Latest / Best of LinkedIn: Digital Transformation & Tech / Best of LinkedIn: Artificial Intelligence CW 03/ 04

Transcript

Best of LinkedIn: Artificial Intelligence CW 03/ 04 · Best of LinkedIn: Digital Transformation & Tech

• 0:00
  Speaker 0: This episode is provided by Thomas Allgeier and Frennus based on the most relevant LinkedIn posts about artificial intelligence from calendar weeks O three and O four.
• 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.
• 0:24
  Speaker 1: It's going to be back.
• 0:25
  Speaker 1: And, you know, looking at the insights from the last couple of weeks in January, The whole tone just feels different.
• 0:32
  Speaker 0: It really does.
• 0:32
  Speaker 0: I mean, we're not in the playground anymore, are we?
• 0:34
  Speaker 1: Not at all.
• 0:35
  Speaker 0: If twenty-twenty-four and twenty-twenty-five were the wow years, you know, everyone gasping because a computer wrote a poem or something.
• 0:41
  Speaker 1: Yeah, the magic show.
• 0:42
  Speaker 1: Exactly.
• 0:43
  Speaker 0: Then early twenty-twenty-six feels like, I don't know, the year of the hangover, or maybe sobering up is a better way to put it.
• 0:49
  Speaker 1: I think sobering up is fair.
• 0:51
  Speaker 1: We're definitely seeing a narrative arc.
• 0:52
  Speaker 1: The hype cycle is settling and we're entering a phase of what I call operating discipline.
• 0:57
  Speaker 1: The conversation is just pivoted so hard.
• 0:59
  Speaker 1: It's not, look what this model can do anymore.
• 1:02
  Speaker 1: Now it's... How do we run this safely, legally, and profitably at scale?
• 1:07
  Speaker 0: And that profitable part seems to be the one everyone's really struggling with.
• 1:11
  Speaker 0: It's the shift from magic tricks to mechanics.
• 1:13
  Speaker 1: For sure.
• 1:14
  Speaker 0: So for this deep dive, we're going to navigate that shift through a few key lenses.
• 1:18
  Speaker 0: We've got to talk about governance, which I know sounds dry, but wait until you hear the numbers.
• 1:23
  Speaker 0: Those
• 1:23
  Speaker 1: numbers are big.
• 1:24
  Speaker 0: Then there's the ticking clock in Europe with the AI Act, and then the tech itself.
• 1:29
  Speaker 0: Because while the lawyers are panicking, the engineers are shifting from chatbots to full-blown agents.
• 1:36
  Speaker 1: To the AI operating system.
• 1:37
  Speaker 1: Yeah, it's a dense couple of weeks, but I think you're right.
• 1:39
  Speaker 1: We have to start with a topic that's probably ruining sleep for every CEO right now.
• 1:43
  Speaker 0: Governance, yeah.
• 1:44
  Speaker 0: Let's unpack this, because usually you say governance and people's eyes glaze over.
• 1:48
  Speaker 0: They
• 1:48
  Speaker 1: tune out, yeah.
• 1:49
  Speaker 0: But looking at a post from Julian Noble, that attitude is basically a death wish in twenty-twenty-six.
• 1:55
  Speaker 1: It is.
• 1:56
  Speaker 1: Noble really laid it on the line.
• 1:57
  Speaker 1: He's saying air compliance has mutated.
• 2:00
  Speaker 1: It's no longer just an IT ticket.
• 2:02
  Speaker 1: Right.
• 2:02
  Speaker 0: It's not, hey, IT, can you check the software anymore?
• 2:04
  Speaker 0: No.
• 2:05
  Speaker 1: It's now a critical enterprise risk.
• 2:07
  Speaker 1: And he's not just talking about the tech breaking.
• 2:09
  Speaker 1: He's talking about exposure.
• 2:11
  Speaker 1: hidden exposure.
• 2:12
  Speaker 0: That's the key part.
• 2:13
  Speaker 0: This is that shadow AI problem.
• 2:15
  Speaker 0: The recruitment tool HR bought that's biased the marketing sauce that hallucinates claims about your product.
• 2:21
  Speaker 0: The board isn't even tracking half of this stuff.
• 2:24
  Speaker 1: And Noble points to the big financial stick here.
• 2:26
  Speaker 1: Yeah.
• 2:26
  Speaker 1: With the EU AI Act in full force.
• 2:28
  Speaker 1: We're not talking about a slap on the wrist.
• 2:30
  Speaker 1: No.
• 2:30
  Speaker 1: We're talking penalties up to thirty five million euros or seven percent of global turnover.
• 2:36
  Speaker 0: Seven percent.
• 2:36
  Speaker 0: That is a resume generating event for a CEO.
• 2:39
  Speaker 0: That is pack your box and leave money.
• 2:41
  Speaker 1: It absolutely is.
• 2:43
  Speaker 1: But here's the nuance, and Joanna Miley had a great insight on this.
• 2:46
  Speaker 1: Just reacting to that fear by trying to comply might also be the wrong
• 2:50
  Speaker 0: move.
• 2:50
  Speaker 0: What do you mean?
• 2:51
  Speaker 1: Well, she argues that viewing this as just AI Act plus done is a massive strategic error.
• 2:58
  Speaker 1: She talks about a layered control stack.
• 2:59
  Speaker 0: A control stack.
• 3:01
  Speaker 0: That sounds like a cybersecurity principle, like defense in depth, but for AI.
• 3:05
  Speaker 1: That's exactly it.
• 3:06
  Speaker 1: Think about a bank today.
• 3:08
  Speaker 1: You're not just juggling the AI Act.
• 3:10
  Speaker 1: You've got GDPR, NIS-II, DeWara's, all at the same time.
• 3:15
  Speaker 0: It's a regulatory minefield.
• 3:17
  Speaker 1: It is.
• 3:17
  Speaker 1: And Mylar's point is, you have to pick a control posture.
• 3:22
  Speaker 1: Are you doing the bare minimum, or are you building control first?
• 3:26
  Speaker 0: designs.
• 3:26
  Speaker 0: Which sounds great in theory, but that creates friction.
• 3:29
  Speaker 0: I mean, I saw Shea Brown's post about procurement and it really hit home.
• 3:33
  Speaker 0: He's seeing deals just stall out.
• 3:34
  Speaker 1: The computer says no moment.
• 3:36
  Speaker 0: Exactly.
• 3:37
  Speaker 0: And it's not because the startups have bad tech.
• 3:39
  Speaker 0: It's because they walk into a meeting and procurement pulls out a fifty page questionnaire.
• 3:43
  Speaker 0: Yep.
• 3:44
  Speaker 1: And Brown says they're failing because they can't answer the boring questions.
• 3:46
  Speaker 1: Yeah.
• 3:47
  Speaker 1: Are you ISO forty two thousand one certified?
• 3:49
  Speaker 1: Have you tested for bias?
• 3:51
  Speaker 1: Do you train on customer data?
• 3:52
  Speaker 1: And
• 3:52
  Speaker 0: the founder is just staring blankly wanting to show off their cool demo.
• 3:56
  Speaker 1: And the deal just dies right there.
• 3:58
  Speaker 1: Brown's insight is that if you can't get past procurement in twenty twenty six, you don't have a business.
• 4:03
  Speaker 1: It doesn't matter how smart your model is.
• 4:05
  Speaker 1: This
• 4:05
  Speaker 0: just feels like a reckoning for the move fast and break things crowd.
• 4:09
  Speaker 0: And Paula Falcon gave it a name I love.
• 4:11
  Speaker 0: Architectural bankruptcy.
• 4:13
  Speaker 1: It's a stinging phrase, isn't it?
• 4:14
  Speaker 1: It
• 4:15
  Speaker 0: hurts because it's true.
• 4:16
  Speaker 1: It is.
• 4:17
  Speaker 1: Falcon says if you spent a ton of money on AI in twenty twenty five without a governance framework, you didn't buy innovation.
• 4:25
  Speaker 1: You've funded architectural bankruptcy.
• 4:26
  Speaker 1: Wow.
• 4:27
  Speaker 1: The failures we're seeing aren't because people chose the wrong LLM.
• 4:31
  Speaker 1: They failed because of bad enterprise architecture and unstructured data.
• 4:35
  Speaker 0: You can't build a skyscraper on a swamp.
• 4:37
  Speaker 0: You just get a faster way to make mistakes.
• 4:39
  Speaker 1: And a faster way to get fined, which brings us, I think, to the geography of those fines.
• 4:44
  Speaker 1: We have to talk about Europe.
• 4:45
  Speaker 1: All
• 4:45
  Speaker 0: right.
• 4:45
  Speaker 0: Theme two, the clock is ticking.
• 4:48
  Speaker 0: Matt Elsie flagged a specific deadline that everyone needs to have circled in red.
• 4:52
  Speaker 1: August, that is the deadline for full obligations on high-risk systems.
• 4:56
  Speaker 1: Which
• 4:56
  Speaker 0: is only what, six months away?
• 4:58
  Speaker 0: And we should clarify, high-risk catches a lot more than just, you know, medical robots.
• 5:02
  Speaker 1: Oh, much more.
• 5:03
  Speaker 1: If your AI agent influences hiring, sets credit limits, determines insurance premiums, or denies refunds.
• 5:09
  Speaker 0: So basically half the back office of any modern company.
• 5:13
  Speaker 1: Pretty much.
• 5:14
  Speaker 1: And Elsie points out you need eight core pillars of compliance ready.
• 5:18
  Speaker 1: But the hardest one, I think, is real human oversight.
• 5:22
  Speaker 0: And that doesn't just mean a person watching a screen.
• 5:24
  Speaker 1: No, not at all.
• 5:26
  Speaker 1: It means a human with the authority and the competence to hit the stop button.
• 5:30
  Speaker 1: If an algorithm denies a loan and a human just rubber stamps it without knowing why, that's not oversight.
• 5:36
  Speaker 1: That's non-compliance.
• 5:37
  Speaker 0: And that connects to this idea of sovereignty, right?
• 5:40
  Speaker 0: For years, it was just about data location.
• 5:42
  Speaker 0: Is it in Frankfurt or Virginia?
• 5:44
  Speaker 1: Yeah, that was the old definition.
• 5:45
  Speaker 1: Yeah.
• 5:45
  Speaker 1: Rob Thomas posted something that suggests that's totally outdated.
• 5:49
  Speaker 1: He argues sovereignty has evolved from geography to operation.
• 5:52
  Speaker 1: He calls it decision authority.
• 5:55
  Speaker 0: So it's about who's the boss?
• 5:56
  Speaker 1: Basically, who owns the outcome?
• 5:58
  Speaker 1: Sovereignty in twenty twenty six means proving where the inference happened and who was in control.
• 6:03
  Speaker 1: If you use a black box model and can't audit the logic, you don't have sovereignty.
• 6:07
  Speaker 1: You just have a European server.
• 6:08
  Speaker 0: That is a critical distinction.
• 6:11
  Speaker 0: It's accountability, not residency.
• 6:13
  Speaker 0: This must be a nightmare for regulated sectors.
• 6:16
  Speaker 0: I saw Eric Van Dyke talking about MedTech.
• 6:18
  Speaker 1: MedTech is the deep end of the pool here.
• 6:20
  Speaker 1: Van Dyke says Brussels is trying to align the AI Act with the existing medical device regulations.
• 6:26
  Speaker 1: But right now, it's a double burden.
• 6:28
  Speaker 0: So for them, AI is less... innovation lab.
• 6:31
  Speaker 1: And more compliance permit.
• 6:32
  Speaker 1: Exactly.
• 6:33
  Speaker 1: But here's the paradox.
• 6:35
  Speaker 1: While the legal teams are demanding human oversight, the tech itself is evolving to operate without it.
• 6:41
  Speaker 1: We're moving from chatbots to agents.
• 6:43
  Speaker 0: This is the exciting part.
• 6:45
  Speaker 0: And we're seeing a lot of buzz from people like Nandan Malikara about the AI operating system.
• 6:49
  Speaker 1: This is a huge conceptual shift.
• 6:51
  Speaker 1: Think about the old swivel chair workflow.
• 6:53
  Speaker 0: Oh, I know it will.
• 6:54
  Speaker 0: Open chat GPT, ask it to draft an email, copy.
• 6:57
  Speaker 0: swivel to Outlook, paste, swivel to SAP, download a CSV.
• 7:01
  Speaker 1: It was exhausting and just so inefficient.
• 7:04
  Speaker 1: Malikar argues the AIOS gets rid of that.
• 7:06
  Speaker 1: The human just says, get the QFOR report and email a summary to the board.
• 7:11
  Speaker 1: And the
• 7:11
  Speaker 0: AI traverses the ERP, pulls the CRM data, does the whole thing.
• 7:15
  Speaker 1: The AI is the orchestrator.
• 7:17
  Speaker 1: The enterprise apps become the back end.
• 7:19
  Speaker 1: Malikar appointed to Claude Coat, hitting a billion dollars in ARR faster than chat GBT as a signal.
• 7:27
  Speaker 1: The market is paying for execution, not just conversation.
• 7:30
  Speaker 0: Simon Taylor had a great analogy for this.
• 7:32
  Speaker 0: He called the modern agent a co-worker, not just a tool.
• 7:36
  Speaker 0: Right.
• 7:36
  Speaker 1: He used an example with Figma and Slack.
• 7:39
  Speaker 1: One prompt.
• 7:40
  Speaker 1: draft the Slack update, visualize the new timeline in Figma, and create the tasks in Asana.
• 7:46
  Speaker 0: One prompt, three different applications.
• 7:48
  Speaker 0: It understands the workflow context.
• 7:50
  Speaker 1: Exactly.
• 7:51
  Speaker 1: And that shift from generating text to executing actions changes how you build these things.
• 7:56
  Speaker 1: BriefCashorePandy had a strong post on this.
• 7:58
  Speaker 1: He says, twenty twenty six is all about building reliable systems.
• 8:02
  Speaker 0: So my library of perfect prompts isn't enough anymore.
• 8:04
  Speaker 1: It's not useless, but it's not enough.
• 8:07
  Speaker 1: Bandy says prompt engineering is being replaced by system engineering.
• 8:11
  Speaker 1: The stack has an orchestration layer, a memory layer, and crucially an evaluation layer.
• 8:16
  Speaker 0: Because you can't just eyeball it anymore, you need automated testing if an agent is sending emails on its own.
• 8:21
  Speaker 1: If you're just tweaking prompts in a playground in twenty twenty six.
• 8:24
  Speaker 1: You're already behind.
• 8:25
  Speaker 0: Okay, but where's the money?
• 8:26
  Speaker 0: Is this just cool tech or does it pay off?
• 8:29
  Speaker 0: I saw John Godel's breakdown for chief revenue officers.
• 8:31
  Speaker 0: Yes,
• 8:32
  Speaker 1: Godel frames this as an operating leverage tool.
• 8:35
  Speaker 0: Operating levers?
• 8:36
  Speaker 0: That's music to a CFO's ears.
• 8:38
  Speaker 1: It is.
• 8:39
  Speaker 1: Think about a deal desk.
• 8:40
  Speaker 1: Usually a huge bottleneck.
• 8:42
  Speaker 1: An agent can auto build that approval packet in seconds.
• 8:46
  Speaker 1: It reduces the cycle time of deals.
• 8:48
  Speaker 0: And if you close a deal three days faster, that's direct revenue impact.
• 8:51
  Speaker 0: That's cash flow.
• 8:52
  Speaker 0: Exactly.
• 8:53
  Speaker 1: But... And there's always a but.
• 8:55
  Speaker 0: There always is.
• 8:56
  Speaker 1: Is it actually working?
• 8:57
  Speaker 1: This brings us to the reality check part of this.
• 9:00
  Speaker 0: Let's play Devil's Advocate.
• 9:01
  Speaker 0: Hina Purohit posted about the productivity paradox.
• 9:04
  Speaker 0: The numbers she cited were
• 9:06
  Speaker 1: weird.
• 9:07
  Speaker 1: They're stark.
• 9:08
  Speaker 1: She cites data showing seventy-four percent of teams report productivity gains from AI.
• 9:13
  Speaker 1: They feel faster, but only eleven percent report measurable value.
• 9:18
  Speaker 0: Wait, what?
• 9:18
  Speaker 0: So people are churning out more stuff, more code, more slides, but it's not moving the needle on the business.
• 9:24
  Speaker 1: We might just be generating noise faster.
• 9:27
  Speaker 1: And Werner Heistek pointed out something even crazier about software development.
• 9:30
  Speaker 1: The
• 9:30
  Speaker 0: coding assistants?
• 9:31
  Speaker 0: I love those.
• 9:32
  Speaker 0: People
• 9:32
  Speaker 1: do, but Heistek points to cases where AI coding caused a nineteen percent slowdown.
• 9:39
  Speaker 0: A slowdown?
• 9:40
  Speaker 0: How is that even possible?
• 9:42
  Speaker 1: It writes the code instantly, but is it the right code?
• 9:45
  Speaker 1: The slowdown is the verify tax.
• 9:47
  Speaker 1: The human developer has to spend hours debugging and fixing the messy code the AI generated.
• 9:53
  Speaker 0: I feel that.
• 9:54
  Speaker 0: It's like being an editor instead of a writer.
• 9:55
  Speaker 0: Sometimes editing takes way longer.
• 9:57
  Speaker 0: So how do we fix
• 9:58
  Speaker 1: this?
• 9:59
  Speaker 1: Aman Ruiz suggests it's an organizational design problem.
• 10:02
  Speaker 1: He talks about the AI native squad.
• 10:04
  Speaker 0: Squad, not just one AI wizard on the team.
• 10:06
  Speaker 1: Exactly.
• 10:07
  Speaker 1: Ruiz argues that having one wizard doesn't help if everyone else is stuck in twenty twenty two.
• 10:11
  Speaker 1: The friction kills the speed.
• 10:13
  Speaker 1: The whole squad needs to level up together.
• 10:15
  Speaker 0: That makes total sense.
• 10:16
  Speaker 0: Culture eats AI for lunch, apparently.
• 10:18
  Speaker 1: Does.
• 10:19
  Speaker 1: And one last thing before we wrap.
• 10:21
  Speaker 1: The huge looming risk with these agents.
• 10:24
  Speaker 1: Security.
• 10:25
  Speaker 1: Specifically, prompt injection.
• 10:27
  Speaker 0: Bernard Maher flagged this.
• 10:28
  Speaker 0: And for a long time, I just thought that was a prank.
• 10:30
  Speaker 0: Tricking a chatbot into saying something
• 10:32
  Speaker 1: similar.
• 10:32
  Speaker 1: When it's just a chatbot, it is a prank.
• 10:35
  Speaker 1: But now connect it to an agent that can access your ERP, your bank account.
• 10:39
  Speaker 0: Right.
• 10:40
  Speaker 0: If I can trick the agent, I'm not getting a poem.
• 10:43
  Speaker 0: I'm transferring money or deleting the customer database.
• 10:46
  Speaker 1: Prompt injection becomes a critical vulnerability.
• 10:49
  Speaker 1: It's not a toy anymore.
• 10:50
  Speaker 1: It's a security breach waiting to happen.
• 10:52
  Speaker 0: So if we pull all this together... the wow phase is dead.
• 10:57
  Speaker 0: We need strict governance.
• 10:58
  Speaker 0: We need solid architecture.
• 11:00
  Speaker 0: And we're shifting to agents that can do real work, but that comes with a whole new set of risks.
• 11:06
  Speaker 1: And we have to do it all while navigating a productivity paradox where faster doesn't always mean better.
• 11:11
  Speaker 0: It sounds like twenty twenty six is the year of the grown-up conversation about AI.
• 11:16
  Speaker 1: I think that is the perfect way to put it.
• 11:17
  Speaker 1: Okay,
• 11:17
  Speaker 0: let's wrap this up.
• 11:19
  Speaker 0: If you enjoyed this episode, new episodes drop every two weeks.
• 11:22
  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
• 11:30
  Speaker 1: tech.
• 11:31
  Speaker 1: And if I can leave you with one final thought, we are moving from the wow phase to the how phase.
• 11:37
  Speaker 1: The winners in twenty twenty six won't be the ones with the best models.
• 11:41
  Speaker 1: The winners will be the ones with the best architecture and governance to run them safely.
• 11:45
  Speaker 0: Thanks for joining us on the deep dive.
• 11:46
  Speaker 0: Don't forget to subscribe.
• 11:48
  Speaker 0: See you next time.