Best of LinkedIn: Microsoft Ignite 2025

00:00:00: This episode is provided by Thomas Allgeier and Frennis, based on the most relevant LinkedIn posts talking about Microsoft Ignite.

00:00:07: Frennis supports enterprises with market and competitive intelligence decoding emerging technologies, customer insights, regulatory shifts, and competitor strategies, so product teams and strategy leaders don't just react, but shape the future of AI.

00:00:21: Welcome to the deep dive.

00:00:22: Today, we're digging into Microsoft Ignite, to give you, well, a strategic playbook.

00:00:29: The big narrative we saw coming out of all the sources is pretty clear.

00:00:32: It is.

00:00:32: AI is not just a bunch of scattered experiments anymore.

00:00:35: It's moving to a single, governed, and scalable operational stack.

00:00:41: Exactly.

00:00:42: And this deep dive, it isn't about just listing shiny new features.

00:00:45: We're really going to dissect the foundational architecture that Microsoft is building.

00:00:49: Okay, so what are we focusing on?

00:00:50: We're

00:00:51: going to unpack three key themes.

00:00:52: First, what's being called the Agentech Frontier, then the new security baseline that's required for all of this.

00:00:57: And finally, the convergence of data platforms.

00:01:00: These are really the big trends for any tech professional for the next year.

00:01:03: Okay, let's jump in.

00:01:05: The Agentech Frontier, that sounds big.

00:01:08: It is.

00:01:09: We've all been using co-pilots, right?

00:01:10: AI assistants.

00:01:11: That feels like phase one.

00:01:13: But people like Sabine van der Linden and Sanchibus Garan are talking about a leap to, well, phase

00:01:19: three.

00:01:19: Autonomous agents.

00:01:20: Exactly.

00:01:21: Agents that can plan, reason, and execute entire workflows with minimal human help.

00:01:27: They're meant to be like... digital colleagues.

00:01:30: And that idea that shift from assisted intelligence to fully autonomous agents, it creates this massive governance problem.

00:01:37: Right.

00:01:38: If you have hundreds of these digital colleagues running around, how do you track them?

00:01:42: How do you manage and secure them?

00:01:43: Which

00:01:44: brings us to probably the biggest announcement in this space.

00:01:47: Agent three sixty five.

00:01:48: Max Goss described this as the comprehensive control plane.

00:01:51: It's basically designed to stop the fragmentation you'd otherwise get

00:01:55: because agents would be popping up everywhere.

00:01:56: Exactly.

00:01:57: Different services, different.

00:01:59: I like that line from Samuel Boulanger that we now need to manage agents like we manage people.

00:02:04: This kind of central control becomes mandatory.

00:02:07: It absolutely is.

00:02:08: I mean, think about HR systems for people.

00:02:10: Agent management has to be just as robust.

00:02:14: Jack Robotham actually detailed the core parts of this.

00:02:16: Oh, interesting.

00:02:17: What are they?

00:02:18: First, you have a registry.

00:02:20: Basically, a single source of truth to discover and track every single agent in the company.

00:02:25: You can't secure what you can't see.

00:02:27: That visibility is key, of course.

00:02:29: But what about control?

00:02:31: Because if an agent is autonomous, its permissions are a huge risk.

00:02:35: That's the second piece.

00:02:36: Access control.

00:02:37: Every agent gets a unique ID.

00:02:39: He uses policy templates to enforce the principle of least privilege.

00:02:43: Okay.

00:02:44: And for security, this is all integrated right out of the box with Microsoft's Entra Defender and Purview.

00:02:49: The governance is baked in, not bolted on later.

00:02:53: So if Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-Agent-A

00:03:14: Denizelki pointed out that it now supports hosted agents, complex multi-agent workflows, and even has built-in memory.

00:03:22: And deployment.

00:03:22: Direct deployment into Microsoft.

00:03:27: It's ready for mission-critical tasks.

00:03:29: And the power source for that factory, the models themselves, that's expanding too.

00:03:32: I saw a lot of talk about the partnership bringing anthropic cloud models into Azure Foundry.

00:03:37: Right, like cloud sonnet four point five and opus four point one.

00:03:40: Yeah,

00:03:41: people like Mike Krieger and Carly Young highlighted this.

00:03:43: It gives enterprises these, you know, high-end frontier models to build with.

00:03:47: And more choices always better.

00:03:48: But an agent, no matter how powerful the model is, is useless without context.

00:03:53: Ah, and that's where the new IQ layer comes in?

00:03:55: Exactly.

00:03:56: Nicole Herskovitz and Annie Alassila talked about work IQ.

00:03:59: It's the intelligence layer that finally lets co-pilot and other agents move beyond just simple Q&A.

00:04:05: OK, so how does this IQ layer actually make things smarter?

00:04:08: What's it doing?

00:04:09: It connects three things.

00:04:10: First, your company's internal data.

00:04:13: Second, the user's personal context and preferences, which they're calling memory.

00:04:17: And third, predictive action or inference.

00:04:20: So it combines what the company knows, what I know, and what it thinks I'll do next.

00:04:24: You

00:04:24: got it.

00:04:25: And by combining those, the agent can anticipate what you need and execute these really contextual workflows.

00:04:31: It's the the connective tissue between raw data and autonomous action.

00:04:35: Okay, that makes sense.

00:04:36: Let's shift gears a little.

00:04:37: To

00:04:37: security.

00:04:38: Yes.

00:04:38: If autonomous agents are the future, then security has to be the floor they stand on.

00:04:43: Shoaib Kudzi said security was maybe the biggest topic at Ignite.

00:04:47: It signals a huge shift.

00:04:49: AI security is moving from being optional to being the new, you know, essential standard.

00:04:54: And there's a huge strategic decision here.

00:04:57: The big news for Microsoft, three sixty-five e five customers is that security popilot is now just included.

00:05:03: Right.

00:05:04: The BLA to Missy Ann and Chris S both noted that removing that procurement barrier is a massive signal.

00:05:10: It's a profound shift in the security baseline.

00:05:13: It really is.

00:05:14: By bundling an AI-powered security agent across Intra, Defender, Purview, and Intune, Microsoft is basically making AI security standard for its enterprise clients.

00:05:24: Not some premium feature.

00:05:26: So

00:05:26: your security operations center is no longer starting from scratch.

00:05:29: They get an AI co-pilot by default.

00:05:31: But let's get specific.

00:05:33: How does Defender actually police these new digital colleagues?

00:05:38: Well, Defender itself has become agent aware.

00:05:41: Vlad Krasansky and Thomas Chengjizu detailed this.

00:05:44: It focuses on three main things.

00:05:46: First, getting a unified inventory and visibility.

00:05:48: Define the shatter agents.

00:05:49: Exactly.

00:05:50: The ones built outside of the official channels.

00:05:52: Second, risk reduction.

00:05:54: So identifying misconfigurations or agents with way too many permissions.

00:05:57: Send

00:05:57: the third.

00:05:58: Direct threat protection against AI specific attacks like prompt injection that tries to compromise an agent's privileges.

00:06:05: So if agents are like employees, they need an identity.

00:06:08: for governance.

00:06:09: Right,

00:06:09: which is why Anujrana and Hanku talked about intra-agent ID.

00:06:13: This gives every agent an identity, just like a human employee.

00:06:16: It's essential for auditing and controlling them.

00:06:19: And this focus on governance is also spilling over into compliance.

00:06:23: It is.

00:06:23: Sammy Eunis pointed out that Microsoft Sentinel is expanding into compliance operations.

00:06:28: It has new solutions for things like GDPR and API.

00:06:31: That's a big deal.

00:06:32: Compliance isn't usually the focus of Seacops tools.

00:06:35: Well, the goal is what he called compliance through operations.

00:06:38: So instead of compliance being a retroactive audit, you're building the checks right into your security workflow.

00:06:44: It minimizes regulatory risk from the start.

00:06:47: And on a more practical note, I saw the new Defender Deployment Tool.

00:06:51: Preview, yeah.

00:06:52: Wesley Vanden Heuvel and Michael Mortensen mentioned it.

00:06:55: It just simplifies onboarding for Defender for Endpoint, especially for Ores using things like GPO or SCCM.

00:07:02: A small thing, but it reduces a lot of friction.

00:07:04: OK, so we've got Agents and Security.

00:07:06: None of this works unless the data is accessible and, more importantly, unified.

00:07:11: Let's pivot to data.

00:07:12: Microsoft Fabric.

00:07:13: Yep.

00:07:14: Robert Kramer talked about its push to be that truly unified analytics platform, and this seems to be powered by something called Fabric IQ.

00:07:21: The term Fabric IQ, which Tafiya Wisi mentioned, is key.

00:07:26: It's more than just a layer.

00:07:27: It's the semantic foundation.

00:07:29: It uses ontology.

00:07:30: A

00:07:30: system of naming and categorizing things.

00:07:32: Exactly.

00:07:33: To create a common understanding across data, meaning, and actions.

00:07:37: This is crucial because it means AI agents and human analysts are working from the same business context.

00:07:44: So instead of just raw data, the data now inherently knows what it means.

00:07:48: Like this is Q for revenue and this is customer retention and that meaning is standard everywhere.

00:07:54: Precisely.

00:07:55: As Rajan C explained, this is what lets you move beyond dashboards to create intelligent experiences.

00:08:00: The AI is working from a foundation of standardized business knowledge.

00:08:04: And on top of that foundation, we saw a huge commitment to interoperability.

00:08:08: Zero copy data sharing was everywhere.

00:08:10: It really was.

00:08:11: Wangi, McKelvie, Omar Khan.

00:08:12: Yeah.

00:08:13: A lot of people were talking about it.

00:08:14: This is where the old data silos really start to break down.

00:08:17: The biggest one I thought was a partnership with SAP.

00:08:20: Oh, for sure.

00:08:21: Thomas Sowers had confirmed the SAP Business Data Cloud Connect for Microsoft Fabric.

00:08:27: It allows bi-directional zero copy sharing of that semantically rich SAP data.

00:08:33: Which is huge because SAP data is often the most complex and critical.

00:08:36: And they're extending the zero copy idea to their competitors too, which is smart.

00:08:41: Azure Databricks can now natively read data from One Lake.

00:08:44: And Snowflake customers can use One Lake as native storage.

00:08:47: So, Fabric becomes this open central hub without forcing painful migrations.

00:08:52: We also saw some serious infrastructure upgrades.

00:08:55: Luke Fangman detailed Azure Horizon DB, which is a new cloud-native post-graceful service built for massive scale.

00:09:01: Oh, massive.

00:09:02: Up to three thousand seventy-two vCore, so it's built for very data-intensive workloads.

00:09:07: And for Dagger Transformation, there's the DBT Fusion integration.

00:09:10: Nina Anderson noted that brings the DPT engine right into fabric.

00:09:13: Yeah, better transformations right where the data lives and to bring this all back down to earth for smaller organizations.

00:09:18: Right.

00:09:18: The new M-E-T-E-S-T-E-F-I co-pilot business guests.

00:09:22: Exactly.

00:09:23: Paul Randazzo and Allison West Hughes emphasize this.

00:09:26: It's now available via CSP, twenty-one dollars per user per month, for orgs up to three hundred users.

00:09:32: That really democratizes access to these enterprise-grade AI tools.

00:09:36: It does.

00:09:37: Okay, so let's talk about how all this changes our daily grind.

00:09:40: I really liked Aaron McGough's point.

00:09:43: AI isn't the story anymore, it's the utility.

00:09:46: Yes.

00:09:47: It's like talking about the internet.

00:09:48: It's just the expected underlying tech.

00:09:50: The focus shifts immediately to the impact these agents have on collaboration and productivity.

00:09:56: The plumbing's done, so what can we build?

00:09:58: And

00:09:59: we saw some really practical updates in Microsoft Teams that support this.

00:10:02: Nitish Gulcha and Saheep Singh detailed a few.

00:10:05: For instance, one chat.

00:10:06: which lets you start a conversation with anyone just using their email.

00:10:09: Right.

00:10:10: It instantly removes that friction with external partners, but that also raises security questions.

00:10:15: Which they address with trust indicators.

00:10:16: So you get these clear labels like external or guest.

00:10:19: So you always know who you're talking to and importantly who an agent might be sharing information with.

00:10:24: And for anyone who juggles multiple accounts, the MTMA Cross Tenant Activity Hub is a lifesaver.

00:10:31: You can respond to chats across different accounts without constantly switching context.

00:10:35: A real efficiency booster.

00:10:38: But beyond the chat tools, the agents need a good internal knowledge base to be truly smart.

00:10:43: The fuel for the agents.

00:10:45: Precisely.

00:10:46: Prashant Chandrasekhar announced the launch of Stack Internal, which is the evolution of Stack Overflow for teams.

00:10:53: It's for engineering teams to manage their own enterprise intelligence.

00:10:57: And that learning layer is getting embedded directly into workflows too.

00:11:01: Eric Lee noted that Cornerstone's AI learning agent is now available in co-pilot.

00:11:05: So it helps employees upskill or find training right when they need it.

00:11:09: These integrated knowledge systems ensure the agents are always acting on the most current, correct internal intelligence.

00:11:15: It's the difference between a generic answer and an answer based on your company's own proprietary best practices.

00:11:20: And that distinction is everything.

00:11:22: If you enjoyed this episode, new episodes drop every two weeks.

00:11:25: Also check out our other editions on cloud insights, sustainability and green ICT, digital products and services, health tech, defense tech, ICT and tech insights and artificial intelligence.

00:11:35: You know to tie this all together we saw these staggering predictions like one point three billion AI agents by twenty twenty eight.

00:11:44: but Alistair Wolcock provided a really necessary counterbalance to that.

00:11:48: He said adoption metrics are becoming a vanity metric.

00:11:51: This just counting usage isn't enough anymore.

00:11:54: Not even close.

00:11:55: The real strategic shift for every leader listening is to move your KPIs immediately from just usage to ROI realization.

00:12:03: The infrastructure is there now.

00:12:05: But if you can't prove productivity gains measured against real business KPIs, you risk a major adoption backlash.

00:12:11: That's the challenge for the next twelve months.

00:12:13: If the infrastructure is now standard, the pressure is squarely on demonstrating real, measurable business value.

00:12:19: That is a great final thought for you to take back to your teams.

00:12:23: Thank you for joining us for the deep dive.

00:12:25: Be sure to subscribe so you don't miss our next episode.