• Everyday AI
  • Posts
  • Ep 791: Microsoft Build Recap: 4 New AI Features That Stood Out

Ep 791: Microsoft Build Recap: 4 New AI Features That Stood Out

Google drops impressive new Gemma 4, Meta hits new AI snags, AI CEOs push Congress on DNA concerns and more

Author

Everyday AI
June 04, 2026

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: Microsoft Build brought a wave of AI updates, from autonomous agents to new models and hardware. Here are the biggest announcements that stood out. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Google drops impressive new Gemma 4, Meta hits new AI snags, AI CEOs push Congress on DNA concerns and more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: IBM and Google Cloud are expanding their enterprise AI partnership, AI is now the leading reason cited for job cuts, and Google DeepMind launched a new Gemma model built to run multimodal AI on laptops. And more. Read on for Byte Sized News.

💪 Leverage AI: Microsoft spent Build pushing AI deeper into the workplace, from autonomous agents and in-house models to new AI-powered devices. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: Microsoft unveiled a new AI-focused operating system for hardware, Meta selling business AI agents and Google tests AI search opt-out for publishers and more. Check it here!

Ep 791: Microsoft Build Recap: 4 New AI Features That Stood Out


Did you miss everything Microsoft announced at its Build conference? 😮

We didn't.

From autonomous agents in Copilot to new models and agent-first hardware, Microsoft went full stack with its lineup.

We break down our four biggest takeaways and give you the implications.

Also on the pod today:

• Copilot Super App teased for summer 🛠️
• Autopilots: always-on AI agents 🤖 
• Scout: proactive Microsoft 365 assistant 📅 

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight –  Empromptu AI builds AI apps that turn real usage into custom models, Keen Code is A CLI-based minimal coding agent with lean context management and skill-driven MCP support, AppWizzy Builds Scalable Apps & Websites with AI

Google Water Saving — Google is pledging to replenish more water than it uses in its data centers by 2030 and has launched new projects across seven states to help protect local water supplies.

Anthropic Cybersecurity — Attackers are getting way more dangerous as they use AI deeper into cyberattacks, and current security frameworks aren't keeping up.

Pinterest and AWS — Pinterest is making its biggest infrastructure move yet, teaming up with AWS to supercharge AI for 600 million users

OpenAI Research — The U.S. is mapping out a plan to strengthen federal oversight for advanced AI, tapping into state laws and new White House directives.

Reve 2.0 Release — Reve 2.0 just dropped, letting you generate and edit 4K images with code-level precision.

Ideogram 4.0 Release — Ideogram 4.0 just dropped as the top open-weight text-to-image model with native 2K resolution and killer text rendering.

Anthropic Statistics — Anthropic shares how they automated 95% of business analytics with Claude, revealing hard-earned lessons for making AI-powered self-serve insights accurate and reliable.

Generalist AI Valuation — Generalist AI just scored a $2 billion valuation after raising $400 million, according to Bloomberg.

Alnylam Drug Discovery — Alnylam and Inceptive just signed a deal worth up to $2 billion to use AI for drug discovery. See how this partnership could accelerate new medicines.

1. IBM and Google Cloud launch new AI consulting practice for enterprise Gemini deployments 🤝

IBM and Google Cloud announced today a new global Google Cloud Practice that will bring thousands of IBM consultants into the push to move enterprise AI from experiments into production.

The partnership expands IBM Consulting Advantage with industry-specific agents optimized for Gemini Enterprise, targeting sectors like banking, government, retail, telecom, energy, insurance, and life sciences. In plain terms, the companies are trying to make it easier for big organizations to modernize old systems, connect messy data, strengthen cybersecurity, and run AI tools safely across hybrid cloud environments.

2. AI is now the top cited reason for May job cuts ✂️

According to Challenger, Gray & Christmas, U.S. employers announced 97,006 job cuts in May, with tech leading the damage at 38,242 cuts and AI cited as the top reason behind layoffs.

The shift is hitting white-collar tech roles especially hard, as companies including Meta, Cisco, Cloudflare and Coinbase reshape teams around AI while tech sector cuts this year have climbed 66% from last year.

3. Google DeepMind launches Gemma 4 12B for laptop AI with native audio 🎧

Google DeepMind today introduced Gemma 4 12B, a mid-sized open model designed to run advanced text, image, and audio tasks directly on consumer laptops with 16GB of memory.


The big shift is efficiency: instead of relying on separate systems to handle vision and audio, the model feeds those inputs straight into its core language system, which should cut memory use and speed up local performance.

4. Meta’s AI API release slips with no firm launch date 🚀

According to The Wall Street Journal, Meta has repeatedly delayed the developer release of its newest AI model API, despite telling developers nearly two months ago that it was coming “soon.”


Meta now says it is testing the tool with partners and plans to release it this month, but the shifting timeline puts pressure on its effort to turn huge AI spending into real revenue.

5. AI CEOs Altman, Hassabis, Amodei push Congress on gene-synthesis safeguards 🫸

OpenAI’s Sam Altman, Google DeepMind’s Demis Hassabis, Anthropic’s Dario Amodei, and Microsoft AI’s Mustafa Suleyman are urging Congress to require DNA and RNA synthesis companies to screen customers and orders.


The call comes as faster AI tools make it easier to design or hide dangerous biological material, raising concern that barriers to creating bioweapons are weakening.

6. Broadcom tumbles after AI chip forecast disappoints 📉

Broadcom shares fell more than 15% Thursday, their worst one-day drop since January 2025, after investors zeroed in on an AI chip sales forecast that came in below Wall Street’s high expectations.


The company beat overall revenue and earnings estimates and guided above consensus for the current quarter, but projected $16 billion in third-quarter AI chip sales versus analysts’ $17.2 billion target.

Microsoft doesn’t need to win the AI race.

It just needs to be unavoidable.

That’s the real story after Microsoft’s latest AI-filled Build developer conference. 

Copilot might not be your favorite AI tool. Microsoft’s MAI models might not catch the frontier tomorrow. Some of the new hardware may still feel early, expensive, or weird.

Doesn’t matter as much as people think.

Microsoft is playing the enterprise game that actually scares competitors: agents where work happens, models it can control, local compute it can bring closer to sensitive data, and devices that push AI past the laptop.

That’s what we tackled on today’s Everyday AI. The four big Build highlights weren’t random announcements. They were Microsoft showing the shape of an AI strategy that can win by being everywhere work already lives.

1. Make Scout earn trust 🔥

Autopilots are Microsoft’s move from reactive Copilot chats to AI workers that can keep running in the background with their own identity, permissions, and access across work systems.

Scout is the first one, and the business utility is painfully obvious: meeting prep, scheduling conflicts, follow-ups, reminders, handoffs, and coordination across Teams, Outlook, OneDrive, SharePoint, calendars, files, local resources, and the browser.

That’s not fluffy productivity glitter. That’s the messy middle where managers burn hours, decisions disappear, and accountability gets buried under 47 Teams threads.

But autonomous work needs guardrails before it needs hype. One percent agent drift sounds tiny until it quietly compounds inside approvals, customer follow-ups, pricing research, or exec prep.

Try This:

Pick one cross-functional workflow where work routinely dies between meetings. Map the systems Scout would need to touch, the actions it could take alone, the approvals that still require a human, and the point where drift becomes unacceptable.

2. Stop grading only models ⚡

Microsoft now has seven in-house MAI models across reasoning, coding, image, transcription, and voice, with MAI-Thinking-1 positioned as its first reasoning model and MAI-Code-1-Flash aimed at coding work.

No, that doesn’t instantly make Microsoft a top AI lab. The sharper read is cost, control, ecosystem leverage, and whether Microsoft can make “good enough” models feel obvious inside the stack companies already pay for.

That’s the part leaders miss when they obsess over model leaderboards. If the model is six to nine months behind but cheaper, governable, integrated, and good enough for casual enterprise users, the harness around the model starts mattering more than the trophy case.

Try This:

Run a bake-off on three boring but expensive workflows: long-document reasoning, coding support, and multimodal content work. Score accuracy, cost, latency, governance fit, and employee adoption friction, because “best model” is useless if it can’t survive procurement, security, or Monday morning.

3. Move edge AI past excuses 🚀

The NVIDIA partnership, RTX Spark laptops, and Surface RTX Spark Dev Box are less about shiny machines and more about Microsoft planting a flag around local inference.

Pair that with smaller local models, and the old excuse starts cracking. Sensitive data can’t leave the environment? Fine. Then the next strategic question is which workflows should move closer to the machine, the edge, or the employee.

That’s where this gets real for regulated, data-heavy, or latency-sensitive teams. Microsoft isn’t just chasing model glory in the cloud. It’s building toward a world where proactive agents can run closer to the work without making every sensitive workflow feel like a compliance bonfire.

Try This:

Ask IT and security for the top three AI use cases currently blocked by data sensitivity, latency, or cloud cost. Split them into cloud-safe, private-environment-only, and local-first candidates, then stop treating all AI workloads like they need the same architecture.

4. Design devices before surveillance wins 🔥

Project Solara is the wildcard because it points beyond apps, tabs, and chat windows.

The idea is agent-first devices: badges, desk devices, frontline tools, and hardware that lets AI show up where laptops are awkward and chatbots are basically useless.

That matters for healthcare, retail, field service, hospitality, industrial work, logistics, and any environment where the job happens on the floor instead of inside a browser. The upside is huge if agents can help overloaded teams move faster without forcing everyone to become a keyboard jockey.

The risk is also obvious. A badge with a camera, mic, identity, transcription, and agent access can improve work or become surveillance with better branding.

Try This:

Pick one frontline role where AI chat doesn’t fit the job. Document the workflow, the environment, the data being captured, the privacy red lines, and the exact moment an agent-first device would reduce friction without creeping people out.

Microsoft’s Build strategy wasn’t about one knockout AI announcement.

It was about coverage.

Agents. Models. Local compute. Devices. All pointing in the same direction.

That should make every enterprise leader ask a much sharper question: are we building an AI strategy around tools, or around where work actually happens?

Reply

or to participate.