Best of LinkedIn: Artificial Intelligence CW 35/ 36

00:00:00: This deep dive is provided by Thomas Allgaier and Frennis, based on the most relevant LinkedIn posts about artificial intelligence in calendar weeks thirty-five and thirty-six.

00:00:10: Frennis supports enterprises with market and competitive intelligence, decoding emerging technologies, customer insights, regulatory shifts, and competitor strategies.

00:00:20: So product teams and strategy leaders don't just react, but shape the future of AI.

00:00:25: Welcome to the deep dive.

00:00:27: Look, if you've been on LinkedIn lately, you know.

00:00:29: It's just... a wash with AI content, an absolute flood.

00:00:33: Oh, completely overwhelming,

00:00:34: yeah.

00:00:34: So our mission today is really to cut through all that noise for you.

00:00:38: Absolutely.

00:00:39: Think of us as, well, your curators.

00:00:41: We've sifted through everything from calendar weeks, thirty-five and thirty-six, to find the really key top artificial intelligence trends seen across LinkedIn.

00:00:50: Distilling it down for you, especially if you're in the ICT and tech industry, giving you those crucial insights.

00:00:55: Exactly.

00:00:55: We want to help you stay ahead of the curve.

00:00:57: We'll be touching on, you know, big strategic shifts.

00:00:59: what these AI agents actually are, the whole governance piece, market feelings, all of it.

00:01:03: It's about helping you shape what's next, not just react to it.

00:01:07: Okay, let's kick things off with theme one, strategy and operating models.

00:01:12: How companies are really having to rethink things at a core level because of AI?

00:01:18: Yeah, it feels like we're past the purely theoretical stage now.

00:01:22: Definitely.

00:01:23: It's getting practical and sometimes painful, right?

00:01:25: Precisely.

00:01:26: And Daniel Paul had a great point on this.

00:01:28: He said, More frameworks won't save your agent.

00:01:31: A clear roadmap

00:01:32: will.

00:01:33: That rings true.

00:01:34: It does, because so many companies, they stumble, they don't know where to start, they jump between things.

00:01:39: And it just leads to chaos.

00:01:41: Total chaos.

00:01:42: And then they give up.

00:01:43: So that clear, strategic approach, it's just non-negotiable now.

00:01:47: Absolutely.

00:01:47: And, you know, Alina Keck shared this, frankly, staggering statistic from MIT.

00:01:52: Oh, yeah.

00:01:53: Ninety-five percent of generative AI projects failing to drive revenue.

00:01:58: ninety-five percent

00:01:59: huge

00:02:00: and often it's because they lead with the tech the shiny object not with a real solid business problem.

00:02:05: you have to ask what are we actually solving here?

00:02:08: first exact which you know connects to what chris kimmel is saying about finding your north star.

00:02:12: before you even think about ai governance policies.

00:02:15: right you need that strategic alignment first otherwise the policies are just floating disconnected precisely and it's changing leadership to isn't it?

00:02:23: matthew?

00:02:24: a mattson had this interesting idea moving from managing people to orchestrating AI systems.

00:02:33: Orchestrating, I like that.

00:02:34: Yeah, it's not about micromanaging.

00:02:36: the AI's choices, but designing the system itself, you know, with ethics, autonomy, intent, build right in.

00:02:42: It's a fundamental shift in what leadership looks like.

00:02:45: And Jens Bucking talked about these frontier firms.

00:02:47: Ah, yes, the frontier firms.

00:02:49: Yes, structured around on-demand intelligence, human AI teams working together, super agile.

00:02:54: He argues this isn't just nice to have anymore.

00:02:57: It's becoming a survival imperative.

00:02:59: That's the phrase he used, survival imperative.

00:03:02: So pulling that all together.

00:03:04: The big takeaway for leaders it seems like it's disciplined investment really tracking the value and Well fundamentally rethinking how work gets done.

00:03:13: Yep

00:03:14: caution against the hype focus on the value.

00:03:16: Okay now Let's get into AI agents.

00:03:20: This is where it gets really interesting.

00:03:21: I think feels like a major leap.

00:03:23: Oh, absolutely.

00:03:24: This is way beyond just simple automation.

00:03:27: Richard Lefebvre defined agentic AI really well.

00:03:30: How do you put it

00:03:32: as systems that can autonomously execute workflows?

00:03:35: crucially correct their own errors and adapt strategies.

00:03:39: He said they act more like a colleague than a calculator.

00:03:42: A colleague, not a calculator.

00:03:43: I like that framing.

00:03:44: Yeah.

00:03:44: And he gave those great examples.

00:03:45: You know, the evolution from Wally, the tireless steward to R-to-D to the problem solver, and then to Jarvis, the sort of strategic augmenter.

00:03:55: It shows that path towards intelligence that really enhances what we can do.

00:03:59: Makes it very tangible.

00:04:00: And these aren't just sci-fi concepts now.

00:04:02: Alex's Cova claimed businesses using these agents, growing twice as fast.

00:04:06: Two times faster, wow.

00:04:08: Yeah.

00:04:08: Siting examples like lead gen agents, sales outreach, content creation, even competitive intelligence.

00:04:14: And they can be built relatively quickly, in weeks, apparently.

00:04:17: So real business value right now.

00:04:18: Exactly.

00:04:19: And Jake Jones added another layer talking about how agents tackle operational friction.

00:04:24: Operational friction, the work between the work.

00:04:27: That's

00:04:27: it.

00:04:28: Triaging requests, gathering info.

00:04:31: All that messy stuff that bogs teams down, not just automating single tasks, but smoothing the whole flow.

00:04:37: Reducing the chaos.

00:04:39: That's huge.

00:04:40: But we need some caution here too.

00:04:43: Sure.

00:04:43: Michael Short pointed out the term AI agent implies agency responsibility.

00:04:48: Yeah, which current tools don't really have.

00:04:50: Human accountability is still absolutely central.

00:04:52: We can't forget that.

00:04:53: Good point.

00:04:54: And Blake Oliver made a useful distinction between like linear AI workflows and true AI agents that have goals and autonomy.

00:05:02: Right,

00:05:02: but he also mentioned their limits, like terrible memories, context window issues.

00:05:07: They're powerful, but not magic.

00:05:09: Still limitations to work around.

00:05:11: For sure.

00:05:12: So this all points towards a pretty massive shift in how we work.

00:05:15: Gustavo Valbuena called it the agentic era.

00:05:18: The agentic era.

00:05:19: makes sense.

00:05:19: And

00:05:19: a lot of Brickner said HR needs to start planning now.

00:05:22: How do you onboard manage, integrate AI agents into the workforce?

00:05:25: Yeah, like hiring a whole new... type of team member almost, fascinating implications for HR.

00:05:30: Okay,

00:05:31: let's pivot slightly.

00:05:32: How is all this AI actually showing up in new products, new platforms, changing how we interact with tech day to day?

00:05:40: Well, one interesting thing was Dan Williams breaking down data bricks, agent bricks.

00:05:44: Agent bricks, okay.

00:05:45: Yeah,

00:05:45: the promises you pointed out your documents and boom, you get a working AI agent.

00:05:49: Pretty cool.

00:05:50: But he noted you need the right setup.

00:05:53: Ah, the catch.

00:05:54: Always a catch needs their unity catalog, that data governance piece and serverless compute.

00:05:59: So specific requirements.

00:06:00: Got it.

00:06:01: And on the creative side, I saw pretty hectic deep dive on Google's nano banana.

00:06:07: Sounds funny.

00:06:08: but looks

00:06:08: powerful.

00:06:09: Banana banana, tell me more.

00:06:11: It's an image editing model.

00:06:12: Does things like identity preservation, prompt to pixel pipeline.

00:06:16: It really seems like the future for creative work, making complex editing much more intuitive.

00:06:20: Very cool.

00:06:21: And Elliot Boreffa had some good advice too saying, great AI content isn't magic.

00:06:26: It's about high quality training data, good references, getting beyond just generic stuff.

00:06:30: Quality in, quality out, still applies.

00:06:33: Absolutely.

00:06:34: And speaking of interaction changes, Robert Stupak had this pretty provocative take on search.

00:06:41: Oh, yeah.

00:06:41: He sees us moving from SEO, you know, search engine optimization to Genio, generative engine optimization.

00:06:48: Genio, okay.

00:06:49: And shifting from browsing websites to a gentic commerce.

00:06:52: He even floated the idea that AI overviews in chat could mean, well, the end of websites.

00:06:59: Whoa.

00:07:00: The end of websites.

00:07:01: That is bold.

00:07:02: It's a big thought, right?

00:07:03: It changes how we access everything online.

00:07:05: Definitely food for thought.

00:07:07: And more practically.

00:07:08: Martin Crowley had a guide on picking AI video tools like Google, VO-II, Sora, Runway.

00:07:14: Yeah,

00:07:14: based on quality, speed, what you need it for.

00:07:16: His point was, don't just stick to one.

00:07:18: Makes sense.

00:07:18: And Rachelle Schmerzel talked about using our retrieval augmented generation.

00:07:22: Ah, REG, combining generative AI with real-time data.

00:07:26: Exactly.

00:07:26: For virtual agents and contact centers.

00:07:29: Making them way more accurate by pulling in live business info.

00:07:32: Super useful application.

00:07:33: Okay, let's zoom out a bit now.

00:07:35: Look at the bigger picture.

00:07:37: The partnerships, the ecosystem moves, what's happening regionally.

00:07:40: Yeah, there's some significant moves there.

00:07:42: Mark Calder had reported on Broadcom partnering with OpenAI.

00:07:45: To design AI accelerator chips, right?

00:07:47: Exactly.

00:07:48: That's a big deal.

00:07:49: Puts them right in the ring within video, basically.

00:07:53: Shows hardware is becoming this co-developed thing.

00:07:55: Interesting.

00:07:56: And in fintech.

00:07:57: Oh.

00:07:58: Seemed like a lot going on.

00:07:59: Oh,

00:07:59: yeah.

00:08:00: Linus Biliunas and Simon Taylor flagged a bunch of stuff.

00:08:04: AI powered digital banks like Right Bank, New Bank Expanding, Gemini's new XRP card.

00:08:10: And Visa dropping its agentic commerce APIs, which was notable.

00:08:14: Right.

00:08:15: Shows how fast things are moving and shifting in finance tech.

00:08:18: Definitely.

00:08:18: And connecting that globally, Dr.

00:08:20: Stefan Samble highlighted what's happening in Europe, specifically the IPAI project in Germany.

00:08:25: The

00:08:26: Dieter Schwartz Foundation thing.

00:08:27: That's the one in Heilbronn.

00:08:29: being called Europe's most ambitious supplied AI project.

00:08:32: Really trying to bridge research and actual application.

00:08:35: Plus

00:08:35: they just launched Jupiter, right?

00:08:37: Europe's fastest supercomputer.

00:08:39: Yep.

00:08:39: So real push for, you know, European AI capability and maybe sovereignty.

00:08:44: And

00:08:45: it's not just business and research.

00:08:46: AI is expanding into defense too.

00:08:48: Yuri Jackson have mentioned a Finnish company, Kelu.

00:08:51: With the autonomous hydrogen airship.

00:08:53: Yeah, participating in a NATO exercise.

00:08:55: Shows AI's reach.

00:08:57: and Borigurr appointed to the Smart AI twenty-twenty-five event, focusing purely on industrial AI, cognitive robotics, smart manufacturing.

00:09:06: So the ecosystem is really branching out everywhere, every sector.

00:09:09: So, okay, let's bring it back down to Earth.

00:09:12: What does this all mean for day-to-day operations?

00:09:14: Let's dive into some real-world adoption examples.

00:09:17: Use cases.

00:09:17: Yeah, good idea.

00:09:19: Shri Elaprolou shared a fantastic AI for good example from Buenos Aires.

00:09:23: Boti, their city assistant.

00:09:24: Exactly.

00:09:25: Using a gentic AI land graph, Amazon Bedrock to handle millions of citizen conversations a month.

00:09:30: accurate culturally relevant responses.

00:09:33: Really impressive stuff.

00:09:34: That is impactful.

00:09:35: Oh.

00:09:35: And maybe closer to home for some listeners, Varun Rana, noted the New York subway is adopting agentic AI.

00:09:41: For efficiency, yeah.

00:09:42: He had that funny line.

00:09:43: MTA, but the A stands for agentic.

00:09:45: Yeah.

00:09:46: It shows even, you know, traditional public infrastructure is getting on board.

00:09:49: Definitely.

00:09:50: And in banking, Vidya Vargavan wrote about using AI for frictionless, personalized services.

00:09:56: She mentioned a FAST framework.

00:09:58: Frictionless?

00:09:59: adaptive, secure, transparent.

00:10:02: That's it!

00:10:03: Moving towards proactive guidance, not just transactions.

00:10:06: And healthcare.

00:10:06: Wow, lots happening there.

00:10:08: Seems like it.

00:10:09: Dr.

00:10:09: Joachim H. shared how InfectoFarm built generative AI in-house.

00:10:15: Proving pharma can do it themselves.

00:10:17: Robert Perl highlighted how AI tools are empowering patients, changing that doctor-patient dynamic.

00:10:23: Making

00:10:23: it more collaborative.

00:10:24: Yeah.

00:10:25: And John Baker had this really counterintuitive finding.

00:10:28: healthcare workers who embrace the paradoxes around AI.

00:10:31: They actually adopt it more and are happier with it.

00:10:33: Interesting psychological angle there.

00:10:35: Isn't

00:10:35: it?

00:10:36: And maybe the biggest news.

00:10:38: Ravichandran Subya reported Russia is starting human trials for an AI made cancer vaccine.

00:10:42: A

00:10:42: cancer vaccine designed by AI.

00:10:45: That's potentially revolutionary.

00:10:46: It's truly

00:10:47: groundbreaking if it pans out.

00:10:49: And let's not forget HR.

00:10:50: Some year Pinocchalapatti's podcast was asking big questions.

00:10:55: How will agents change recruitment?

00:10:57: Rejecting candidates.

00:10:59: Predicting quits.

00:11:01: What is the recruiter of twenty thirty?

00:11:03: Yeah, HR seems ripe for this agentic shift.

00:11:06: But despite all these amazing examples, Martinez Nicolaeva threw a bit of cold water on things.

00:11:11: He's skeptical about the high adoption rates we hear about.

00:11:14: Oh, why is that?

00:11:15: He thinks many businesses are just scratching the surface.

00:11:18: You know, using AI for basic search, summarizing, not really embedding it deeply.

00:11:24: So the real potential is still untapped.

00:11:26: That's his view.

00:11:27: He thinks the big gains will come from fusion with other tech making AI almost invisible in the stack.

00:11:32: Yeah, got a ways to go

00:11:33: make sense.

00:11:34: There's hype and then there's deep integration.

00:11:36: right which brings us neatly to Governance legal policy.

00:11:41: This is where the conversation about responsibility really heats up the guardrails.

00:11:45: Absolutely

00:11:45: crucial.

00:11:46: And the big one here, mentioned by both Dr.

00:11:48: Nils Rower and Mirot Dermis, is the EU AI

00:11:51: Act.

00:11:51: Because of its global reach.

00:11:53: Right.

00:11:53: Exactly.

00:11:54: Its extraterritorial effect means businesses pretty much everywhere need to pay attention.

00:11:59: It's becoming the benchmark, the de facto global standard.

00:12:03: Huge implications for compliance worldwide.

00:12:05: Definitely.

00:12:06: And underpinning all that, Michael Charles Borelli stressed the need for solid data governance and record keeping.

00:12:12: Audit-ready AI programs.

00:12:13: Yeah.

00:12:14: You need that traceability.

00:12:16: Proof the system does what you say it does and behaves ethically.

00:12:19: Can't just let it run wild.

00:12:21: No way!

00:12:22: Which ties into copyright too.

00:12:23: Dr.

00:12:24: Barry Scannell reported on that massive anthropic settlement.

00:12:27: A one point five billion dollar one?

00:12:29: for using pirated books.

00:12:31: That's the one.

00:12:31: It basically put a concrete price on creative works used for training AI, signals a big shift towards structured licensing, pay upfront, basically.

00:12:39: Shipper than court.

00:12:40: Yeah.

00:12:40: Pretty

00:12:40: much.

00:12:41: And Jan Bernordeman noted that AIPPI, the Intellectual Property Association, is looking at global rules for AI and copyright, harmonization, fair pay for creators.

00:12:50: Important

00:12:50: work.

00:12:51: And beyond legal, Matthew Kay introduced his AI standards prompt map.

00:12:56: Integrating AI with global standards like ISO.

00:12:58: Yeah, for HR, IT, finance, making sure AI outputs are compliant, repeatable, scalable, building operational excellence.

00:13:07: Smart.

00:13:08: Ties it back to existing best practices.

00:13:09: And Giuseppe Stigliano had that wonderful analogy.

00:13:12: Raise AI like you would raise a child.

00:13:15: Meaning.

00:13:16: Meaning that early guidance values, mindset, attitude is critical for its ethical development, socially, politically, economically.

00:13:24: How we shape it now matters immensely.

00:13:27: Profound point.

00:13:28: It needs ethical foundations from the start.

00:13:30: Exactly.

00:13:31: So, okay, let's get into those foundations.

00:13:32: Data, infrastructure, MLO memes, the plumbing behind it all.

00:13:37: The unsung heroes, yeah.

00:13:39: And Ellie Gareffa's point comes back here.

00:13:41: Good AI content.

00:13:43: It's all about the training data, the references.

00:13:45: And retrieval data labeling.

00:13:46: You need that infrastructure at scale.

00:13:47: You do.

00:13:48: But Damian Millay added a really crucial nuance.

00:13:51: Improving AI results isn't just about better models or cleaner data.

00:13:55: What else

00:13:55: then?

00:13:55: It's often about refining the process, the human-AI interaction, the guardrails.

00:14:00: He shared an example where a big mistake was caught because the process failed safe, not because the model was perfect.

00:14:05: That's powerful.

00:14:06: Process over perfection.

00:14:08: Right.

00:14:08: And architecturally, Dr.

00:14:09: Sebastian Ronecki suggested we ditch the old AI onion model.

00:14:13: For something more practical.

00:14:14: Yeah.

00:14:15: He proposed two modes.

00:14:16: machine learning mode, optimizing specific outcomes from structured data, and AI mode, developing broad capabilities for unstructured stuff.

00:14:24: Helps make smarter investment and design choices.

00:14:27: More practical for building things out.

00:14:28: Definitely.

00:14:29: And we saw advanced infrastructure in action with Prem Narina's demo.

00:14:33: Use the Grock LPUs, those super fast chips.

00:14:36: Yeah, with Humane and Katonic AI.

00:14:38: real-time workflow automation.

00:14:40: Turning conversations straight into JIRA tickets, he said, intent is now programmable.

00:14:45: Wow, shows that seamless integration.

00:14:47: Exactly.

00:14:48: So it really underscores that robust infrastructure choices aren't just about speed.

00:14:53: They're fundamental for performance, security, compliance,

00:14:58: all of it.

00:14:58: And as Shalini Rao pointed out, you need that AI-ready data and governance baked in right from the start.

00:15:04: Okay,

00:15:04: let's shift gears one last time to the human element.

00:15:07: workforce, skills, leadership.

00:15:09: How is AI reshaping all of that?

00:15:11: Well, leadership is definitely changing.

00:15:14: Frederick De Bruyck was clear.

00:15:15: AI and leadership is non-negotiable

00:15:17: now.

00:15:17: Non-negotiable.

00:15:18: Yeah.

00:15:19: Leaders need that blend of empathy and smart data insights to stay relevant.

00:15:23: It's about building trust, not just, you know, ticking AI boxes.

00:15:26: Right.

00:15:27: And Gustavo Valbuena added that AI isn't just a side topic in leadership discussions anymore.

00:15:32: It's central.

00:15:32: Integral.

00:15:33: Focusing on maximizing benefits for everyone, the organization and the employees.

00:15:38: Support, productivity, better experience.

00:15:41: That's a positive framing.

00:15:42: It is.

00:15:43: And HR's role here is just massive.

00:15:46: Alana Brickner, drawing on Service Now's playbook, talked about HR becoming AI enablers.

00:15:51: AI enablers meaning.

00:15:52: Reinvesting the time AI frees up, building a culture of transparency, tackling that huge skills agenda.

00:15:57: A skills

00:15:57: gap is huge, right?

00:15:59: What was the stats?

00:16:00: Sixty percent of workers needing reskilling by twenty thirty.

00:16:03: Sixty percent.

00:16:04: Oh, that's urgent.

00:16:05: Extremely.

00:16:06: And David Green curated resources on actually deploying AI in HR, reinventing performance management for this new reality.

00:16:14: Of course, this inevitably brings up job displacement concerns.

00:16:17: Felicity Menzies raised a really interesting point about entry-level roles.

00:16:21: What was your concern?

00:16:22: That AI displacing those roles threatens the whole talent pipeline.

00:16:27: You lose that foundational learning ground.

00:16:29: It creates a delayed feedback problem.

00:16:31: efficiency now, but knowledge gaps later.

00:16:34: That's a really thoughtful point.

00:16:35: The long-term impact.

00:16:36: Yeah.

00:16:37: On the flip side... Pascal Burnett and Joanna Masijuska argued AI should take the mundane tasks.

00:16:42: To free us up from more creative work.

00:16:44: Exactly.

00:16:45: But they also warned about AI obesity.

00:16:48: Getting lazy and letting the AI do everything.

00:16:51: Huh.

00:16:51: AI obesity.

00:16:52: Need to watch out for that.

00:16:53: We do.

00:16:54: And Tim Evans talked about AI in education.

00:16:57: Going beyond just summarizing text, using it with things like Bloom's taxonomy for deeper transformative learning.

00:17:02: Empowering educators and students, not just automating tasks.

00:17:06: Precisely.

00:17:07: And the whole human-AI relationship is still unfolding.

00:17:10: Bernard Morris chat with Ameca, the humanoid robot.

00:17:13: Yeah, raise those big questions.

00:17:14: Is it authentic connection or just assistance?

00:17:18: What are the ethics as they become part of daily life?

00:17:21: Deep questions.

00:17:22: But Georgina Morris had a more grounded hot take.

00:17:25: Oh, yeah.

00:17:25: She argued, AI won't replace us soon because it lacks true creativity.

00:17:30: It relies on past data, human curiosity, specialized insight.

00:17:35: that's still irreplaceable.

00:17:36: A good reminder of unique human value.

00:17:38: Definitely.

00:17:40: Okay, let's wrap up the insights by looking at the market.

00:17:42: Sentiment, investment, what's the vibe out there?

00:17:45: Well, James Gosling and Eric Vesto both kind of sounded a cautionary note, drawing parallels to past tech bubbles like VR.

00:17:53: Seeing the hype is ahead of reality.

00:17:54: Pretty much.

00:17:55: They acknowledged AI tech is valuable, but the hype, way ahead of sensible outcomes, lots of startups they weren't, are probably doomed to lose.

00:18:04: It is.

00:18:05: And Dr.

00:18:06: Jeffrey Funk echoed that, comparing the spending on AI talent, those huge salaries to the dot-com era.

00:18:11: Questioning the actual commercial return.

00:18:13: Yeah,

00:18:14: from G-Wiz technology that can't be trusted, as he put it.

00:18:17: But Craig Skroggy offered a bit more balance.

00:18:20: He observed that AI hype and AI reality can coexist.

00:18:25: Quoting Altman and Gates, we overestimate the short-term, underestimate the long-term, his advice, long-term discipline in investing.

00:18:33: Makes sense.

00:18:34: Don't get swept away, but don't ignore the potential either.

00:18:36: Exactly.

00:18:37: And in this whole environment, Cassie Kozierkoff asked that killer leadership question about accountability.

00:18:42: When you build on foundation models you didn't create.

00:18:44: Right.

00:18:44: Who's accountable for the outputs?

00:18:46: Her answer.

00:18:47: The leaders deploying the AI.

00:18:49: you can't outsource all accountability.

00:18:51: That's

00:18:51: a crucial takeaway.

00:18:52: The buck stops with the Deployer.

00:18:54: If you enjoyed this deep dive, new deep dives drop every two weeks.

00:18:57: Also check out our other editions on ICT and tech, digital products and services, cloud and sustainability and green ICT.

00:19:03: So as we've kind of journeyed through this incredible AI momentum, I mean, from self-correcting agents, managing workflows all the way to potentially an AI made cancer vaccine.

00:19:14: It feels like one question really stands out for everyone, shaping this future.

00:19:17: Are we, you know, actively designing the guardrails, the ethical foundations for these incredibly powerful systems, or are we just sort of reacting?

00:19:26: Reacting to how fast they're evolving and just, well, trusting they'll grow upright all on their own.

00:19:30: Something to think about.

00:19:31: Thank you for joining us for this deep dive into the latest AI insights.

00:19:35: Make sure you subscribe so you don't miss our next deep dive and continue shaping the future with us.