Learning Brief — 2026-06-24

What we covered

• AI news: The Token Rationing Era: How AI Cost Management Is Reshaping Enterprise Product Strategy
• PM news: A 9-Person Team Now Ships What Used to Take 90: Inside Laurel's AI-Native Playbook
• PM learning: How to Build a Company OS in Claude Code: Shipping 10x Faster with AI-Native Teams

Mental model

When you build around AI capabilities, your bottleneck moves from execution to clarity—optimise for decision velocity and intent-setting, not role coverage.

Summary

Companies are now implementing token budgets and rationing systems to prevent employees from exhausting API spend on low-value tasks. What started as unlimited experimentation with AI tools has shifted to cost controls and governance frameworks as organizations face real bill shock from production deployments.

Laurel just published a breakdown of how they built a company operating system using Claude Code—and the numbers are worth paying attention to. A team of nine is now shipping work that previously required ninety people. That's not a typo, and it's not theoretical. This is a real product team documenting how they actually operate.

Here's the PM angle: this isn't about replacing engineers with AI. It's about fundamentally restructuring how a team coordinates, decides, and ships. Jiaona Zhang, their Chief Product Officer, is essentially showing us what happens when you stop thinking about AI as a tool and start thinking about it as a change to your operating model itself. The question isn't "how do we use Claude to code faster?" It's "what does our org chart look like if we assume AI handles the work that used to require multiple people?"

For you as a senior PM, this matters because it's forcing a real reckoning. If your competitors—especially in spaces where execution velocity is the moat—are operating at this kind of leverage, you need to understand what they're actually doing differently. It's not just about hiring fewer people. It's about how decisions flow, how you spec work, how you handle quality gates, how you think about product iteration cycles.

The practical implication: if you're building a roadmap for the next eighteen months, you need to be asking whether your team structure and decision-making processes are designed for this new reality. Are you still organizing around the assumption that engineering capacity is scarce? Because if your competitors aren't, you're already behind.

This matters to your product strategy because it changes what's actually possible to build, and how fast.

Here's the thing that should change how you think about team composition and velocity: a nine-person team at Laurel is now shipping what used to require ninety people. That's not a marginal improvement. That's a structural shift in how work gets done, and if you're targeting a Senior PM or GPM role, understanding the mental model behind it matters more than the AI tool itself.

The core insight is this: when you build a company OS around AI capabilities, you're not just automating tasks. You're fundamentally changing what your team needs to be good at. You're collapsing layers of work that used to require specialisation, handoffs, and coordination. What that means in practice is that your bottleneck moves from execution to clarity.

In a traditional org, you might have product managers writing specs, engineers building, QA testing, and ops coordinating. Each handoff introduces lag and translation loss. But when your OS is built around Claude Code or similar AI systems, the game changes. One person can now articulate intent, iterate design, write and audit code, and validate outcomes in a single flow. The team doesn't get smaller because you're lazy—it gets smaller because you've eliminated the friction points that made specialisation necessary in the first place.

The move here is to think about your team not as a collection of roles, but as a collection of decision gates. What decisions actually require human judgment? What requires domain expertise that can't be augmented? Everything else becomes a candidate for AI-native workflows. At Laurel, they clearly identified that their core value came from product intuition and customer understanding, not from the mechanics of execution.

This has a direct implication for how you staff and structure teams going forward. If you're hiring for a Senior PM role, you're not hiring for someone who can manage nine people doing traditional work. You're hiring for someone who can set clear intent, make fast calls on what matters, and know when to push back on the AI's output. That's a different skill set. It's closer to a founder than a traditional line manager.

The concrete action this week: map your current team's work into two buckets—decision-making and execution. For everything in the execution bucket, ask: could this be faster and cheaper if we built a Claude Code workflow around it? You might not rebuild your entire OS tomorrow, but identifying one or two high-friction handoffs and prototyping an AI-native alternative will teach you whether this model actually works for your product and team.