Latest / Best of LinkedIn: Digital Transformation & Tech / Best of LinkedIn: Artificial Intelligence CW 49/ 50

Transcript

Best of LinkedIn: Artificial Intelligence CW 49/ 50 · Best of LinkedIn: Digital Transformation & Tech

• 0:00
  Speaker 0: This deep dive is provided by Thomas Allgaier and Frennus, based on the most relevant LinkedIn posts about artificial intelligence in calendar weeks, forty nine and fifty.
• 0:09
  Speaker 0: Frennus supports enterprises with market and competitive intelligence, decoding emerging technologies, customer insights, regulatory shifts and competitor strategies.
• 0:19
  Speaker 0: So product teams and strategy leaders don't just react, but shape the future of AI.
• 0:25
  Speaker 0: Welcome to the deep dive.
• 0:28
  Speaker 1: So this week, we're digging into a huge pile of insights from LinkedIn, really looking at the top artificial intelligence trends that we're defining the end of the year.
• 0:37
  Speaker 0: Yeah, and if you're expecting another chat about the latest model, writing better poetry, it's not what's happening.
• 0:43
  Speaker 0: The whole conversation is shifted.
• 0:44
  Speaker 0: It's no longer about the novelty of it all.
• 0:46
  Speaker 0: It's now about control.
• 0:47
  Speaker 0: It's about execution and building actual systems.
• 0:50
  Speaker 0: We're seeing leaders, engineers, regulators, they're all grappling with how to govern, orchestrate, and you know, critically, how to secure AI when it leaves the lab.
• 1:00
  Speaker 1: So it's moved from what if to how to.
• 1:02
  Speaker 1: Exactly.
• 1:03
  Speaker 0: The focus is one hundred percent on measurable value now.
• 1:06
  Speaker 0: OK, so let's unpack all that.
• 1:07
  Speaker 0: Our mission here is to pull out the most important nuggets for you if you're a professional in the ICT and tech industry.
• 1:13
  Speaker 1: And we're going to cover everything from why these huge pilots are failing to which global regulations you absolutely cannot ignore.
• 1:21
  Speaker 0: And that failure point, I mean, that's where we have to start.
• 1:23
  Speaker 0: The data is pretty sobering.
• 1:26
  Speaker 1: It really is.
• 1:27
  Speaker 0: Let's get into it.
• 1:28
  Speaker 0: The hard truth about adoption.
• 1:31
  Speaker 0: We're calling this first theme AI strategy and the execution gap.
• 1:36
  Speaker 0: The big consensus we're seeing is that AI underperformance, it isn't really a model quality problem.
• 1:43
  Speaker 0: It's a failure of governance and just, well, weak execution.
• 1:47
  Speaker 1: It's even more than that.
• 1:48
  Speaker 1: It's actually a retreat.
• 1:49
  Speaker 1: John Nordmark shared this truly surprising data point.
• 1:52
  Speaker 1: I
• 1:52
  Speaker 0: saw that one.
• 1:53
  Speaker 1: AI adoption in big companies is actually declining, not rising, not even plateauing.
• 1:57
  Speaker 1: They tried it and now they're pulling back.
• 1:59
  Speaker 0: Declining, so not just slowing down, but actually going backwards.
• 2:02
  Speaker 0: That completely flies in the face of the industry narrative.
• 2:05
  Speaker 1: It does.
• 2:05
  Speaker 1: And Stephanie Gradwell frames the why perfectly.
• 2:09
  Speaker 1: She says most leadership teams don't really have an AI problem.
• 2:12
  Speaker 1: OK.
• 2:12
  Speaker 1: They have an execution problem.
• 2:14
  Speaker 1: The models, they work beautifully in the lab, you know, and a proof of concept.
• 2:17
  Speaker 1: In a clean room.
• 2:18
  Speaker 1: Exactly.
• 2:19
  Speaker 1: But then you drop it into the messy reality of business workflows and legacy systems.
• 2:25
  Speaker 1: and the core business KPIs, they don't change.
• 2:28
  Speaker 1: They just can't seem to bridge that gap from a successful pilot to actual enterprise value.
• 2:34
  Speaker 0: So the challenge isn't building the shiny new tool.
• 2:38
  Speaker 0: It's redesigning the entire factory floor to actually use it.
• 2:41
  Speaker 1: That's the perfect analogy.
• 2:43
  Speaker 1: And the solution we're seeing mentioned everywhere is a return to basics.
• 2:46
  Speaker 1: Start small, stay focused, iterate.
• 2:48
  Speaker 1: Right.
• 2:49
  Speaker 1: Kaushik Bay suggests this very practical four-step approach.
• 2:52
  Speaker 1: First, you rigorously map your workflow pain points.
• 2:55
  Speaker 0: Okay, standard stuff.
• 2:56
  Speaker 1: Right.
• 2:56
  Speaker 1: But then second, you pick a high-frequency low-risk task, something that won't, you know, crash the business if it fails.
• 3:01
  Speaker 1: Third, you automate it with a controlled agent.
• 3:04
  Speaker 1: And then finally, you measure the impact for a very strict thirty days.
• 3:07
  Speaker 0: Wait, but how is that different from what we were doing five years ago?
• 3:10
  Speaker 0: You know, mapping pain points, running thirty-day tests.
• 3:14
  Speaker 1: That's a great question.
• 3:16
  Speaker 1: The difference is the type of automation.
• 3:18
  Speaker 1: That this isn't just scripting a task.
• 3:20
  Speaker 1: It's about introducing a controlled agent that can operate with some autonomy.
• 3:24
  Speaker 1: And if you get that quick, quantifiable ROI win, that's your first.
• 3:30
  Speaker 1: internal success story.
• 3:32
  Speaker 0: And that's what unlocks the budget.
• 3:33
  Speaker 1: That's what unlocks the budget and the will for a bigger transformation, which brings us all back to leadership.
• 3:39
  Speaker 1: Neil Harrison reflected on this.
• 3:41
  Speaker 1: He said the failure in twenty twenty-five adoption wasn't a lack of ambition.
• 3:45
  Speaker 0: It was a lack of leadership capability.
• 3:47
  Speaker 1: Exactly.
• 3:48
  Speaker 0: Real
• 3:48
  Speaker 1: success means redesigning the work itself for human AI collaboration, not just trying to bolt a new tool into a broken process and that change management, that's the hard part.
• 3:58
  Speaker 0: That idea of integrated systems, of systems that collaborate, that's a perfect segue into our next big theme.
• 4:03
  Speaker 1: Absolutely.
• 4:04
  Speaker 1: The focus has really moved away from just a single model's performance and toward what everyone's calling the agentic shift and orchestration.
• 4:11
  Speaker 0: Okay, the word agent is everywhere right now.
• 4:15
  Speaker 0: What does that actually mean for someone in tech?
• 4:17
  Speaker 1: Well, Breesh Gashore Pandey points out that the foundational models, the LLMs, they're becoming commodities.
• 4:23
  Speaker 1: They're interchangeable.
• 4:24
  Speaker 1: Right.
• 4:24
  Speaker 1: The real value, and so the real engineering focus, has shifted to what he calls agentic stacks.
• 4:31
  Speaker 1: These are the system layers that handle things like memory, complex execution, and the orchestration that controls the whole process.
• 4:39
  Speaker 0: This is where it gets really interesting for me.
• 4:41
  Speaker 0: Nandan Malakara defines what agents do differently.
• 4:45
  Speaker 0: They plan, so they break big tasks into smaller steps.
• 4:49
  Speaker 0: They coordinate systems, they learn from feedback, and this is the key part, they intelligently invoke other tools to act.
• 4:56
  Speaker 0: They're not just tools, they're intelligent partners.
• 4:58
  Speaker 1: And this isn't just hype, it's delivering real economic results.
• 5:01
  Speaker 1: Richard Turen reported on an anthropic survey.
• 5:04
  Speaker 1: It found eighty percent of technical leaders are already reporting measurable economic returns from agents.
• 5:09
  Speaker 0: Eighty percent, that's huge.
• 5:11
  Speaker 1: It is.
• 5:12
  Speaker 1: The most mature area, no surprise, is software development.
• 5:15
  Speaker 1: Adoption there is at ninety percent.
• 5:18
  Speaker 1: We're seeing examples like Neobank N-twenty-six achieving seventy percent automation in certain processes.
• 5:24
  Speaker 0: Okay, eighty percent measurable return.
• 5:26
  Speaker 0: Are we talking about small pilot projects here?
• 5:28
  Speaker 0: or is this scalable, sustained ROI?
• 5:31
  Speaker 1: That is the big question.
• 5:33
  Speaker 1: And the industry's answer is to build the infrastructure to make it scalable.
• 5:37
  Speaker 1: That's where orchestration comes in.
• 5:39
  Speaker 1: You can't just have a fleet of autonomous agents running wild.
• 5:42
  Speaker 0: You need a control plane.
• 5:43
  Speaker 1: You need a control plane.
• 5:45
  Speaker 1: Vishal Mishra announced Microsoft's agent three sixty five framing it exactly as that the new enterprise control plane
• 5:52
  Speaker 0: control plane.
• 5:53
  Speaker 0: so the foundational models are becoming interchangeable.
• 5:55
  Speaker 0: isn't creating a Microsoft controlled highway just trading one kind of vendor lock-in for another?
• 6:00
  Speaker 0: that's
• 6:00
  Speaker 1: the strategic risk for sure but it shows how urgent the need is for a system of systems they call it the highway traffic control and safety system for all your AI agents.
• 6:10
  Speaker 1: the point is leaders know they need centralized management to scale this safely
• 6:14
  Speaker 0: because that creates a massive amount of risk.
• 6:16
  Speaker 1: Massive risk.
• 6:18
  Speaker 1: Rob Vanderveer flagged cross user prompt injection as the single biggest threat in agentic AI right now.
• 6:24
  Speaker 0: So what is that exactly?
• 6:26
  Speaker 0: Imagine
• 6:26
  Speaker 1: you have a privileged automated agent scanning say a public forum.
• 6:32
  Speaker 1: An attacker can sneak a malicious instruction, the prompt injection, into a totally normal looking comment.
• 6:38
  Speaker 0: And because the agent is designed to follow instructions.
• 6:40
  Speaker 1: It follows the malicious one.
• 6:42
  Speaker 1: That instruction could tell it to leak sensitive data, delete files, change configurations, anything.
• 6:48
  Speaker 0: Wow.
• 6:48
  Speaker 1: So the recommended defense is pretty harsh, but it's necessary.
• 6:52
  Speaker 1: You have to limit an agent's access to untrusted data dynamically.
• 6:56
  Speaker 1: You have to treat all agents as potential bad actors and operate with zero trust.
• 7:01
  Speaker 0: That security concern flows right into the regulatory landscape, which is trying to build a cage around all this
• 7:06
  Speaker 1: power.
• 7:06
  Speaker 1: Let's turn to that.
• 7:07
  Speaker 1: governance, trust, and the regulatory reality.
• 7:11
  Speaker 1: The EU AI Act is really anchoring this whole global debate.
• 7:14
  Speaker 0: Yeah, but there's a lot of, let's say, paranoia out there.
• 7:18
  Speaker 0: Before teams panic and start huge compliance projects, Alex Sureski gives some simple advice.
• 7:23
  Speaker 0: Just as the first question, have you checked whether the AI Act even applies to your system at all?
• 7:28
  Speaker 0: So many teams just slap AI enabled on a marketing page and create their own regulatory headache.
• 7:34
  Speaker 1: Right.
• 7:34
  Speaker 1: You have to look at the actual definitions.
• 7:36
  Speaker 1: Kimberly Friesen shared some key indicators for figuring this out.
• 7:40
  Speaker 1: A regulated system has to be machine-based, have a certain level of autonomy, and that threshold can be pretty low.
• 7:46
  Speaker 1: Okay.
• 7:47
  Speaker 1: And most importantly, it has to have a real influence on a physical or virtual environment.
• 7:53
  Speaker 1: A lot of simple prediction tools just fall outside of that scope.
• 7:56
  Speaker 0: But if your system does fall into that high-risk category, things like credit scoring, underwriting, hiring tools, what are the immediate requirements?
• 8:04
  Speaker 1: Hegbert Weatherborne lists them out.
• 8:05
  Speaker 1: You need rigorous risk management, demonstrable data quality controls, detailed technical documentation, mandated human oversight, and verifiable accuracy.
• 8:14
  Speaker 1: It's non-negotiable.
• 8:15
  Speaker 0: And the penalties are serious.
• 8:17
  Speaker 0: We're talking up to fifteen million or three percent of your global annual revenue.
• 8:21
  Speaker 0: So what's the blueprint for actually doing all that?
• 8:24
  Speaker 1: Gurdiep Singh Chopra positions ISO forty two thousand one as that blueprint.
• 8:30
  Speaker 1: It's the international standard for an AI management system.
• 8:33
  Speaker 1: It basically operationalizes the law.
• 8:35
  Speaker 1: It tells you how to systematically meet those legal obligations.
• 8:38
  Speaker 0: And for all of us who rely on third-party vendors for this stuff, what do we need to be demanding from them?
• 8:44
  Speaker 1: Patricia Bertini says you have to proactively demand responsible AI documentation.
• 8:50
  Speaker 1: That means getting training data summaries, a clear copyright compliance policy, and model documentation forms.
• 8:56
  Speaker 1: If a vendor can't give you that.
• 8:58
  Speaker 1: It's a huge red flag.
• 8:59
  Speaker 1: It's a huge red flag.
• 9:00
  Speaker 1: And it means the legal exposure falls squarely back on you, the client.
• 9:03
  Speaker 0: What's fascinating is the global fragmentation here.
• 9:06
  Speaker 0: Lexi Reese discussed the new one rule AI executive order in the U.S.
• 9:10
  Speaker 0: from the Trump administration.
• 9:11
  Speaker 0: It favors voluntary standards.
• 9:13
  Speaker 1: A completely different philosophy from the EU.
• 9:15
  Speaker 1: Exactly.
• 9:16
  Speaker 0: And that creates a massive strategic headache for global companies.
• 9:19
  Speaker 0: The reality is you have to build your systems to the strictest standard, which is the EU AI Act, and then kind of work backwards from there.
• 9:25
  Speaker 1: So the EU is setting the global bar effectively.
• 9:29
  Speaker 1: They
• 9:29
  Speaker 0: are.
• 9:29
  Speaker 0: And then you have countries like Italy.
• 9:31
  Speaker 0: According to Bethesimo Novec, going even further, they're passing their own national laws to criminalize harmful deep fakes, which goes beyond even the EU framework.
• 9:41
  Speaker 1: It's a fractured landscape.
• 9:43
  Speaker 0: It really is.
• 9:44
  Speaker 0: And all that complexity hits the engineering teams directly.
• 9:48
  Speaker 0: Which brings us to our next theme.
• 9:50
  Speaker 1: Right.
• 9:50
  Speaker 1: Engineering, IP, and the future of work.
• 9:53
  Speaker 1: The tools are changing so fast, the roles have to change with them.
• 9:57
  Speaker 0: Absolutely.
• 9:57
  Speaker 0: For developers, the shift is huge.
• 10:00
  Speaker 0: Dr.
• 10:00
  Speaker 0: Ingir Ganem described it as moving from the busy bee who does hands-on coding to the air traffic controller who oversees, reviews, architects, and orchestrates the whole system.
• 10:10
  Speaker 0: The value isn't in how many lines of code you type anymore.
• 10:13
  Speaker 1: And that's all being driven by AI coding assistance.
• 10:16
  Speaker 1: Hague Lupesco shared some stats.
• 10:17
  Speaker 1: Ninety percent adoption reported.
• 10:19
  Speaker 1: Developers shipping sixty percent more pull requests.
• 10:22
  Speaker 0: Sounds amazing.
• 10:23
  Speaker 1: It does, but here's the catch.
• 10:24
  Speaker 1: Quality metrics like change failure rate, they don't consistently improve.
• 10:28
  Speaker 0: So adoption doesn't equal impact.
• 10:30
  Speaker 1: Not automatically?
• 10:32
  Speaker 1: You need a structured rollout and proper enablement to make sure you're getting high quality alongside that high velocity.
• 10:38
  Speaker 0: Which means the old problems haven't gone away.
• 10:40
  Speaker 0: Nick Kay pointed out that classic vulnerability, SQL injection, cross-site scripting are still common.
• 10:46
  Speaker 1: And worse, the LLM assistance can actually amplify those bad patterns if the developer doesn't have the fundamental knowledge to spot insecure code in the output.
• 10:55
  Speaker 0: Wow.
• 10:56
  Speaker 0: Okay, and adding another layer of risk to all this is the IP battle brewing in Europe.
• 11:02
  Speaker 1: Yes, the details from this landmark ruling in Germany from the Munich regional court against open AI They're absolutely vital for anyone in this space.
• 11:11
  Speaker 0: So what does the German court actually rule?
• 11:14
  Speaker 1: It was a case brought by Jima, the German music rights organization.
• 11:17
  Speaker 1: The court found that the AI's memorization and then reproduction of copyrighted song lyrics was infringement.
• 11:23
  Speaker 0: And they rejected the usual defense.
• 11:25
  Speaker 1: They rejected the text and data mining defense.
• 11:28
  Speaker 1: The argument that training on copyrighted data is fair use.
• 11:31
  Speaker 1: It didn't hold up.
• 11:32
  Speaker 1: The court basically said if the model reproduces the material, it means it memorized it and that's infringement.
• 11:38
  Speaker 0: which creates a ton of legal uncertainty because that ruling goes directly against a recent decision in the UK on the same issue.
• 11:45
  Speaker 1: Exactly.
• 11:46
  Speaker 1: It fractures the legal landscape for generative AI across Europe.
• 11:50
  Speaker 0: Let's zoom out for a second to the macroeconomic impact.
• 11:53
  Speaker 0: Envato Ordax reported a staggering statistic.
• 11:56
  Speaker 0: AI infrastructure was responsible for ninety-two percent of US GDP growth in the first half of twenty twenty-five.
• 12:03
  Speaker 1: Ninety-two percent.
• 12:04
  Speaker 1: That's not just a tech story.
• 12:05
  Speaker 1: That's the entire macroeconomic growth engine.
• 12:08
  Speaker 0: It's immense.
• 12:09
  Speaker 0: But Zemina Ahmad warns that it's leading to what she calls jobless growth.
• 12:13
  Speaker 0: Productivity is skyrocketing.
• 12:15
  Speaker 1: But employment in IT and business services is actually declining.
• 12:20
  Speaker 1: Companies are quietly pursuing a strategy of more output with fewer people, and that's hitting junior staff and freelancers the hardest.
• 12:26
  Speaker 0: A really sobering perspective.
• 12:28
  Speaker 0: It connects that initial execution gap we talked about to this ultimate goal of massive, efficient, and potentially job-displacing growth.
• 12:34
  Speaker 0: We've covered
• 12:35
  Speaker 1: a lot of ground, so to summarize this whole deem dive, the AI battleground It's not about the models anymore.
• 12:41
  Speaker 1: It's about the systems, the governance, and the organizational change you need to actually get real value.
• 12:47
  Speaker 0: Right.
• 12:47
  Speaker 0: The focus now is all on control, security, and global compliance.
• 12:53
  Speaker 0: And if we think about the long-term human cost of this, there's an interesting point from Borogur.
• 12:58
  Speaker 1: What's that?
• 12:59
  Speaker 0: He raised the risk of Gen Z. of younger workers heavily relying on AI for critical social health and career decisions.
• 13:07
  Speaker 1: And for us, as the professionals building and deploying these systems we just said need to be treated with zero trust, that raises a really important question.
• 13:14
  Speaker 1: Which is?
• 13:15
  Speaker 1: What kind of organizational liability are we creating when we inherit a workforce that's potentially de-skilled in critical thinking and independent judgment?
• 13:24
  Speaker 1: That vulnerability created by a reliance on the very systems we're trying to lock down, that's something every leader needs to be thinking about urgently.
• 13:31
  Speaker 0: If you enjoy this deep dive, new episodes drop every two weeks.
• 13:34
  Speaker 0: Also check out our other editions on ICT and tech, digital products and services, cloud, sustainability and green ICT, defense tech and health tech.
• 13:43
  Speaker 0: Thank you for joining us for this deep dive.
• 13:45
  Speaker 0: Don't forget to subscribe.