Best of LinkedIn: ICT & Tech Insights CW 17/ 18

00:00:00: This episode is provided by Thomas Allgeier and Frennis, based on the most relevant LinkedIn posts about ICT and tech insights from CW-Seventeen and Eighteen.

00:00:09: Frennes supports ICT enterprises in the form of delivering precise ICT market and pricing intelligence that analyst subscriptions and existing databases cannot provide.

00:00:19: you can find more info in the description.

00:00:21: so uh keeping that in mind let's establish exactly what we are doing today.

00:00:27: Over the past two weeks just a massive wave of insights has flooded professional networks regarding, you know The actual state-of-tech infrastructure.

00:00:36: We've synthesized the critical signals from all that noise.

00:00:39: right

00:00:39: because well imagine You've just spent millions building a fleet of cutting edge supersonic jets

00:00:45: okay?

00:00:45: I'm picturing it.

00:00:46: The engineering is flawless.

00:00:47: You roll them out onto the tarmac, you fire up the engines and suddenly realize that you never paved the runway.

00:00:52: Oh wow!

00:00:52: Yeah...you're trying to launch a Mach-III aircraft from a dirt road.

00:00:56: And that scenario's exactly what global enterprises are facing right now with their technology strategies.

00:01:01: I mean..that's a phenomenal way of visualizing this shift we're tracking here.

00:01:06: For last two years the tech market was just Consumed by this hype and experimentation phase.

00:01:11: Oh, totally just playing around

00:01:13: right?

00:01:13: we're all marveling at what generative AI models could do inside controlled sandboxes But reading through the insights were analyzing.

00:01:20: today for this deep dive that honeymoon phase is officially over.

00:01:24: it's done.

00:01:25: Yeah We are now entering the era of structured enterprise execution

00:01:29: And Execution is messy.

00:01:32: Today, we're guiding you through the four interconnected tectonic shifts that are dominating this reality check

00:01:38: which our AI execution The evolution of cyber resilience the impending arrival of quantum readiness and finally digital sovereignty.

00:01:47: right

00:01:47: And we really need to start with AI execution because the terminology is shifting like right under our feet.

00:01:52: We're no longer just talking about building AI agents, we were suddenly talking about governing them!

00:01:56: We

00:01:57: are I mean... The architectural conversation has matured incredibly fast.

00:02:01: Look at Rohit Sharma's recent analysis of Google Cloud Next.

00:02:04: Oh yeah that was a sharp observation.

00:02:06: It really was.

00:02:07: He pointed out last year every keynote was about the Agents themselves Like what they could write or code or generate Just

00:02:15: the flashy stuff Exactly.

00:02:17: But this year, the spotlight moved entirely off of agents and onto layers wrapping around them.

00:02:22: Let's actually break down what those layers are because it can sound a bit abstract for, you know some folks.

00:02:28: Shauna highlighted the Gemini agent platform.

00:02:31: It is no longer just a model and now ships with registries to index your agents

00:02:36: Which is huge

00:02:37: Yeah!

00:02:37: And cryptographic identities so that they trace what they do and centralized policy enforcement.

00:02:43: Vasu Jakal shared exact same signal from Microsoft Cam.

00:02:46: Yes,

00:02:47: with Microsoft Agent-Agent.

00:02:51: And the reason for this sudden focus on governance is, well honestly it's fear.

00:02:55: IT leaders are absolutely terrified of shadow AI.

00:02:58: Right and in same way we used to talk about Shadow IT where employees were you know using unapproved software On their laptop.

00:03:04: The

00:03:04: shadow AI is much faster

00:03:06: Oh!

00:03:06: Way Faster.

00:03:07: You have autonomous agents spinning up across endpoints Cloud environments Internal apps.

00:03:13: An employee might deploy an agent To scrape financial data And IT has zero visibility into It.

00:03:19: None.

00:03:19: That's why organizations are demanding a unified control

00:03:22: plane.

00:03:23: Yeah, Control Plane gives you that centralized dashboard to observe throttle or just kill an agent access entirely.

00:03:30: But here is where we hit your supersonic jet on a dirt road problem.

00:03:34: Ah the unpaved runway.

00:03:36: Exactly

00:03:37: all of this.

00:03:37: sophisticated multi-agent governance Is effectively useless if The underlying data foundation is fractured.

00:03:44: That is the massive bottleneck right

00:03:45: now.

00:03:46: I want to pull in a real-world example of this, actually because Aditya Ghoul shared brilliant observation on the reality enterprise data.

00:03:52: he was looking at recent integration between SAP and Google Cloud

00:03:56: specifically their multi agent AI for customer experience or CX

00:03:59: Right And On paper The tech is just breathtaking.

00:04:02: you have agents coordinating bidirectionally across sales platforms and marketing clouds In Real Time.

00:04:07: But Ghoul points out the fail flaw in large BTC implementations.

00:04:13: If an enterprise has duplicated customer records or, you know their transaction histories are siloed in legacy systems the multi-agent AI doesn't magically fix that.

00:04:23: It just executes on the broken data.

00:04:25: it Just makes faster more confident bad decisions.

00:04:28: Exactly if you deploy and autonomous agent on top of a messy database?

00:04:33: It might start firing off aggressive retention discounts to your most loyal customers Or I don't know sending onboarding emails to people who churn three years ago

00:04:41: which is a disaster.

00:04:43: Goal's warning is really clear.

00:04:45: You have to fix your data ownership and cleanliness before you hand the keys over to multi-agent workflows,

00:04:50: right?

00:04:50: And we have to be careful not to view this as just a simple database cleanup exercise like oh let's just deduplicate some rows

00:04:56: definitely not.

00:04:57: Mark O'Gar presented a framework at Deda Gokhan that elevates This to a strategic level.

00:05:03: he argues The technology alone cannot bridge the gap between messy Data an AI execution.

00:05:08: He mentioned five levels didn't they?

00:05:09: yeah you actually had To move through five distinct levels of quality.

00:05:12: Okay, walk me through that because usually companies stop at data quality and just call it a day.

00:05:17: Exactly!

00:05:17: And that's why they fail.

00:05:19: Data Quality is the baseline prerequisite like...is the data accurate?

00:05:23: Goyer says you then have to graduate into context-quality than information-quality or knowledge-quality until we finally reach business decision quality.

00:05:32: Wow okay so your transforming data governance into decision government

00:05:36: exactly.

00:05:36: But practically speaking, if an autonomous agent is making a thousand decisions a second to personalize a web experience the speed at which we process that govern data has to radically change.

00:05:48: It really does.

00:05:49: legacy batch processing where you update your database every night at two AM That cannot support and edge in acting in real time.

00:05:56: it absolutely can't And Kai wainer shared a masterclass on how to solve this.

00:06:00: looking at Wix dot com Wix powers around seven percent of the entire internet Which is wild to think about It is.

00:06:06: To deliver AI personalization at that scale, they had to completely abandon traditional batch processing.

00:06:12: They rebuilt their entire machine learning feature store using Apache Kafka and Apache Flink.

00:06:17: Okay for the business leaders listening who might not be deep in data engineering How does that actually differ from the old way of doing things?

00:06:25: Well think of legacy batch processing what we usually call ETL extract transform load.

00:06:31: it's like reading yesterdays newspaper to make stock trades today.

00:06:35: You're always a step behind!

00:06:36: Exactly, you are always reacting to historical states.

00:06:40: what Wix built with Kafka and Flink is the equivalent of standing right on the trading floor.

00:06:44: Kafka ingests the raw events as they happen.

00:06:47: we're talking over seventy billion events a day across their network.

00:06:51: Wait,

00:06:51: seventy billion?

00:06:53: That scale is hard to even comprehend.

00:06:54: It IS staggering.

00:06:56: So Kafka streams those events continuously And Flink processes it aggregates them.

00:07:01: This architecture serves up fresh data to the AI models with millisecond latency.

00:07:07: Yeah,

00:07:08: the core takeaway here is that if you want to deploy real-time AI your underlying architecture has become streaming first.

00:07:14: Which honestly brings us to a rather terrifying realization and natural pivot of our second major theme for you, cyber resilience.

00:07:22: because if Wix is processing seventy billion events to personalize websites at millisecond latency well threat actors are using the exact same speed in autonomous reasoning to probe enterprise networks.

00:07:35: that's the chilling reality.

00:07:37: AI isn't just accelerating business workflows, it has completely dropped the barrier to executing highly sophisticated cyber attacks.

00:07:45: Chitanya Gojini highlighted a situation that really puts this into perspective.

00:07:49: It involves an unreleased model from Anthropic internally known as Claude Mythos.

00:07:54: I read about this.

00:07:55: Yeah apparently This model demonstrated such a profound leap in autonomous reasoning That they deemed too dangerous for release.

00:08:03: The capabilities were unprecedented complex, unfamiliar systems without any human prompting.

00:08:10: None at all!

00:08:10: It was finding systemic cybervulnerability so severe that researchers feared it could threaten internet-wide integrity if weaponized...

00:08:17: Then when you pair with insights from Jay Chaudry who by the way has spent thirty years in cybersecurity The picture gets really stark.

00:08:25: Yeah he warned about this scale.

00:08:26: He warned because of AI accelerated discovery.

00:08:29: we can easily see a twenty fold spike in software vulnerabilities Not a twenty percent increase A twenty-fold spike.

00:08:36: Traditional security architectures simply break under that kind of mathematical pressure.

00:08:41: Francis Odom provided an excellent breakdown why the standard zero trust model is currently failing in the agentic era.

00:08:48: Wait, let me stop you there because for last five years Zero Trust has been ultimate buzzword.

00:08:53: Every single vendor sells it.

00:08:55: Why does this suddenly fail?

00:08:57: because the logic of zero trust was built for humans.

00:09:01: Think about it.

00:09:01: in a traditional setup, A human employee logs-in you authenticate their identity to evaluate their predictable behavior and you grant them access to a segment of the network.

00:09:11: but today nonhuman identities AI agents API keys automated scripts they drastically outnumber people across cloud environments

00:09:21: And an AI agent doesn't act like a human.

00:09:22: It might spin up, chain three different cloud tools together, execute a complex data query and completely disappear in three

00:09:28: seconds.".

00:09:29: Precisely!

00:09:30: Traditional controls don't even register that activity fast enough.

00:09:34: Odin points out we have to evolve from zero trust... ...to JIT Trust just-in-time trust.

00:09:41: Instead of granting a block access, JITTrust is continuous behavior based evaluation.

00:09:47: The Agent has to earn trust for every single microaction And the moment its behavior drifts from the expected baseline, that trust is instantly revoked.

00:09:56: So if I am a CISO listening to this deep dive right now what does the actionable move?

00:10:00: Do i just need go on board and ask for massive budget increase by bigger stack of AI powered security tools?

00:10:07: That's the most dangerous trap an enterprise can fall into.

00:10:11: Yeah, Dr.

00:10:11: HP articulated this beautifully.

00:10:13: fixing cyber risk in this new era is primarily a leadership and organizational challenge not A technical tool gap.

00:10:19: interesting

00:10:20: organizations constantly try to buy their way out of risk with new software But tools do not solve executive accountability

00:10:26: And Jane Franklin echoed that exact sentiment.

00:10:28: regarding board level conversations She noted that boards are still asking outdated questions like Are we compliant or Is our insurance policy up today?

00:10:39: Right Totally missing the point.

00:10:40: She argues, The only question that matters now is about business survivability

00:10:44: and she breaks survivability down into a really fascinating framework of five layers leadership culture governance defense And collaboration

00:10:54: notice what?

00:10:55: Is missing from that list?

00:10:56: yeah Only one of those layers defense is primarily technical.

00:11:00: Four of them are profoundly human, yet corporate boards spend ninety percent their time inspecting the technical layer assuming that culture and leadership will just sort themselves out.

00:11:11: But when a massive breach happens it is always the human layers who fail first Always.

00:11:16: We do need to highlight that AI isn't purely an offensive threat though we're seeing incredible defensive applications moving into execution at scale.

00:11:24: Johanna Haydn shared the twelve-month results of Vodafone Spam Warner, which is a brilliant use case.

00:11:29: I love this example because it solves the problem invisibly.

00:11:33: In just one year, SpamWarner successfully blocked sixty four point five million AI driven fraud call attempts.

00:11:40: But The Genius Of It Is The Architecture!

00:11:43: It operates entirely natively on the network side Exactly...the

00:11:47: end user doesn't have to download a clunky third party app or update their settings, or jump through authentication hoops.

00:11:54: The infrastructure simply protects itself

00:11:56: and capital markets are signaling that this kind of infrastructure resilience is the next gold rush.

00:12:01: I mean service now recently announced a planned seven point seventy five billion dollar acquisition of armist.

00:12:07: That's a massive move

00:12:08: right specifically to bring that level of critical infrastructure visibility To the enterprise scale.

00:12:13: it makes total financial sense yeah But here is where we need to connect the dots to our third theme quantum readiness.

00:12:20: Because while AI is stress testing our defenses today, quantum computing is the impending tectonic shift that will fundamentally rewrite the rules of infrastructure tomorrow.

00:12:29: This a topic always feels like it belongs in a sci-fi movie.

00:12:31: I mean quantum has been five years away.

00:12:33: for last twenty years It

00:12:34: really does.

00:12:35: But if you look at research shared by Jan Michalon and Dr.

00:12:38: Cara Antoine based on recent McKinsey report The commercial tipping point no longer theoretical...it's happening now.

00:12:46: The numbers back that up.

00:12:47: They track over three hundred companies actively operating in the quantum space, moving aggressively past academic research into deployable products.

00:13:00: And Honeypink and Jay Gambera shared a milestone that really grounds this reality.

00:13:04: It has been exactly ten years since IBM put the first quantum computer on the cloud.

00:13:09: Ten years?

00:13:10: Yeah, what started as an incredibly niche experiment has compounded into an ecosystem of six hundred thirty thousand users operating on ninety one deployed systems.

00:13:20: But The Breakthrough That Truly Changes The Game For Enterprise Architecture was shared by Vajoy Pandey in Chinnton Patel.

00:13:25: They detailed Cisco's new universal Quantum Switch.

00:13:28: To understand why this matters, you really have to understand the core limitation of quantum hardware.

00:13:33: Which is that they are incredibly fragile right?

00:13:35: Historically you had these powerful quantum processors but their completely isolated.

00:13:40: They use different encoding modalities and if try connect them send quantum information the stage just collapses

00:13:46: Precisely.

00:13:47: What Cisco's team successfully built Is a prototype switch That can route entangled photons The actual carriers of Quantum Information.

00:13:57: It translates the data between entirely incompatible quantum systems.

00:14:01: And, The mechanism here is what glue my mind.

00:14:04: it does all of this at room temperature routing the data over standard existing telecom fiber optic cables with less than a four percent degradation and fidelity which

00:14:13: is huge.

00:14:14: right because usually Quantum anything requires temperatures colder then deep space?

00:14:18: The fact that it runs over existing Telecom fibers to true breakthrough means distributed quantum computing The literal foundation of the quantum internet is architecturally possible without digging up cities to lay new proprietary cables.

00:14:31: I have to play this skeptic for a moment though, if i am an IT leader listening to this deep dive right now my hair is on fire.

00:14:39: dealing with AI governance messy data migrating to five g. why should i dedicate any mental bandwidth?

00:14:45: two entangled photons and quantum switches today?

00:14:48: Because of a deeply concerning dynamic, Hemigomas Garcia refers to as crypto procrastination.

00:14:55: Crypto

00:14:55: procrastination?

00:14:56: Yeah you cannot wait for quantum hardware to sit on your desk before you update your defenses.

00:15:02: John Saville and Philip Intellora are sounding massive alarms regarding post-quantum cryptography or PQC.

00:15:09: Right because the harvest now decrypt later strategy.

00:15:13: Exactly, threat actors often stay sponsored are actively infiltrating networks.

00:15:17: today they were stealing deeply encrypted highly sensitive data.

00:15:21: we're talking health records financial ledgers national security communications

00:15:25: stuff that can't even read.

00:15:26: yet

00:15:26: right?

00:15:26: They know they can't break the encryption Today but there just hoarding it

00:15:30: their sitting on terabytes of stolen data Just waiting for a quantum computer to mature enough to shatter the mathematical algorithms protecting it.

00:15:38: Philip and Laura puts in very stark financial terms Quantum risk is already sitting on your enterprise balance sheet.

00:15:44: You just haven't accounted for it yet.

00:15:46: Wow migrating an entire enterprise to quantum safe security Isn't a simple software patch you push on a Friday afternoon?

00:15:53: It's as a massive multi-year transformation program.

00:15:57: if you wait till the quantum thread Is visibly at your door?

00:16:00: You are already mathematically out of time

00:16:02: which honestly provides A really powerful sobering bridge To our fourth and final theme digital sovereignty and governance.

00:16:10: When you combine the autonomous decision-making power of agentic AI with the incoming reality bending processing power of quantum, You are forced to ask The ultimate question

00:16:21: who actually controls this infrastructure?

00:16:22: Yes

00:16:23: Who controls it And when it inevitably breaks who is legally and financially responsible?

00:16:28: It Is a defining Question for enterprise leadership in This decade and we're seeing Europe take A very aggressive lead In answering.

00:16:36: Digital sovereignty is moving out of policy papers and into actual hardware deployments.

00:16:40: Byrne Wagner highlighted a phenomenal partnership between Siemens & Schwartz Digits.

00:16:44: They are building sovereign private, five-G networks tailored specifically for critical infrastructure And they aren't just slapping the sovereign marketing label on standard cloud products.

00:16:54: No!

00:16:54: The architecture's what matters here...they're utilizing air gap systems

00:16:58: Which?

00:16:59: Just to clarify for our listeners means that network has no physical or wireless connection with public internet.

00:17:05: it is entirely isolated.

00:17:07: Exactly true isolation.

00:17:09: and they are ensuring the entire value chain, hardware or software.

00:17:13: cloud hosting is predominantly German.

00:17:17: Wagner's overarching point is that digital sovereignty isn't decided by compliance checklists in a boardroom.

00:17:23: It is decided who physically controls servers on switches Ensuring there no hidden back doors from foreign entities.

00:17:29: We're

00:17:29: seeing same demand for sovereign control extend deeply into supply chains too.

00:17:34: Elky Andwell shared an update on Ford who's bringing roughly seventy of its top global suppliers into the Katina X data ecosystem.

00:17:41: Ah,

00:17:41: Katina x is fascinating because a goal to enable standardized fully traceable CO-II data sharing across the automotive supply chain

00:17:50: Right but without giving up control.

00:17:52: they are moving away from centralized monolithic databases where single tech giant holds everyone's data.

00:17:58: instead it's collaborative sovereign infrastructure.

00:18:01: The supplier retains absolute control over who sees their data and for how long.

00:18:06: But as we layer AI on top of these sovereign networks, the governance challenges multiply.

00:18:12: Nikhil S brought up a vital concept of AI supply chain responsibility.

00:18:17: It really is!

00:18:18: Every enterprise is racing to embed AI into their workflows.

00:18:21: but let's say an AI model hallucinates and denies legitimate medical claim or breaks automated manufacturing line.

00:18:28: Who owns that failure?

00:18:30: That is a massive legal blind spot.

00:18:32: Is it the data scientist who trained model?

00:18:34: Is that cloud provider hosting or identity management software granted AI access in first place?

00:18:40: Nikhil S argues simple provider versus deployer.

00:18:43: binary is obsolete, we need multi-layered risk frameworks.

00:18:46: you have to govern.

00:18:47: compute layer chip level software stack and organizational readiness.

00:18:52: hallucination security failure at any one of those layers cascades rapidly top.

00:18:57: Fortunately, we are starting to see the major players step up with actual blueprints for this.

00:19:02: David Kleinman noted that IBM just released their secure... AI Controls Framework, version two.

00:19:08: It's specifically designed to help highly regulated industries like the financial sector safely govern these autonomous AI agents

00:19:15: and it moves past theoretical compliance.

00:19:18: The framework embeds governance physically inside this system design.

00:19:22: It mandates things like autonomy tiering where an agent can only make decisions upto a certain financial threshold And literal kill switch design patterns so human can sever the agents access instantly.

00:19:34: It really brings us full circle back to the beginning of our deep dive.

00:19:38: The raw technological capability, this supersonic jet has arrived... ...the real work now is building the runways and control towers….

00:19:45: …and safety protocols to handle

00:19:46: it.".

00:19:47: The technology has matured!

00:19:48: The open question whether are organizational structures or data foundations in leadership incentives have matured enough safely execute

00:19:56: on them?

00:19:57: I think that's a perfect thought to carry forward.

00:19:59: But before we wrap up today, I want to leave you the listener with one final slightly uncomfortable reality check to mull over.

00:20:08: As you strategize how to build out these incredible new AI workflows and quantum ready defenses consider a warning shared by Eric Kimberling.

00:20:17: Kimberly focuses on traditional enterprise transformations right?

00:20:20: Yes!

00:20:20: And he noted that traditional ERP Enterprise resource planning projects still fail at a staggering rate of over eighty percent.

00:20:30: That is incredible, given all the advancements we have in cloud computing and software engineering.

00:20:34: Exactly!

00:20:34: And the reason they fail isn't because technology is flawed but it's because incentives are fundamentally misaligned.

00:20:41: The software vendors optimize for recurring revenue or ecosystem lock-in.

00:20:46: The system integrators pay to turn on the software not necessarily to ensure that you deliver your specific business value... ...and the client ends up carrying operational risk.

00:20:57: It is a critical dynamic to recognize.

00:20:59: As we shift from AI experimentation into massive enterprise execution, the vendors are going push you toward their proprietary stacks.

00:21:07: So as you sit in your next strategy meeting You have ask yourself very hard question Are you truly architecting a system optimized for business outcomes?

00:21:17: Your data cleanliness and digital sovereignty Or just feeding multi-million dollar machine optimized vendor lockin?

00:21:25: keep that in mind as you navigate this transition.

00:21:28: The infrastructure choices you lock-in today will entirely dictate the autonomy and security your business has tomorrow.

00:21:34: If you enjoy this episode, new episodes drop every two weeks.

00:21:37: Also check out our other editions on cloud defense tech digital products & services artificial intelligence sustainability And green ICT Defense Tech and HealthTech.

00:21:47: Thank You so much for joining us On this deep dive into insights shaping our industry.

00:21:51: Make sure you hit subscribe so that you don't miss the next one.

00:21:53: Until then, take a hard look at your data runways and stay secure!