In The Loop Episode 17 | Google I/O & Microsoft Build In 10 Minutes: What We Learned From The Two Biggest AI Conferences

In just seven days, Google and Microsoft launched 150 new AI features between them at the two biggest AI conferences of the year. Microsoft alone mentioned agents 300 times over the course of a few days. And while both companies are clearly obsessed with AI, they've taken very different approaches — something that came through clearly at their respective events.

Google has gone all in on weaving AI into consumer products, while Microsoft painted a vision of an open, agentic web transforming the world of work.

In today’s episode, I’m going to explore what I found most interesting in their announcements — and what it might mean for the future direction of AI.

This is In The Loop with Jack Houghton. I hope you enjoy the show.

‍

‍

Microsoft Build 2025: Enterprise AI strategy

For today’s episode, I’m going to do my best to round up the major announcements from the conferences. As I mentioned, there were over 150, so I won’t go through all of them. Instead, I’ll cherry-pick what I found most interesting — especially in terms of what it means for the future of AI over the next three, six, or nine months.

I’ll start with Microsoft Build — their flagship developer event — because I think it was actually more compelling. They introduced a new vision they’re calling the "frontier firm." This is an organization that embeds AI from the ground up, where humans are surrounded by teams of agents and everyone becomes an agent manager. Traditional org charts get scrapped, and people evolve into AI bosses.

Microsoft CEO Satya Nadella said they’re going beyond just sharing case studies with customers. Their goal is for 50% of all code to be written by AI within the next year. To put that into perspective, they’re currently at 20 to 30%. They've also just announced layoffs — about 3% of their workforce, or roughly 6,000 people. Honestly, it wouldn’t surprise me if this is part of a broader move to reinvest in AI growth — a case of eating their own dog food.

Microsoft is positioning itself as customer zero for all the AI solutions it's pitching to the enterprise. And at the heart of that vision is the “open agentic web,” which dominated the first day of BUILD.

So, what is the agentic web? It's a collection of ideas we’ve talked about on this podcast before. They’re not new, but Microsoft’s backing and the narrative they’re building around them could speed things up significantly. The idea is to have AI agents that can interact with online products and services — booking appointments, collaborating with other agents, purchasing products on your behalf.

The “open” part is key. Microsoft doesn’t want to build a closed ecosystem. They’re advocating for open, interoperable standards that allow any company to embed their agents, collaborate, and transact. Microsoft’s ecosystem would help facilitate that.

As I’ve said before, these big companies want to own the AI landscape — and this is another major step in that direction.

To bring this vision to life, Microsoft introduced several foundational components and standards. None of them are new on their own, but combining them under a unified vision — and rallying other companies around them — is what could make the difference.

First up is the Model Context Protocol (MCP). Think of it as the USB-C of agent-to-platform communication. It lets agents interact with other systems and share data — for example, allowing my agent to place an order on Amazon. I did a deep dive on MCP not long ago, so check that out if you want more detail. Microsoft’s support here is a big leap forward.

The second is the Agent-to-Agent Protocol. This open standard enables agents from different vendors to collaborate — say, a Microsoft-based agent working directly with one from Google or Amazon. The aim is to make it easy for developers like Mindset AI and others in the space to build compatible ecosystems. Again, I covered this in a previous episode.

The important part is that Microsoft is actively supporting these protocols. As we’ve discussed before, protocols only succeed when lots of different players adopt them and bake them into how they build.

On the infrastructure side, Microsoft has rolled out tools to support these standards. They’re upgrading Copilot Studio for multi-agent systems, so people can start thinking in terms of agent teams. They’re launching the Azure AI Foundry — essentially a service for monitoring agents. And they’re introducing something called Agent ID, which acts like identity management for agents. Just like human employees have roles and permissions, Agent ID will handle access control for AI systems.

Together, these tools form the backbone of Microsoft’s vision for the open agentic web. Again, none of this is brand new — we’ve covered these topics — but Microsoft’s support could be the catalyst that brings it all to life.

A major piece of that vision is something called the NL Web Initiative, short for Natural Language Web. This initiative is about making websites more agent-friendly. I talked about this before when we looked at multi-agent systems and the challenges agents face when trying to "go shopping" for us online.

Right now, websites are designed for humans. The NL Web wants sites to expose certain functions and data to agents — through APIs and protocols — so they can do things like buy tickets or access services. If websites aren’t agent-friendly, agents can’t interact with them effectively.

What’s interesting is that companies like Shopify, Snowflake, and even institutions like the Chicago Public Library are already on board. And as I’ve said before, for any of these protocols or standards to work, they require collaboration. Companies have to opt in and build around them.

So that’s the open agentic web — and it’s a fascinating concept.

‍

‍

Google I/O 2025: Consumer AI strategy

Now, turning to the Google I/O conference — their flagship developer event, which happened just days after Microsoft Build — what stood out was how clearly it highlighted the difference in strategies between the two companies. Google is leaning hard into the consumer AI strategy, while Microsoft is doubling down on enterprise AI.

Google’s announcements were heavily focused on consumer applications of AI — AI in Search, Gmail, Photos, Android, even shopping. Many of their developer updates also revolved around consumer-facing tools, like Android and Firebase for app development. And the biggest, splashiest demos? Again, all in the consumer space — think AR glasses that translate speech in real time, or their “magic” photo editor.

Contrast that with Microsoft Build, which was laser-focused on enterprise and business use cases. What’s clear is that while Microsoft has a well-defined vision for enterprise AI, Google still hasn’t settled on what AI means for the open web. Instead, they’re taking an experimental approach — throwing a lot of spaghetti at the wall to see what sticks.

Take Project Astra, for example — it's an initiative to enable AI agents to control computer screens. Sometimes called “computer-use agents,” these systems can interact with a UI like a real user: clicking buttons, filling out forms, all without human assistance. Nothing radically new, but a clear step toward autonomous agents.

There’s also “Agent Mode,” which uses something called the Model Context Protocol (MCP). This lets users issue broader commands like “Plan my weekend trip,” and the AI can then create and execute the plan — booking an Airbnb or grabbing tickets — by accessing third-party services via MCP.

And then there are all the experiments with AI in search and shopping. Google said its AI Overviews — which many of us are already seeing in search — now serve over 1.5 billion people across 200+ countries. That technically makes it one of the most widely used generative AI products in the world. According to Google, AI Overviews have increased overall search usage by 10%.

That’s a big part of Google’s narrative to Wall Street: AI isn’t going to cannibalize their core search business — it’s going to expand it. More searches. More ad revenue. Faster, better, bigger.

They’ve even launched direct-to-AI products. One notable example: a £200/month license for access to NotebookLM, powered by their top-tier Gemini model. That’s a bold move — and one of the few examples of a Big Tech company selling AI as a clearly separate line item, rather than hiding it inside existing products or business units.

They’re also testing new AI-powered shopping tools — like a conversational shopping assistant to help refine product searches, and a virtual try-on tool where you can upload a photo and see how clothes might look on you. Again, nothing brand new, but it signals that Google is aggressively experimenting to find what sticks in this emerging “agentic” web.

They’re even getting back into consumer hardware. At I/O, they previewed Android XR glasses (developed with Samsung) and teased Project Moohan, an AR/VR headset they’re developing in partnership with Warby Parker. Hopefully they’ve learned something from the Google Glass days — mainly that they’re not great at building hardware alone — and are now taking a page out of Meta’s playbook by partnering with a proven brand, like Meta did with Ray-Ban.

If you’ve been listening to previous episodes, you might notice a trend here — I’m hedging my bets. But I do wonder: could the smartphone eventually be replaced entirely by some kind of AI-powered glasses? Maybe not soon, but it’s hard to shake the feeling that the phone won’t be our main consumer device forever.

‍

‍

Closing thoughts

So, to wrap things up — it’s clear that Google and Microsoft are taking very different approaches, and you can see that reflected in the types of announcements they’ve made. Sure, there’s some overlap when it comes to developer tools and a few other areas. These are two massive companies launching a huge volume of new things. But overall, there’s a pretty stark divide.

As for who’s going to “win”? Who’s doing it best? Honestly, it’s too early to tell. But — and I hate to admit this, because I’ve never been a huge Microsoft fan — I do think they’re making huge strides in the enterprise space. And that’s a fascinating area to watch.

Anyway, that’s it for today. Hope you enjoyed the episode — and I’ll catch you next week.

‍