Learning Brief — 2026-06-02

What we covered

• AI news: Microsoft's AI Infrastructure Push and the Cost Reality Check
• PM news: AI's Multiplayer Moment: Why the Next Wave Isn't About Individual Productivity
• PM learning: From 10x people to 10x teams: The multiplayer shift in AI

Mental model

Shift from optimizing individual productivity with AI to redesigning workflows so AI becomes the connective tissue that compresses friction between people.

Summary

Microsoft announced a new reasoning model and released Adaptive Spec-driven Scoring, an open source framework that lets developers spin up AI evaluations and behavior tests using plain text descriptions instead of hand-coded test suites. This cuts the friction on a major pain point — most teams today are still manually defining how to measure AI quality. Perplexity AI unveiled a hybrid local-cloud inference orchestrator that dynamically routes AI workloads between a user's device and cloud servers in real time, optimizing for latency and cost mid-task. This is the first production system that actually solves the split-inference problem that's been theoretically interesting for two years. Uber capped employee AI spending after burning through its budget in just four months, despite encouraging unlimited usage. This is a real signal: even well-resourced companies are hitting cost walls faster than expected when AI tooling becomes frictionless.

So here's what's shifting in how teams are actually using AI, and it matters for how you're thinking about your roadmap right now. Ravi Mehta just laid out something pretty important: we're moving from the first wave of AI—which was all about making individual contributors faster—to a second wave that's fundamentally about multiplayer workflows. Single-player AI was Copilot writing code faster, or ChatGPT drafting your emails. That's real, but it's also kind of a local optimization. The next wave is different.

The multiplayer framing changes what you should be building for. Instead of asking "how do we make our users 10x faster individually," you're asking "how do we make teams 10x more coordinated?" That's a completely different product problem. It means thinking about handoffs, context sharing, async collaboration, and how AI agents can actually reduce friction between roles—not just within them.

For a senior PM, this is the inflection point where your competitive advantage shifts. If you're still designing features around individual productivity gains, you're optimizing for yesterday's AI narrative. The teams winning right now are the ones building for coordination breakdowns. How does your product help a designer brief an engineer? How does it let a PM stay aligned with their squad without constant sync meetings? How do you use AI to compress the information loss that happens between departments?

This also reframes the AI arms race you're probably feeling pressure to join. It's not about having the fanciest model or the most features. It's about understanding where your users actually lose time together—and building AI workflows that fix that specific friction point. That's a much harder problem to solve, but it's also the one that actually sticks.

If you're building a product where teams collaborate, your next roadmap conversation should probably start here.

Here's the thing that's going to reshape how you think about AI adoption in your org: the era of AI making individual contributors ten times more productive is ending. We're moving into the era of AI making teams ten times more effective. And that's a completely different problem to solve.

The first wave of AI tools — Copilot, ChatGPT for individual use cases — they optimized for solo work. A designer generates mockups faster. An engineer writes code quicker. A PM drafts strategy documents in half the time. All real productivity gains, but they're additive. You're still one person doing one person's work, just faster.

But here's where it gets interesting for you as a leader: the multiplayer wave changes the unit of analysis. Now AI is a medium for coordination, not just execution. It's how teams compress the friction of async work, how you run discovery at scale without drowning in research debt, how you align five stakeholders on strategy without seventeen meetings.

What that means in practice is this — your job shifts from "how do I get my team using AI tools" to "how do I restructure workflows so AI becomes the connective tissue between people?" That's a product design problem, not a tool adoption problem.

Think about it like this: if you're a PM and you use AI to write better specs faster, you've won maybe ten percent of the game. But if you use AI to turn your spec into an interactive prototype that your eng team can iterate on in parallel, that your design team can extract component requirements from, that your marketing team can start planning launch around before code review — now you've fundamentally changed how work flows through the organization.

The move here is to audit one workflow on your team this week. Pick something that currently requires sequential handoffs — discovery to strategy to design to build, whatever your pattern is. Map where the friction actually lives. It's usually in translation. Someone writes something, someone else interprets it, context gets lost, you loop back. That's where multiplayer AI lives. Not in making one person faster, but in making the handoff instantaneous and lossless.

Start there. Don't think about "AI tools for my team." Think about "where does my team lose velocity to communication friction?" That's your lever.