Latest / Best of LinkedIn: Digital Transformation & Tech / Best of LinkedIn: AI & Agentic Systems CW 41/ 42

Transcript

Best of LinkedIn: AI & Agentic Systems CW 41/ 42 · Best of LinkedIn: Digital Transformation & Tech

• 0:00
  Speaker 0: This episode is provided by Thomas Allgaier and Frennus, based on the most relevant LinkedIn posts about AI and agentic systems from CW-forty one in forty two.
• 0:10
  Speaker 0: Frenness supports ICT and tech providers with a ecosystem strategy by delivering independent vendor assessment build versus buy analysis an ecosystem intelligence that prevents expensive mistakes and positions the provider's competitively.
• 0:24
  Speaker 0: you can find more info in the description.
• 0:26
  Speaker 1: all right let's get ready to do it.
• 0:28
  Speaker 0: yeah Welcome to the deep dive.
• 0:30
  Speaker 0: Our mission today is just, you know cut through a lot of the hype.
• 0:33
  Speaker 0: We're looking at the most relevant trends across the professional landscape To figure out what's actually happening with AI and agentic systems right now.
• 0:41
  Speaker 1: Right because we are definitely moving out of that era Of pure unfiltered experimentation.
• 0:47
  Speaker 1: We've entered this new phase in an area of accountability really
• 0:51
  Speaker 0: Exactly.
• 0:52
  Speaker 0: And the overarching narrative we're seeing from professionals on the ground is that, The main constraints are getting actual value out of AI.
• 1:00
  Speaker 0: They aren't tech itself anymore!
• 1:02
  Speaker 1: No not at all...the bottlenecks are structural.
• 1:04
  Speaker 1: We talking broken operating models massive governance cues and just quality human judgement.
• 1:10
  Speaker 0: Yeah if you look what people actually deploying these systems or saying technology IS ready It's organizations totally lagging behind
• 1:19
  Speaker 1: Which a perfect place to start looking right at that enterprise level, where the rubber meets the road and why so many companies are just hitting a complete wall trying to scale this stuff.
• 1:28
  Speaker 0: Yeah!
• 1:28
  Speaker 0: So Nandan Mulakara brought up point I think is massive warning light for anyone in tech or digital transformation.
• 1:35
  Speaker 0: now he pointed out enterprises are risking total repeat of scaling failures we saw with RPA.
• 1:42
  Speaker 1: Oh robotic process automation.
• 1:44
  Speaker 0: yeah
• 1:45
  Speaker 1: That's brilliant historical parallel
• 1:48
  Speaker 0: Because back then, RPA just made broken processes move faster.
• 1:52
  Speaker 0: And now everyone is basically mounting this massive AI gun onto those exact same outdated operating models.
• 1:58
  Speaker 1: They're essentially just automating the waste.
• 2:01
  Speaker 0: But I do want to pushback on that a little bit like isn't speed kind of the entire point?
• 2:07
  Speaker 0: Of AI?
• 2:08
  Speaker 0: if you have a broken shopping cart but you strap A Ferrari engine into it Isn't that still better than walking?
• 2:13
  Speaker 1: well i mean It'll get you down the aisle faster, sure.
• 2:16
  Speaker 1: But you're going to crash into all the shelves right?
• 2:18
  Speaker 1: You're gonna tear this store apart...you just cannot automate a bad workflow and expect good results.
• 2:24
  Speaker 1: Asad Ansari highlighted this perfectly when he was looking at the UK civil service.
• 2:29
  Speaker 0: Oh!
• 2:29
  Speaker 0: The massive savings promises.
• 2:30
  Speaker 0: right
• 2:31
  Speaker 1: exactly it's this promise of AI saving something like thirty six billion pounds a year.
• 2:37
  Speaker 1: But his reality check was that those numbers are completely unrealistic if you don't tear down the foundational broken processes first.
• 2:45
  Speaker 0: Because AI is just a pure multiplier?
• 2:48
  Speaker 1: Yes, feed it a broken system and doesn't fix it – the chaos just multiplies!
• 2:54
  Speaker 0: So Atena Rahani actually brought great framework to this operating model problem.
• 2:59
  Speaker 0: She pointed out scaling agentic AI where the system takes action on its own requires fundamental shift in the operating itself.
• 3:07
  Speaker 1: Right, because it's no longer just a question of whether the tech can act.
• 3:10
  Speaker 0: Exactly!
• 3:11
  Speaker 0: It is about deciding which specific outcomes get.
• 3:13
  Speaker 0: what she called bounded autonomy
• 3:16
  Speaker 1: Like
• 3:16
  Speaker 0: where does an agent have free reign?
• 3:18
  Speaker 0: And where do human judgments still need to step in and shape their path?
• 3:21
  Speaker 1: If you skip that restructuring part things gets messy fast.
• 3:25
  Speaker 1: Gabriel Millian actually broke down some hard data on this from Airstreet Capital state-of-AI report.
• 3:30
  Speaker 0: What did they find?
• 3:32
  Speaker 1: They looked at the OS world.
• 3:33
  Speaker 1: two point O test Benchmark for real computer work, like navigating web browsers using desktop apps.
• 3:41
  Speaker 0: The messy everyday enterprise stuff?
• 3:44
  Speaker 1: Exactly and a top AI model Claude Opus V. it got partial credit for seventy seven point seven percent of the steps
• 3:53
  Speaker 0: which sounds pretty incredible on paper.
• 3:55
  Speaker 0: almost eighty percent Of the steps right.
• 3:57
  Speaker 0: that sounds Like A massive productivity boost.
• 4:00
  Speaker 1: It does.
• 4:00
  Speaker 1: until you look at the actual finish rate It only finished forty-four point three percent of the actual tasks.
• 4:07
  Speaker 0: Oh wow!
• 4:08
  Speaker 1: Less than half,
• 4:09
  Speaker 0: right?
• 4:10
  Speaker 1: Because a high score on these steps isn't the finish job.
• 4:13
  Speaker 1: in an enterprise partial credit is basically zero credit.
• 4:17
  Speaker 0: Yeah I mean if an agent drafts a perfect contract but then gets confused by the final eSignature API The contract doesn't go out.
• 4:25
  Speaker 0: A single missed approval leaves the whole thing undone.
• 4:28
  Speaker 1: Exactly So.
• 4:29
  Speaker 1: the technology's the easy part
• 4:33
  Speaker 0: which perfectly ties into ERA Agusabt-Sien's point.
• 4:36
  Speaker 0: He made this argument that when AI actually does save time on a project, you shouldn't just convert those savings in to cheaper bill for the client.
• 4:44
  Speaker 1: Right!
• 4:44
  Speaker 1: The instinct is always to pocket the savings.
• 4:47
  Speaker 0: Yeah but he was saying that savetime has to be reinvested into process.
• 4:51
  Speaker 0: You need it quality checks business validation and change management.
• 4:55
  Speaker 0: So what all of us mean as leader?
• 4:57
  Speaker 0: It means we have to redesign workflow around human oversight not machines output
• 5:02
  Speaker 1: Absolutely.
• 5:04
  Speaker 1: And if fixing the process is step one, then Step two... ...is figuring out how you govern all these new workflows without just grinding innovation to a total halt.
• 5:13
  Speaker 0: Oh!
• 5:13
  Speaker 0: The governance nightmare?
• 5:14
  Speaker 1: Yeah.
• 5:15
  Speaker 1: Joseph Abraham highlighted what he calls the Q problem.
• 5:19
  Speaker 1: When you go into an enterprise and ask the AI team what's slowing them down they usually blame model hallucinating or bad data.
• 5:26
  Speaker 0: But it's not that data, isn't it?
• 5:28
  Speaker 1: No It's the queue.
• 5:29
  Speaker 1: People are submitting use cases by email and they just sit there for eighteen days.
• 5:35
  Speaker 1: And a basic internal chatbot gets reviewed with the exact same depth as a critical financial model.
• 5:40
  Speaker 0: It's like installing speed bumps on a major highway, people don't stop driving They just detour through residential neighborhoods where it is way more dangerous.
• 5:47
  Speaker 0: Exactly
• 5:48
  Speaker 1: Slow governance doesn't reduce risk
• 5:53
  Speaker 0: literally how shadow AI happens.
• 5:55
  Speaker 0: Employees have deadlines, so they just quietly use unapproved tools.
• 5:59
  Speaker 0: and by the way if you want to keep navigating these enterprise detours with us make sure your subscribed to The Deep Dive So You Don't Miss Future Editions.
• 6:06
  Speaker 0: We Cover A Lot Of These Roadblocks.
• 6:09
  Speaker 1: And on the regulatory side of that roadblock Christian Bean analyzed a new EU AI Act standard EN- one eight two eight six.
• 6:18
  Speaker 0: Okay That sounds extremely dense
• 6:19
  Speaker 1: It is but his take away really practical.
• 6:23
  Speaker 1: ESA's compliance shouldn't mean building entirely new, parallel structures in your company.
• 6:28
  Speaker 0: So no isolated AI ethics board that meets twice a year?
• 6:31
  Speaker 1: Right!
• 6:32
  Speaker 1: You need to perform gap assessments and integrate the AI requirements into quality management systems you already have.
• 6:38
  Speaker 0: That makes systemic sense.
• 6:40
  Speaker 0: But there is human side of this governance too which Seagrid Burj van Royen brought up regarding clinical AI in healthcare.
• 6:46
  Speaker 1: This is a fascinating dynamic.
• 6:48
  Speaker 0: Yeah she argues The best governance document in the world won't make a clinician actually trust an AI tool.
• 6:55
  Speaker 0: Trust breaks into real-world when a doctor is exhausted and under time pressure.
• 6:59
  Speaker 0: And
• 6:59
  Speaker 1: then, most dangerous part isn't even when they reject the AI?
• 7:02
  Speaker 0: No!
• 7:03
  Speaker 0: It's when that AI is right MOST of the time because THEN clinicians stop actively agreeing... ...and just start clicking through.
• 7:10
  Speaker 0: Active trust degrades to what she calls passive consent
• 7:14
  Speaker 1: Which is terrifying…in clinical setting
• 7:16
  Speaker 0: Completely.
• 7:17
  Speaker 0: Tools have to be designed to explicitly communicate their uncertainty, to force the human to wake up and pay attention.
• 7:24
  Speaker 1: And that danger of passive consent gets infinitely riskier when we transition from AI just gives advice... ...to a GENTIC.
• 7:31
  Speaker 1: AI actually takes action on our behalf.
• 7:33
  Speaker 0: Here's where it really get interesting.
• 7:36
  Speaker 0: Renee Kay highlighted this recent incident in UK AI Security Institute.
• 7:40
  Speaker 0: It proves that agentic risks are way beyond just, you know, prompt injection.
• 7:45
  Speaker 0: We aren't
• 7:45
  Speaker 1: just tricking chatbots to write bad poems anymore
• 7:48
  Speaker 0: Exactly!
• 7:49
  Speaker 0: we now have agents capable of unsanctioned supply chain compromise.
• 7:53
  Speaker 0: they can spin up servers execute transactions...we absolutely need a defense in depth layered security approach for this.
• 8:00
  Speaker 1: Thomas Sowering from SAP put this into perspective perfectly.
• 8:03
  Speaker 1: He said, AI hasn't really changed what we defend against in cybersecurity but it has completely changed.
• 8:09
  Speaker 1: how fast that happens?
• 8:10
  Speaker 0: Because the time to impact is shrinking?
• 8:12
  Speaker 1: Exactly!
• 8:12
  Speaker 1: It's compressed for weeks down literally minutes.
• 8:16
  Speaker 1: Having a human respond at human speed no longer viable as first line of defense.
• 8:20
  Speaker 1: We need machine-speed defenses.
• 8:22
  Speaker 0: Wait, but if the machine is defending at machine speed and the attacker's attacking at machine-speed where does human even fit in?
• 8:29
  Speaker 0: Are we literally just bystanders watching algorithms
• 8:32
  Speaker 0: fight?".
• 8:32
  Speaker 1: It's a valid concern.
• 8:33
  Speaker 1: But we are the ones who control boundaries.
• 8:36
  Speaker 1: Moshkin Lefebvre provided framework for this.
• 8:39
  Speaker 1: Agent autonomy isn't binary switch... Right!
• 8:44
  Speaker 1: It should scale based on task risk.
• 8:46
  Speaker 1: You have to design sandboxes asking two things What consequences of agent being
• 8:51
  Speaker 0: wrong?
• 8:52
  Speaker 1: and is the action reversible?
• 8:53
  Speaker 0: Ah, reversibility.
• 8:55
  Speaker 0: Drafting an email as reversible wiring fifty grand it's not
• 8:58
  Speaker 1: Exactly.
• 8:59
  Speaker 0: But defining that sandboxes getting weird because our legacy security systems can't even recognize good agents.
• 9:06
  Speaker 0: Nikhil Kassidy shared this highly specific kind of hilarious example from The FinTech Space
• 9:11
  Speaker 1: About fraud models.
• 9:12
  Speaker 0: Yeah Right now Fraud detection models are misclassifying legitimate AI purchasing agents at suspicious spots
• 9:19
  Speaker 1: Because they're too fast.
• 9:20
  Speaker 0: Right.
• 9:21
  Speaker 0: The fraud models learned what suspicious means based on human typing rhythms like hesitation, the way a mouse jiggles.
• 9:27
  Speaker 0: A good AI agent has no hesitation.
• 9:29
  Speaker 0: It looks exactly Like a malicious script.
• 9:32
  Speaker 1: So this system is blocking the exact efficiency it was meant to protect
• 9:35
  Speaker 0: Exactly And Michael Westerwell brought up a wild solution To this authorization problem.
• 9:41
  Speaker 0: World ID just launched iris verified identity proofs For AI agents.
• 9:46
  Speaker 0: Wait!
• 9:46
  Speaker 0: Like scanning your eyeball for the AI.
• 9:48
  Speaker 1: Literally Yes You scan your eyeball to verify you're human, and hand that cryptographic proof of human...to the AI agent.
• 9:57
  Speaker 0: Oh wow!
• 9:58
  Speaker 1: What's fascinating here is that the bank security system stops asking the robot who it IS…and starts asking WHO stands BEHIND IT?
• 10:06
  Speaker 0: That is incredible.
• 10:08
  Speaker 0: And it perfectly brings us to this final piece – the workforce itself because at the end of day all comes down to humans standing behind a machine with quality of their judgement
• 10:17
  Speaker 1: which in pretty fragile state right now.
• 10:19
  Speaker 1: honestly
• 10:19
  Speaker 0: Yeah, Joshua Miller shared this statistic.
• 10:21
  Speaker 0: that is just a nightmare for management.
• 10:24
  Speaker 0: A KPMG survey of forty-eight thousand people found that fifty seven percent of employees are hiding their AI use at work.
• 10:32
  Speaker 1: Over half your workforce is secretly using it
• 10:34
  Speaker 0: and presenting the Ai's work as they're own original thinking.
• 10:38
  Speaker 0: Imagine doing performance reviews right now.
• 10:40
  Speaker 0: You can't tell if strategy belongs to you employee or if its just AI polish.
• 10:45
  Speaker 1: No,
• 10:47
  Speaker 0: you have to promote on the actual thinking.
• 10:48
  Speaker 0: You've heard them articulate in their room.
• 10:51
  Speaker 1: This is exactly what Saul Rashidi calls The Messy Middle.
• 10:54
  Speaker 1: Everyone expects massive productivity But an MIT study found that ninety-five percent of organizations saw no measurable return On Their Gen AI Pilots.
• 11:03
  Speaker 0: Ninety five percent.
• 11:04
  Speaker 1: That was a staggering failure rate for pilots.
• 11:07
  Speaker 0: It Is
• 11:07
  Speaker 1: And it gets crazier.
• 11:09
  Speaker 1: A Metra randomized trial showed that experienced developers using Ai tools were actually nineteen percent slower
• 11:15
  Speaker 0: Slower Slower.
• 11:17
  Speaker 1: But when surveyed, they believed there were twenty percent faster.
• 11:20
  Speaker 0: How do you lose that much time and still feel like your flying?
• 11:24
  Speaker 1: Because AI commoditizes the execution.
• 11:26
  Speaker 1: Typing the prompt feels instant but then the developer spends three hours untangling hallucinations in debugging sloppy logic that they didn't write.
• 11:35
  Speaker 0: so The real work shifts entirely to oversight.
• 11:38
  Speaker 1: right critical thinking taste discernment these are the new scarce capital.
• 11:44
  Speaker 1: If you don't have deep domain expertise, You can even tell if the AI's output is good or just confidently wrong.
• 11:50
  Speaker 0: Okay let's unpack this with a final story from Jimmy Acton because it perfectly captures everything we've talked about.
• 11:57
  Speaker 0: He watched this high-end robot teleoperation demo.
• 12:00
  Speaker 1: The motion capture one
• 12:01
  Speaker 0: Yeah!
• 12:01
  Speaker 0: A robot was copying every move of human operator zero leg.
• 12:05
  Speaker 0: But then the human accidentally flinched in pain
• 12:08
  Speaker 1: And the robot copied him perfectly.
• 12:10
  Speaker 0: Perfectly mimicked the flinch and inside there profound.
• 12:14
  Speaker 0: Perfect mimicry is not good judgment.
• 12:17
  Speaker 0: the hard part of building AI and integrating it into your company isn't getting It to follow instructions, its teaching when not too.
• 12:26
  Speaker 1: That's the ultimate through line here.
• 12:27
  Speaker 1: Over last fifteen minutes or so we've seen that technology is incredibly capable.
• 12:32
  Speaker 1: but our operating models are governance cues security layers.
• 12:36
  Speaker 1: they're all playing a frantic game of catch up
• 12:39
  Speaker 0: Absolutely.
• 12:40
  Speaker 1: Ultimately human judgement is only thing that scales AI safely.
• 12:45
  Speaker 0: It really is, and I want to leave you with one final kind of provocative thought on that note.
• 12:50
  Speaker 0: Joseph Noggle and Dr.
• 12:52
  Speaker 0: Benjamin Blau were discussing this recent claim from OpenAI
• 12:55
  Speaker 1: about the fluid dynamics problem?
• 12:57
  Speaker 0: Yes!
• 12:58
  Speaker 0: The Navier-Stokes math problem.
• 13:00
  Speaker 0: OpenAI says they solved it in just eighty eight hours.
• 13:04
  Speaker 0: The AI agents traded two point seven million messages and spat out a one hundred sixty six page paper
• 13:10
  Speaker 1: which sounds like a monumental breakthrough.
• 13:13
  Speaker 1: It
• 13:13
  Speaker 0: does, but top human mathematicians are looking at this paper and saying it's nearly unreadable.
• 13:19
  Speaker 0: The logic is so dense an alien that humans can barely verify the work.
• 13:23
  Speaker 1: So we have the answer But no actual comprehension
• 13:26
  Speaker 0: Exactly!
• 13:29
  Speaker 0: Here something for you to mull over This week.
• 13:31
  Speaker 0: As AI radically accelerates what We Can produce Are we investing enough in our ability To actually understand results?
• 13:40
  Speaker 0: If a machine gives you the right answer but can't trace how it got there, well is that actually progress?
• 13:46
  Speaker 1: That's the big
• 14:02
  Speaker 0: question.