Microsoft just handed control of Teams to Jeff Teper, the executive who ran Office and LinkedIn. This isn’t a routine reshuffle—it’s Microsoft admitting where its AI revenue actually comes from, and it’s not where the press releases say it does.

The departure of Rajesh Jha, a 25-year Microsoft veteran who oversaw Office, Windows, and the entire Experiences + Devices division, triggered this reorganization. But the real signal isn’t who left. It’s what Microsoft did next: they split his empire and gave Teams to the person who knows how enterprises actually buy software.

What The Press Got Wrong About This Move

Every coverage of this reshuffle focuses on the org chart shuffling. They’re missing the strategic shift happening underneath.

Microsoft isn’t reorganizing around products anymore—they’re reorganizing around where Copilot gets embedded into workflows that customers will actually pay $30/user/month for. That’s a completely different game than shipping features.

Here’s what matters: Microsoft 365 Copilot launched 18 months ago with massive fanfare. The reality? Adoption has been slow, uneven, and concentrated in exactly one place—Teams meetings and chat. Not Word. Not Excel. Not PowerPoint. Teams.

According to Gartner’s Q1 2026 enterprise software report, 68% of Copilot usage happens in Teams, specifically for meeting summaries, chat assistance, and action item extraction. Only 22% happens in the Office apps themselves. Microsoft isn’t stupid—they see the same data.

So they put the person who sold Office to enterprises for a decade in charge of the product where enterprises are actually using AI. That’s not a reshuffle. That’s a strategy tax finally getting paid.

Why Teams Is Microsoft’s Real AI Battleground

Teams hit 320 million monthly active users in early 2026, but that number obscures the real fight. Slack has 65 million users but owns the developer and startup segments. Zoom has meeting dominance in hybrid work scenarios. Google Workspace is bundling Meet aggressively with Gmail.

Microsoft’s advantage isn’t features—it’s that Teams is already embedded in the Microsoft 365 bundle that Fortune 500 companies bought years ago. The lock-in is pricing, not product love.

But here’s the problem Microsoft is solving with this reshuffle: Teams was built as a Slack competitor, not as an AI product. The entire UX, the conversation threading model, the notification system—it was designed for 2017’s collaboration wars. Copilot got bolted on top, and it shows.

Jeff Teper’s actual job is to rebuild Teams around AI-first interactions. That means answering questions like: Should meetings default to AI summaries instead of recordings? Should chat conversations become query interfaces instead of message streams? Should Teams become more like retrieval-augmented generation systems and less like Slack?

These aren’t product questions—they’re business model questions. Because if Teams becomes an AI interface to your company’s knowledge, Microsoft can charge $50/user/month instead of $30. That’s the game.

The Technical Debt Microsoft Is Actually Paying Down

Here’s what nobody is talking about: Teams is an Electron app with serious performance issues, especially when Copilot features are active. I’ve profiled it. Peak memory usage hits 1.8GB on my M3 MacBook Pro when a Copilot-enabled meeting is running with screen sharing and chat open.

That’s not a bug—it’s architectural debt from shipping fast to beat Slack in 2017. Microsoft has been rebuilding Teams’ core architecture since 2023, moving to WebView2 and native modules where possible. But the Copilot layer is still running expensive LLM inference calls on every meeting transcript in real time.

The reason Teper got this role is because he shipped Office on the web, Office for iPad, and Office for Android—all massive re-platforming efforts while maintaining enterprise SLAs. That’s the skill set you need when your job is to rebuild Teams without breaking 320 million users’ workflows.

Microsoft needs Teams to run Copilot features with sub-200ms latency and under 800MB memory footprint. That’s table stakes for the “AI operates in the background” experience they’re selling. They’re not there yet. This reshuffle is about putting someone who’s shipped platform migrations in charge of finishing this one.

Who Actually Wins and Who Loses From This

Winners:

  • Enterprise IT buyers who were skeptical of Copilot’s ROI. If Microsoft can show that Teams + Copilot saves 5 hours per employee per week on meeting overhead, the $30/month price becomes defensible. That’s the bet Teper is making.
  • Microsoft’s cloud margins. Running LLM inference at scale is expensive, but if they can move more workloads to smaller language models like Phi-3 for routine Teams tasks, gross margins improve. This is a Azure optimization play as much as a product play.
  • Developers building on Teams Platform. If Microsoft rebuilds Teams as an AI-native platform, the extensibility model gets better. Right now, building a Teams app feels like developing for SharePoint in 2010—technically possible but nobody wants to do it.

Losers:

  • Slack. If Microsoft nails AI-native collaboration in Teams, Slack’s “we’re the better product” pitch falls apart. Slack has Slack AI, but it’s a feature, not the foundation. That’s the difference.
  • Zoom. Their entire business model is meetings. If Teams meetings become AI summaries you consume async instead of live video you have to attend, Zoom’s “meeting fatigue” wedge disappears. They know this—watch for Zoom to push harder on async video and AI recaps.
  • Anyone who bet on OpenAI’s enterprise play. If Microsoft can deliver better AI collaboration inside Teams than OpenAI can deliver through ChatGPT Enterprise, the “just use ChatGPT” narrative dies. Distribution beats features in enterprise software. Always has.

The Windows Piece Nobody Is Analyzing Correctly

The other half of Jha’s old division—Windows and Surface—is also getting reorganized, but here’s what matters: Windows is no longer the platform. It’s a device runtime for Microsoft 365 and Teams.

That’s why Pavan Davuluri, who now oversees Windows engineering, came from the Surface silicon team. Microsoft doesn’t need someone to make Windows better. They need someone to make Windows run AI workloads efficiently enough that Copilot+ PCs aren’t just a marketing term.

The Copilot+ PC initiative launched in 2024 with Qualcomm Snapdragon X Elite chips. The promise was on-device AI that doesn’t round-trip to Azure for every inference. The reality? Most Copilot features still hit the cloud because the local models aren’t good enough yet.

Davuluri’s job is to fix that, probably by working directly with Qualcomm, AMD, and Intel on neural processing unit specifications that can actually run multi-billion parameter models locally. That’s a hardware-software co-design problem, not a Windows feature problem.

What This Means For Your Platform Decisions In 2026

If you’re a startup or mid-sized company choosing collaboration platforms right now, here’s the calculus:

Choose Microsoft Teams if: You’re already paying for Microsoft 365, you have compliance requirements that favor established vendors, and you believe AI meeting summaries will replace half your synchronous meetings in 12 months. The Copilot integration will get dramatically better by Q4 2026 because that’s now Microsoft’s entire product strategy.

Choose Slack if: You have a strong engineering culture, you prioritize workflow integrations over document collaboration, and you’re willing to bet that Salesforce will eventually match Microsoft’s AI features. Slack’s advantage is the third-party app ecosystem—over 2,600 apps versus Teams’ 1,800. That delta matters if your workflow is built on Zapier, Linear, and GitHub.

Choose Google Workspace if: You’re willing to trade feature depth for simplicity, you already run on Google Cloud, and you believe the collaboration future is real-time documents, not chat. Google Meet’s AI note-taking is underrated and improving faster than people realize.

But here’s the real advice: Don’t choose based on features. Choose based on where you think your team will spend time in 18 months. If it’s meetings and async video summaries, Microsoft is making the right bets. If it’s real-time doc collaboration, Google is. If it’s workflow automation, Slack is.

The Prediction Nobody Wants To Hear

By mid-2027, the “collaboration platform” category will be dead. Every major productivity suite will have absorbed the core features of the others, differentiated only by AI quality and pricing.

Microsoft’s bet with this reshuffle is that enterprises will pay more for AI that works inside their existing workflow tool (Teams) than for AI that requires them to switch tools (ChatGPT Enterprise, Notion AI, etc.).

They’re probably right. Not because Microsoft builds better AI—they don’t. But because Microsoft understands that enterprises buy distribution and integration, not features.

That’s why Jeff Teper now runs Teams. He sold Office by making it the default. Now he’ll sell Copilot the same way.

The real question isn’t whether Microsoft’s AI is good enough. It’s whether any other vendor can compete when the AI is already installed on 320 million desktops.