Best of LinkedIn: Google Cloud Summit DACH 2026

00:00:00: This episode is provided by Thomas Allgeier and Frennis, based on the most relevant LinkedIn posts about Google Cloud Summit DACH.

00:00:09: Frennes has built a sovereign cloud market radar for ICT providers with weekly hot news monthly reports quarterly leadership presentations and AI podcasts for field teams.

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

00:00:21: so uh imagine an autonomous system that doesn't just draft an email but actually analyzes live weather report Cross references it with your SAP supply chain data, identifies a typhoon that's disrupting your tier two suppliers and then instantly reroutes you inventory across three different continents.

00:00:39: Which sounds totally like a sci-fi pitch right?

00:00:41: Exactly!

00:00:41: It sounds like sci-Fi but all without human ever clicking single button.

00:00:45: And uh... That is exactly what top companies in the DACH region are literally putting into production now.

00:00:51: Yeah its wild.

00:00:52: Welcome to The Deep Dive.

00:00:53: today we're tearing this massive stack of curated LinkedIn posts from ICT and tech professionals who just walked out at the Google Cloud Summit DACH in Frankfort.

00:01:04: Yeah, you know scrolling through their reactions to people that are actually building this infrastructure... The energy shift is undeniable.

00:01:14: Oh absolutely!

00:01:15: For ICT professionals tone has fundamentally changed.

00:01:19: I mean, Luca Caligari the managing director of Google Cloud DACH.

00:01:24: He stated at the event that the industry has officially moved from The Era Of Eventually to The Era of Immediately.

00:01:33: Right and that perfectly captures the mood in these posts.

00:01:35: everyone is just like getting to work.

00:01:38: so i guess the mission for this deep dive Is To Sort-of Cut Through the vendor noise, right?

00:01:43: And look at the ground level reality.

00:01:45: We want you, The Listener to understand exactly how leading enterprises are moving past- He passed

00:01:51: all those endless proof of concept.

00:01:52: Yeah, endless posseists Moving pass them and into production grade sovereign agentic execution.

00:01:57: because almost every post from the summit made one thing crystal clear uh...the chatbot era is dead.

00:02:03: It's so dead!

00:02:10: generate text or write a nice poem, we're talking about autonomous agents that actually execute complex workflows across enterprise software.

00:02:19: Which

00:02:19: is huge jump!

00:02:20: Huge.

00:02:21: and what's fascinating here?

00:02:22: just the sheer velocity of this transition in the real world.

00:02:26: We saw some incredible insights shared about delivery hero for example.

00:02:29: Well I saw that post from Nehal Shah right?

00:02:31: Yeah Nehal Shah and Mario S. they demonstrated how they deployed AI agent into production In just one hundred and fourteen days.

00:02:39: Wait

00:02:39: hold on A hundred and fourteen days.

00:02:41: Yeah, a hundred and fourteen days.

00:02:42: Frankly, one-hundred and fourteen to production for an enterprise of that scale seems almost suspiciously fast to me!

00:02:50: I mean how does an ICT team actually build something that touches millions of customers without completely breaking their core infrastructure?

00:02:59: It sounds risky...

00:03:00: Are we talking about some massive monolithic AI system they just shoved in there?

00:03:04: No,

00:03:05: no and that's actually the secret to their speed.

00:03:06: They aren't building a monolithic brain at all... ...they build what they call the CX Agent Studio.

00:03:11: CX agent studio okay

00:03:13: yeah it's a modular architecture.

00:03:14: so they are using smaller highly specialized language models That are tightly integrated like via APIs into their existing customer service platforms.

00:03:24: Ah!

00:03:24: Okay.

00:03:24: So let's plug-in play

00:03:26: Exactly.

00:03:26: Instead of trying to train one massive model on literally everything, the agent uses our Reiki Retrieval Augmented Generation To pull the exact policy needed for a specific customer issue.

00:03:38: It makes a decision based on that tiny slice of context and then triggers an API call to resolve it.

00:03:43: And that modularity.

00:03:44: That's how they are resolving millions of customer contexts right now.

00:03:47: That makes way more sense from an architectural standpoint.

00:03:50: And, you know they aren't just using it for customer experience either.

00:03:53: I read that their finance teams are using agentic AI to autonomously generate executive summaries and really complex variance analyses.

00:04:02: Yeah which is traditionally a total nightmare for finance teams.

00:04:05: Right

00:04:05: They're literally taking the brutal manual end of month reporting cycle and handing data extraction synthesis over

00:04:15: And the scale of deployment across the board is shifting dramatically.

00:04:19: It's no longer just like a specialized tool for developers or data scientists locked in the basement somewhere, look at Signal Iduna.

00:04:27: According to posts highlighting Patrick Darby and Nutz Engerling, they rolled out Gemini Enterprise... ...to their entire workforce.

00:04:34: The entire work force?

00:04:35: A whole

00:04:35: company!

00:04:36: They completely democratized AI access

00:04:39: Which is a huge structural shift.

00:04:41: I mean let's look at Martin Geisler perspective from Bosch too.

00:04:44: Oh

00:04:45: yeah the Bosch use case is wild.

00:04:46: They are scaling AI to one hundred thousand employees.

00:04:50: One hundred thousand.

00:04:51: He noted that Bosch treats agentic AI not as an IT rollout, but has a massive organizational change project.

00:04:57: That just happens to be powered by tech.

00:04:59: Yeah exactly.

00:05:00: But I have to play devil's advocate here for sure.

00:05:02: go for it.

00:05:03: every major software deployment gets called and Organizational Change Project By leadership.

00:05:08: It's the classic corporate speak.

00:05:11: so what are they actually doing differently with those hundred thousand employees?

00:05:14: Okay its very fair pushback, but The difference Here is that Think about a traditional IT rollout like migrating to new ERP system.

00:05:23: Right,

00:05:24: everyone hates

00:05:24: that!

00:05:26: But it just requires teaching employees to click new buttons in the interface.

00:05:30: It's still the same basic task.

00:05:33: Agentech.

00:05:34: AI requires teaching employers to fundamentally redesign their own workflows.

00:05:38: Oh I see.

00:05:39: So change management at Bosch is really shifting the workforce mindset from how do i do this task manually?

00:05:46: How do I orchestrate an agent?

00:05:48: for me

00:05:49: So, it's a shift from execution to management.

00:05:52: Exactly!

00:05:52: It is like we've upgraded from really smart dictionary that just answers your questions... ...to digital intern who actually fills out the paperwork files and emails client.

00:06:01: Okay, I love the digital intern analogy but that actually raises a massive operational red flag for me.

00:06:06: Oh well.

00:06:07: A digital intern is great But if my physical intern has access to the wrong filing cabinets They're gonna send the wrong maybe confidential files to the client.

00:06:17: so if we empower agents To execute tasks autonomously across an enterprise how do we ensure they aren't pulling?

00:06:24: you know, outdated pricing data or just straight-up hallucinating inventory numbers.

00:06:28: And that operational red flag you just raised?

00:06:30: That is exactly why the conversation at The Summit instantly pivoted to the fuel source for these agents which is data.

00:06:37: Data right!

00:06:38: Yeah and agent as only reliable in business context it can instantly access.

00:06:43: if your data layer The agent's actions won't just be wrong, they will be catastrophic at machine speed.

00:06:49: Yeah executing bad decisions ten thousand times a minute.

00:06:52: Exactly you cannot have non-deterministic AI models running enterprise workflows without highly deterministic rock solid data foundations.

00:07:00: So the focus kind of has to shift away from the flashy AI models themselves and back to the data engineering underneath Precisely.

00:07:06: And this is why the Schwartz Group presentation by Mark Fabian Mezger and Matthias Julek got so much traction in the ICT community.

00:07:13: Oh i saw people raving about that one.

00:07:15: Yeah, they showcased how highly structured data and rigorous continuously validated testing frameworks are just the absolute baseline for their agentic document processing.

00:07:25: Right They're reading continuous automated testing on the data inputs before the agent is even allowed to touch The document

00:07:32: guardrails And I noticed a massive emphasis on integrating this directly into the SAP ecosystem too.

00:07:39: Oh absolutely It seems like AI isn't just being bolted onto the side of the enterprise as an afterthought anymore.

00:07:45: It's being wired directly into the operational core,

00:07:48: it has to be.

00:07:49: take Cortex seven which was heavily discussed at The Summit.

00:07:52: they are merging SAP data with Google AI To build truly resilient supply chains.

00:07:57: mm-hmm the AI is constantly reading the live SAP ledger yeah to predict bottlenecks before They even happen.

00:08:03: and we also saw Deutsche telecom successfully migrating their legacy SAP systems to the cloud.

00:08:08: And didn't they see a massive efficiency bump from that?

00:08:10: Huge.

00:08:11: By modernizing that data foundation, they actually saved forty percent in energy costs.

00:08:17: So the Modern Data Foundation isn't just an enabler for AI It's driving massive operational efficiency all on its own

00:08:25: Right and you see this impacting the end consumer directly too.

00:08:28: I mean Mike Hullberg posted about OTTOs conversational product search.

00:08:32: Oh yeah That is a prime example of AI moving directly into the retail shopping experience.

00:08:39: But you know, think about the mechanism there

00:08:41: like the plumbing?

00:08:41: Yeah

00:08:41: The plumbing an agent finding the right product in a split second relies entirely on a real time data pipeline.

00:08:49: because if the vector database that's powering the AI isn't syncing instantly with the live inventory API then

00:08:56: the agents are going to recommend a product it has out of stock

00:08:58: exactly and Then It becomes worse than useless actively damages the brand trust.

00:09:02: That's spot-on, The intelligence is functionally limited by the latency of data pipeline underneath it.

00:09:08: But okay here an observation shared with attendees like Daniel Hurd that I thought completely flipped script on this whole data thing.

00:09:15: Okay what was?

00:09:16: For years we've been told datafragmentation and data silos were biggest roadblocks to AI adoption.

00:09:22: Right...the classic.

00:09:23: We need clean our data first

00:09:25: argument.

00:09:25: Exactly, but at this summit the sentiment was that AI agents are actually becoming the cure for data silos.

00:09:32: Let's unpack how our agent is curing the silos.

00:09:36: Well think

00:09:36: back to scenario I mentioned in the very beginning of this deep dive

00:09:40: The typhoon one

00:09:41: Yeah, the semiconductor use case they discussed.

00:09:44: They deployed an agent that simultaneously analyzed SAP master data CRM Data from Salesforce and live meteorological updates.

00:09:53: Wow all at once.

00:09:54: All

00:09:54: it wants.

00:09:55: It used all of those disparate data streams to instantly calculate the impact of a typhoon on their tier two suppliers Identifying exactly who was affected, And what the alternate routing options were.

00:10:05: That

00:10:06: is nuts!

00:10:06: If breaks down silos Instantly The Agent Is doing the complex cross-system API querying and cross referencing You know, used to take human analyst days of polling and joining different Excel reports.

00:10:18: That is a phenomenal example.

00:10:20: what we mean by business context?

00:10:21: The agent isn't just parsing text.

00:10:23: it's actively reasoning across fragmented legacy systems To deliver an immediate actionable business outcome.

00:10:30: It is incredible but uh...it does bring up the massive unavoidable elephant in the room for the DICH region.

00:10:35: Ah yes Sovereignty.

00:10:37: Yeah,

00:10:38: if we are taking all this hypercritical enterprise data patient records from hospitals banking ledgers global supply chain vulnerabilities and pumping it through US based hyperscaler AI models We immediately run into the U S cloud

00:10:52: Act.

00:10:53: Yep

00:10:53: for European ICT providers digital sovereignty isn't just a buzzword.

00:10:57: It is a hard inflexible legal requirement

00:11:04: And you could really see it dominating the discussions at this summit.

00:11:07: I mean Google addressed the physical infrastructure side directly,

00:11:10: right?

00:11:10: The data center?

00:11:11: Yeah highlighting their five point.

00:11:12: five billion investment in Germany through twenty-twenty nine which includes construction kicking off on new AI data centers and Hanau.

00:11:19: So they are physically putting the compute power inside the German borders.

00:11:23: Okay But let's be real here physical location doesn't solve the legal jurisdiction problem.

00:11:29: No, it doesn't.

00:11:30: Just because a server is sitting in Hanau Doesn't mean that US intelligence agency can't subpoena data under the cloud act if an American company operates it

00:11:39: Exactly which is why the biggest partnership news of this summit was so critical to the ICT audience.

00:11:45: Christoph Ruffner, the CEO of Fairleys Germany alongside Mark Bachmann announced they are launching a legally independent German entity to operate Google Cloud dedicated.

00:11:56: Okay hold on let me stop you there.

00:11:57: for those of us who Don't spend our weekends reading BSI compliance manuals.

00:12:02: Fair enough?

00:12:02: What does

00:12:03: a legally independent entity actually mean?

00:12:05: mechanically, like how does that actually stop outside

00:12:08: access?

00:12:08: OK think of it like renting space in a highly secure building.

00:12:11: Google built the building they provide physical hardware and software stack But Thales Germany owns only Kate at front door.

00:12:19: They manage security cameras And strict German law dictates who is allowed to walk into the lobby Fully decoupling technology from public google cloud.

00:12:28: Ah, I see.

00:12:29: So the day-to-day operations—the identity and access management, the encryption keys — it is all handled exclusively by Thaley's personnel inside Germany.

00:12:39: They are targeting these C-three A framework.

00:12:41: And what exactly does?

00:12:42: C three a mandate?

00:12:43: so The cloud computing compliance criteria catalog specifically the highest tier mandates absolute structural separation.

00:12:51: It ensures that the hyperscaler literally cannot access the customer's plain text data or operations, even if compelled by a foreign warrant.

00:12:59: Wow!

00:13:00: Yeah it insures the data stays strictly within the German legal and operational framework.

00:13:04: So its not just a policy document they sign?

00:13:07: Its a cryptographic an operational barrier?

00:13:09: Yes exactly And we are already seeing how this unlocks innovation in highly regulated sectors.

00:13:15: like Professor Jens Scholz from UKSH, the University Hospital Schleswig-Holstein made a really profound point during our fireside chat.

00:13:22: Oh what did he say?

00:13:22: He noted that sovereignty and innovation are no longer a contradiction.

00:13:26: they are actively deploying AI right at the patient bed today because this sovereign infrastructure now actually exists to support CREDUS.

00:13:37: You know critical infrastructure requirements.

00:13:39: That's huge for health care.

00:13:40: And it's not just hospitals either.

00:13:42: Sebastian Angerstein and his team at a ok-needers accent actually won the Google Cloud Innovator Award For building a highly secure sovereign AI ecosystem for Health insurance data, but I do have to ask you appointed question about this

00:13:56: okay shoot

00:13:57: historically in this industry Choosing an air-gapped or strictly sovereign cloud meant you were stuck with technology that was like two years out of date.

00:14:04: Oh, absolutely the sovereign line right?

00:14:06: You sacrifice the cutting edge features for compliance.

00:14:09: Are they ICT professionals at this summit actually convinced that this new fail as Google model breaks That historic compromise

00:14:17: based on the deep technical sentiment in these posts?

00:14:19: Yes They Actually are.

00:14:20: The feedback indicates if This isn't some watered down legacy version Of the Cloud.

00:14:25: this is the full modern hyperscaler stack just operated locally.

00:14:30: That is a game changer,

00:14:32: it is.

00:14:32: however while the sovereign cloud solves the legal and compliance side of security transitioning to an egentic enterprise requires completely new cybersecurity architecture internally because you are now introducing autonomous machines.

00:14:49: be actors into your own network

00:14:51: right?

00:14:51: Because using agents to execute workflows sounds great in control vacuum.

00:14:57: But if I'm an enterprise architect and i have thousands of these Autonomous digital interns running around my network

00:15:03: doing whatever they want

00:15:04: executing api calls writing to databases sending emails to clients how do I police them?

00:15:09: We used to secure the perimeter to keep hackers out.

00:15:12: now The potential threat like an agent misinterpreting a prompt in dumping A database is already inside.

00:15:17: The house the

00:15:18: scale Of, the threat landscape has fundamentally shifted which Is forcing security teams.

00:15:22: you use AI just To survive.

00:15:24: Look at the operational reality shared by Florian Schurfeld from Travago.

00:15:28: Okay

00:15:29: He revealed that a security team of just five people Manages ten thousand alerts daily using whiz on Google Cloud.

00:15:35: Wait, stop!

00:15:36: Five People Handling Ten Thousand Alerts A Day.

00:15:39: Yeah That mathematically doesn't work...that usually leads to massive alert fatigue Where The Team Just Ignores The Dashboard Until A Breach Happens.

00:15:47: How Are They Physically Doing That?

00:15:49: They are Using AI to instantly triage all the noise.

00:15:54: The system auto-correlates logs, identifies misconfigurations across their entire cloud footprint, weeds out false positives and automatically closes low priority alerts.

00:16:06: So it surfaces only actual verified threats that require human intervention.

00:16:10: And through this automation those five people achieve zero critical vulnerabilities.

00:16:15: That is wild.

00:16:16: So with concepts like Google's AI threat defense, which Ola Noa talked about we are essentially deploying security agents.

00:16:22: Yeah We have digital security guards hunting threats and misconfigurations at machine speed.

00:16:27: Right They're using agents to police the agents.

00:16:29: that's

00:16:30: The operational reality now.

00:16:31: but you also need structural governance At the architectural level.

00:16:35: Seny Vasan Devassanan from Commerce Bank detailed a really fascinating strategy To tame what he called agendic ai sprawl.

00:16:42: Agentic Ai Sprawl I love that term, because if marketing builds an agent and HR builds a agent.

00:16:50: And finance builds an agents... It's

00:16:52: chaos!

00:16:53: Yeah suddenly you have shadow IT on steroids and the software is actively making decisions.

00:16:59: how does highly regulated banks stop them?

00:17:02: Are they just limiting who can use API?

00:17:04: No And that's the key.

00:17:05: They aren't restricting innovation, they're enforcing architecture.

00:17:09: Kemmer's Bank is scaling AI responsibly by implementing three specific layers... Which are?

00:17:14: An agent registry and an AI gateway in a strict classification system.

00:17:18: So think of the agent registry like an employee ledger for code.

00:17:22: The department cannot just deploy a multi-agent system into the wild, that system has to be officially registered given identity token and its permissions are hard coded in to AI gateway.

00:17:32: Ah!

00:17:32: so if marketing departments agents suddenly tries to query HR payroll database?

00:17:37: The AI Gateway instantly blocks API call based on strict identity and access management policies.

00:17:43: Exactly They're proving banking.

00:17:46: innovation must adapt to the governance framework, not the other way around.

00:17:51: You really need a centralized ledger of every autonomous entity in your

00:17:55: network.".

00:17:56: That makes complete sense and it leads perfectly into how this changes the ways software is actually built.

00:18:01: on first place There was massive shift in developer culture discussed at The Summit by Matt Thompson.

00:18:07: he called it Shift Down Engineering.

00:18:09: Yes, Shift down.

00:18:11: The traditional siloed cycle of develop, deploy and operate is just collapsing into a single seamless motion.

00:18:17: Right

00:18:17: we saw multiple posts talking about product managers vibe coding.

00:18:21: Vibecoding?

00:18:22: So they're getting used to that phrase

00:18:23: I know.

00:18:23: wait They literally call it vibecoding.

00:18:25: It sounds ridiculous But they are literally using natural language AI To rough out code in system concepts Shrinking the development cycle from three months Of endless meetings And debating specs down to Just Three weeks of actual building.

00:18:38: Yeah

00:18:38: And Pavelis Bauscher from Delivery Hero showcased hydrogen, which completely blew my mind.

00:18:44: It is a multi-agent system that autonomously takes engineering work form a Jira ticket directly to a pull request.

00:18:50: and think about the mechanism of a multi agent system doing that.

00:18:54: it's not just one big bot.

00:18:55: That does everything right?

00:18:56: A jura ticket comes in One agent parses the requirements It hands that to a second agent, who writes the actual function.

00:19:03: A third agent writes unit tests for this function and a fourth agent reviews code against the company's repository guidelines.

00:19:11: before creating the pull request The human developer is no longer just writing syntax.

00:19:17: they are orchestrating a system of specialized agents that write test and deploy their codes.

00:19:23: It's a total paradigm shift for ICT providers.

00:19:25: So what does this all mean to you, the listener?

00:19:28: If we synthesize everything that has been pulled from this massive stack of summit insights... What is the ultimate takeaway from

00:19:34: Frankfurt?".

00:19:34: The core message was that the proof-of-concept phase is dead.

00:19:38: In twenty-twenty six enterprise value is defined by very specific three part intersection.

00:19:44: It is the intersection of egentic workflows that autonomously execute tasks, highly trusted deterministic data foundations that provide flawless business context and uncompromising digital sovereignty.

00:19:56: That ensures you legally an operationally own your destiny.

00:20:00: if your architecture's missing any one Of those three pillars You simply cannot scale.

00:20:05: in The modern enterprise environment.

00:20:06: it's

00:20:06: no longer about experimenting its About execution which leaves you with something highly provocative to ponder.

00:20:13: Several posts highlighted the rapid rise of agent-to-agent or A to a communication powered by frameworks like The Model Context Protocol, or MCP.

00:20:22: Oh

00:20:22: yeah, MCPs huge!

00:20:23: For context, MCP allows different AI models to securely connect and share contexts across different data sources.

00:20:30: We are entering a world where your company's AI agents will seamlessly negotiate cross reference inventory and trade data with your suppliers' AI agents in real time completely autonomously.

00:20:39: It is a profound shift in how BDB commerce and date exchange will function at the structural level Exactly,

00:20:45: so as humans increasingly move out of the execution loop And become pure orchestrators of these autonomous systems you have to ask yourself A fundamental question about your architecture.

00:20:55: Yeah Are you still designing your IT infrastructure As a workspace for your human employees?

00:21:01: Or are you actively preparing it To be an optimized highly governed environment for your new digital workforce.

00:21:08: It completely reframes what enterprise architecture even means moving forward.

00:21:12: it really does.

00:21:12: the digital interns are already in building and they're ready to get work

00:21:16: absolutely.

00:21:16: if you enjoyed this episode, new episodes drop every two weeks.

00:21:19: also check out our other editions on ICT & Tech Digital Products & Services Artificial Intelligence Cloud Insights Sustainability And Green ICT Defense Tech And Health.

00:21:29: Thank You So Much For Joining Us On This Deep Dive.

00:21:32: Don't Forget To Subscribe.

00:21:33: Catch You On The Next One.