing This Clash Now?

00:00:53: It'S The Automate.

00:00:54: Everything Crowd Versus The People Trying To Use AI To Make Humans Well More Dangerous.

00:00:59: So we've

00:00:59: Got Three Big Areas To Unpack Today.

00:01:01: Right

00:01:01: Yeah, three main clusters.

00:01:03: First the whole confusing mess around AISDRs.

00:01:07: are they dead?

00:01:08: Are They just evolving?

00:01:09: what's going

00:01:10: on right?

00:01:10: and then this move away from Just writing prompts to something called context engineering

00:01:15: Which I know sounds like a buzzword but it's uh It's critical

00:01:17: get into it.

00:01:18: And finally What we're calling The agentic web and Why your entire SEO strategy is probably obsolete.

00:01:25: okay

00:01:25: Let's start with that controversy.

00:01:26: Then theme one AIS DRS.

00:01:29: i mean if you just rewind till last year The big prediction was that by now the human SDR would be, you know

00:01:36: extinct.

00:01:36: Oh yeah We were all supposed to be on a beach somewhere while the bots booked our meetings.

00:01:40: And

00:01:40: yet here we are and data is telling very different story.

00:01:44: A post from John Idle really kicked things off.

00:01:48: He dug into what happened at ramp

00:01:50: Ramp as huge player Massive fintech Very forward thinking

00:01:55: Exactly.

00:01:56: They had this AI-SDR program called OTs At its peak.

00:02:01: This thing was generating thirty percent of their entire pipeline.

00:02:04: Which is, yeah... That's a staggering number!

00:02:06: Most sales leaders would kill for that I know

00:02:08: right?

00:02:09: But here's the kicker.

00:02:10: They shut it down

00:02:12: Turned off

00:02:13: turned off.

00:02:13: machine idle explains that The ROI It just plateaued.

00:02:17: it turns out.

00:02:18: you can't Just scale outreach like that forever.

00:02:21: The quality started to drop and they realized they needed To empower there.

00:02:25: five hundred human sellers actually close the deals.

00:02:28: That such critical point Yeah.

00:02:30: The tech didn't break, the strategy broke and this lines up perfectly with what Jason Bay pointed out.

00:02:36: he found This little piece of irony that I think is pretty telling.

00:02:39: Oh

00:02:39: it's good.

00:02:40: He noted all big AI companies We're talking open

00:02:43: A.I.,

00:02:44: Clay People who actually build these stuff.

00:02:47: They are hiring human SDRs Like right now Aggressively

00:02:53: Wait so the chefs are refusing to eat their own cooking.

00:02:56: Pretty much.

00:02:56: If the people building the AI don't trust it to replace their own sales teams, why should anyone else?

00:03:01: That's the question isn't it?

00:03:02: and Arnaud Weiss kind of gave this psychological Why behind this?

00:03:06: he calls it reciprocity reciprocity.

00:03:07: Yeah his point is that an AI email communicates zero effort as a buyer.

00:03:13: You know Abat wrote in less than a second so you feel like Zero social obligation to even reply.

00:03:19: That's

00:03:19: so true.

00:03:20: if a human researches me I at least feel like I owe them and no thanks.

00:03:23: If a bot spams, me I owe it nothing

00:03:25: exactly.

00:03:26: we buy from humans because there's that shared investment of time.

00:03:29: Okay So let me play devil's advocate.

00:03:31: is the AI SDR just dead?

00:03:34: was it?

00:03:34: Just a failed experiment?

00:03:36: No, I don't think so.

00:03:37: you have to be careful not to swing too far the other way.

00:03:39: It's not dead.

00:03:39: It's just that the first wave of implementation Was Lazy.

00:03:44: Lee!

00:03:44: Yeah, C.T Leong had this amazing breakdown of the math that we really can't ignore.

00:03:49: Go on

00:03:50: He points out that a traditional human SDR team costs about one point eight million dollars before they book a single qualified meeting.

00:03:57: Wow, when you factor in salary turn over all that

00:04:00: everything.

00:04:01: he argues That AI agents plus just one human manager Can do the same work for ten times cheaper?

00:04:08: Okay So the argument isn't that AI is better at selling it's that its infinitely cheaper at failing.

00:04:13: In a way, yes.

00:04:14: It's about runway.

00:04:15: for startup.

00:04:16: that kind of efficiency is literally survival.

00:04:19: But the smart companies are changing how the AI shows up.

00:04:23: Ryan Tiesel shared a great example from qualified.

00:04:26: They have this agent called Piper.

00:04:28: I saw this.

00:04:29: they didn't hide it in little chat bubble and moved to what they call pinned mode on website sidebar

00:04:34: Right instead hiding bot.

00:04:36: they made it a core part of the experience.

00:04:39: And just by changing the UX, just by making it visible and accessible They double their conversations in meetings In one week.

00:04:46: So its not about replacement It's building better human middleware.

00:04:50: That is the term Niko Gillard and Marcus Chu are using And that seems to be consensus.

00:04:55: Automate grunt work The list-building research.

00:04:58: But for complex deal or handling C-suite objection you need a human.

00:05:02: I loved Summit End's take on this, he said an AISDRs main job shouldn't be to write more emails.

00:05:07: it should be to delete more e-mails.

00:05:09: the infrastructure mindset yeah

00:05:11: yes He uses AI to filter out sixty or seventy percent of bad outreach before it gets sent.

00:05:18: The AI defines this super narrow ICP window.

00:05:22: so your not spam cannon and your sniper rifle

00:05:24: which is how its should be.

00:05:26: And that brings us maybe the funniest story from Dan Briscoe.

00:05:30: If we're gonna treat these AIs like colleagues, We have to onboard them.

00:05:34: Like colleagues.

00:05:35: Oh

00:05:35: this was brilliant.

00:05:36: He on boarded an AI agent named Alice from a company called Eleven X and he treated her exactly like a new hire.

00:05:43: So

00:05:43: what did they do?

00:05:44: He gave her a thousand pages of Company reading material To get up-to-speed

00:05:48: Which she probably read in.

00:05:49: What three minutes

00:05:50: about that?

00:05:51: yeah but then And This is the best part He tried to get her watch the company's training videos.

00:05:56: Named that?

00:05:56: She refused!

00:05:58: The AI literally said she is more of a reader.

00:06:00: No way, That's incredible.

00:06:01: It's

00:06:02: hilarious but it such good lesson.

00:06:03: You can't just plug these things in and walk away.

00:06:06: She still needed human onboarding To understand companies nuance.

00:06:17: I

00:06:21: am seeing this term pop up everywhere.

00:06:23: Twenty-twenty five was the year of prompt engineering, right?

00:06:26: Learning how to ask chat GPT nicely but Maja Voje put it so well.

00:06:31: she said if you give an AI that same prompt twice and get different results You don't have a problem.

00:06:36: You've got context problems.

00:06:38: Exactly!

00:06:38: Prompt Engineering is a tactic.

00:06:40: Write me in email Context.

00:06:42: Engineering is strategy.

00:06:43: It's about building whole environment around the AI Feeding your business rules Your brand voice Your customer data before you ever ask it to do anything.

00:06:52: Brendan Short quoted Andres Karpathy on this, he called it the delicate art of filling the context window.

00:06:58: but what does that actually look like?

00:06:59: Are we just pasting massive documents into the chat?

00:07:02: No!

00:07:02: That's the amateur hour version.

00:07:04: Brendan highlighted a much smarter approach from Joe Rue.

00:07:07: He is using Markdown files in GitHub.

00:07:09: Wait hold on...GitHub isn't that tool for software developers?

00:07:12: It IS.

00:07:13: But here why its brilliant for marketers Markdown clean structured text that AIs can read perfectly.

00:07:20: More importantly, it creates a single source of truth.

00:07:26: changes his product pricing in that one markdown file, every single AI agent connected to it.

00:07:33: Whether its writing emails or answering support tickets updates instantly.

00:07:37: Wow!

00:07:38: So you're not retraining five different bots?

00:07:40: You are just updating the central brain.

00:07:41: Youre updating the Central Brain.

00:07:43: That is a system and I think people get confused by all of their terminology here but Caroline Healy had great breakdowns.

00:07:50: for hierarchy.

00:07:51: She sees three levels.

00:07:52: Level One is Generative AI.

00:07:54: Thats The Creator.

00:07:55: You asked ChatGPT to write a poem?

00:07:57: That's Gen AI.

00:07:57: It's simple enough.

00:07:58: Level two is A.I agents.

00:08:00: They're the specialists, they perform specific tasks like Book of Meeting or Scrape this website for me.

00:08:05: And third level

00:08:06: Agentech-AI.

00:08:08: that's The Orchestrator.

00:08:09: it's system coordinates multiple specialist agents To achieve big goal Like reduce customer churn.

00:08:15: It figures out which agent to deploy and when

00:08:17: But Sandeep Gulati had warning about.

00:08:20: You can't just buy a bot and call it agentic.

00:08:23: No, exactly!

00:08:24: A real system needs memory?

00:08:26: It needs to handle errors?

00:08:27: It need

00:08:28: guardrails?!

00:08:29: If your bot crashes because a spreadsheet column was named wrong... ...it's not an agent — it is just fragile

00:08:35: script.".

00:08:36: We saw some pretty wild examples of this in action though….

00:08:39: Jonathan MK shared his co-workplace.

00:08:42: He's saving he claims over forty hours per week.

00:08:45: One of them …was his competitive war room.

00:08:48: I loved this one.

00:08:49: Yeah, so instead of a human having to check the competitor's pricing page every day He has an agent that just watches the market.

00:08:55: when a competitor changes their pricing The agent spots it automatically updates the internal battle cards and pings.

00:09:01: the sales team

00:09:02: is moving from right copy-to run operations.

00:09:06: And Marina Mugilco took it even further with Airtable's new super agent!

00:09:10: She gave one prompt, build a six-month strategy for my brand Silicon Valley Girl.

00:09:16: You'd expect you know...a generic fluffy paragraph back?

00:09:19: Right instead in twenty five minutes built full competitive analysis distribution model and hiring plan acted like a strategy consultant.

00:09:28: Here is the catch of all this All this amazing internal orchestration, the war rooms.

00:09:33: The markdown files.

00:09:35: It means nothing if your customers can't find you in first place.

00:09:39: Which brings us to theme three and This is probably the most urgent one...the agentic web!

00:09:47: I've spent twenty years getting good at Google Keywords backlinks h-one tags all of it.

00:09:54: But Scott Brinker shared this stat from Uberol.

00:09:57: that should be a wake-up call for everyone.

00:09:59: What was it?

00:10:00: Sixty eight percent of brands are completely absent from AI recommendations.

00:10:05: Invisible.

00:10:05: think about the buyer asks chat GPT or Claude For The Best CRM, for small business and if you're not in That answer You don't exist, you're not even in the running.

00:10:15: And Nathan Hirsch's point was that buyers are making a decision inside of AI's answer and they aren't clicking through to your website anymore!

00:10:21: It just completely destroys traditional funnel.

00:10:24: Simrish data backs this up too... AI doesn't reward for having those blog posts.

00:10:29: it rewards context and credibility.

00:10:32: So The Million Dollar Question How do we get found?

00:10:35: If backlinks isn't our main signal what is?

00:10:37: Ryan Law from RFS He dropped the golden nugget of the entire two weeks, he analyzed the data and found a single strongest predictor of AI visibility.

00:10:48: It's not domain authority... What is it?

00:10:50: ...it's YouTube mentions.

00:10:52: YouTube!

00:10:53: Seriously?!

00:10:53: A zero

00:10:53: point.

00:10:54: seven three-seven correlation that is an incredibly high number in this kind of analysis.

00:10:59: Why YouTube?

00:11:00: is the AI watching videos?

00:11:03: It's reading the transcripts.

00:11:04: These models are hungry for high quality conversational data to understand what a brand or product is, and YouTube transcripts were goldmine.

00:11:14: the AI decides you're relevant.

00:11:16: So, the strategy shifts from write more blogs to get mentioned in more videos?

00:11:20: And freshness is critical.

00:11:21: Brian Law also found that AI overviews change every two point one five days.

00:11:25: You can't just set it and forget it.

00:11:27: But we have to worry about where the AI gets its information right.

00:11:30: Natalia Garcia had some stats on this.

00:11:32: Yeah The classic garbage-in-garbage out problem.

00:11:35: She found ChatGPT sites Wikipedia at nineteen percent of time but its site's Reddit thirteen percent.

00:11:43: That's terrifying for a B-to-B brand.

00:11:45: Your reputation could be shaped by some anonymous reddit thread!

00:11:49: Exactly, and it is becoming a platform war.

00:11:51: Chris Lawn pointed out that Bing Webmaster Tools just launched an AI performance report.

00:11:56: Finally...some visibility.

00:11:57: We can finally see what they call grounding queries.

00:12:00: Okay

00:12:00: translate that for me Grounding Queries.

00:12:04: It shows you the exact question.

00:12:06: A human asked the AI that caused The bot to pull data from your website To ground its answer.

00:12:12: It's the first real peak under the hood of AISEO.

00:12:15: And at the same time, Hans Cronenberg was talking about this Super Bowl ad from Anthropics

00:12:19: and Makers Of Claude.

00:12:20: yeah they're positioning themselves as The No Ads Clean Answers Alternative To Google An Open AI.

00:12:26: it's a battle for trust!

00:12:28: And

00:12:28: Trust all comes back to data quality.

00:12:30: Georgie Frenedjev made point.

00:12:31: we can't miss All these agents ,all these systems.

00:12:34: They break if the data is bad.

00:12:37: He said static lists are dead.

00:12:38: You need live Real-Time Data.

00:12:41: If your agent is working off a six-month old spreadsheet, it's going to embarrass you.

00:12:46: Speaking of embarrassment let's hit few quick practical tips before we wrap.

00:12:50: Nicole Leffer had great one on preventing hallucinations.

00:12:54: The I don't know.

00:12:54: rule It so simple?

00:12:56: You have to explicitly tell the AI in your prompt.

00:13:00: do not make anything up and I don't know is a valid answer.

00:13:03: It just drastically cuts down on the errors.

00:13:06: i also love tom winter's idea of The critique stage.

00:13:09: it's like robot fight club for content.

00:13:10: Yeah,

00:13:11: you have one ai.

00:13:12: maybe claude generate?

00:13:13: First draft.

00:13:14: then You Have A different Model Like gpt-four Critique That Draft Against A set Of Rules.

00:13:18: Then The first Model Rewrites it.

00:13:19: The Output Is Just so much Better

00:13:21: and that all Feeds Into What Kieran Flanagan Was Saying About the new Models Like opus four point six.

00:13:26: He Said They're Starting to develop taste TV yeah Judgment.

00:13:30: They're making creative choices that sometimes beat human intuition.

00:13:34: It's not just pattern matching anymore,

00:13:36: which is both exciting and a little unnerving Okay?

00:13:40: Let's pull this all together.

00:13:41: What does the one big takeaway for everyone listening?

00:13:44: I think The Takeaway Is That The Easy Button Phase Over.

00:13:47: You Can't Just Plug In An AI SDR And Expect Magic.

00:13:52: The Winners This Year Are The Ones Doing The Hard Backend Work Of Context Engineering

00:13:57: Building the systems, the markdown files.

00:13:59: The central brain that makes the AI actually smart?

00:14:02: Exactly and at the same time they are fighting for visibility on this new agentic web

00:14:07: Making sure your brand exists in those YouTube transcripts and Reddit threads And wherever else the AI is learning

00:14:13: Because if the AI doesn't know you exist Your buyers never will.

00:14:17: If you enjoyed this episode, new episodes drop every two weeks.

00:14:20: Also check out our other editions on account-based marketing field marketing channel marketing MarTech.

00:14:26: go to market and social selling.

00:14:27: Thanks for listening And don't forget to subscribe!

00:14:30: I want leave with one final thought from a post by Thomas Ross.

00:14:33: He said So

00:14:39: who is shaping your story when not in the room?

00:14:41: Something to think about.

00:14:43: See ya next

00:14:43: time.