“The Classic Product Playbook Is Dead”: What Anthropic’s Head of Product Learned Building Claude

In a wide-ranging interview released this week, Anthropic’s Head of Product sat down to explain how AI has completely shattered traditional product development — and what actually works inside one of the fastest-scaling AI companies on the planet.

The core message was blunt: the old 6-to-12-month planning cycles, rigid PRDs, and endless alignment meetings are obsolete. Speed has gone from months to days, and the entire game has changed.

Speed Is the New Normal

“Roadmaps used to be a one-year plan,” he said. “Now they’re a living system of priorities that can shift in a day.”

At Anthropic, the time from idea to shipped feature has collapsed from 6 months to as little as **one day**. The team no longer waits for perfect spec documents. They ship, measure, and iterate at the speed of the underlying models — which themselves improve every 3–6 months.

Product Taste > Execution

With AI making code nearly free, the real bottleneck has moved. It’s no longer “can we build this?” — it’s “should we build this, and what exactly should it be?”

“The most effective unit now is a single engineer with strong product taste,” the Head of Product explained. One person who deeply understands user needs and can ship fast beats an entire traditional PM + engineering team stuck in sync meetings and alignment theater.

PM and engineer roles are merging. The best product people are now those who can both define the vision *and* execute it instantly with Claude.

Claude Code vs. Cowork: Two Different Flavors of AI Work

Anthropic has split its internal tools into specialized modes:

Claude Code — when the output is code. The desktop version shines here: instant live previews for frontend work, perfect for rapid iteration.
Cowork — when the output is not code. Think Slack threads, emails, research reports, or full presentations. One real example: an employee connected Cowork to their Drive and Slack. Two hours later it had assembled a polished 20-page presentation in Anthropic’s exact brand style.

The future workplace, he believes, won’t be a single AI chat window. It will be a **suite of specialized agents**, each optimized for a different type of work.

Build on the Bleeding Edge — or Fall Behind

A key philosophy at Anthropic: ship products that are “almost broken” today. Capabilities jump so fast that if you wait for the model to be perfect, your competitors will already have distribution, data, and user habits locked in.

“The Classic Product Playbook Is Dead”: What Anthropic’s Head of Product Learned Building Claude “You have to build on the edge of what the model can do right now,” he said, “so you’re ready the moment the next leap arrives.”

95% Automation Is Actually Worse Than Nothing

One of the most important lessons: 95% automation is not automation.

If the AI does 95% of the work but still requires constant human oversight for the last 5%, you haven’t saved time — you’ve created new operational drag.

True automation only happens at 100%. Only then do you actually free up the 20% of time that matters for strategy and creativity.

Customers don’t pay for “almost.”

Evals Are the New Product Superpower

In traditional SaaS you could write acceptance criteria. In AI you can’t.

The Head of Product now spends significant time formalizing:

What does a “good” answer look like?
What are the critical edge cases?
How do we objectively compare v1 vs. v2 of an agent?

Without rigorous evals, you’re not managing quality — you’re praying.

Personality Is Part of the Product

Claude’s success isn’t just about raw intelligence. It’s also about character — how it argues, admits mistakes, and stays helpful without being sycophantic.

A striking stat: 79% of Anthropic’s enterprise customers also pay OpenAI. People aren’t choosing one model. They’re choosing different models for different tasks — and different “personalities.”

The Real Secret: Mission Alignment Removes Friction

How did Anthropic go from roughly $1B to over $14B ARR in less than a year? The answer isn’t just better models.

“The Classic Product Playbook Is Dead”: What Anthropic’s Head of Product Learned Building Claude “80% of energy in big companies goes into politics and alignment,” he said. “Here, mission alignment is so strong that friction almost disappears.”

The internal mantra is simple: “Just do things.”

The cost of a failed experiment has dropped from months of wasted effort to a prototype you can test in an afternoon. The bigger risk is overthinking and delaying.

Build What You Use Yourself

AI demos look magical the first time. Real products are the ones you want to open every single day.

“If you don’t actually want to use the thing you built every day,” he warned, “it’s not a product — it’s a demo.”

The Gap Between Believers and Skeptics Is Widening

“The Classic Product Playbook Is Dead”: What Anthropic’s Head of Product Learned Building Claude Weekly Active Users for Claude Code have doubled organically since January 2026 — with almost no marketing. The people living and breathing AI every day are moving at light speed. Everyone else is still 6–12 months behind in their strategic thinking.

And the biggest winners of all this chaos?

Product managers.

As the Head of Product put it with a grin: “PMs always wanted to ship 100 different features. The only thing stopping them was slow engineering. Now you can build ten products in a couple of days — if you actually know what to build. The hard part was always knowing what to want.”

The classic product development playbook didn’t just evolve.
It broke.

And the new rules are being written in real time — by the teams that ship fastest and think deepest.

Full interview (highly recommended): YouTube — Anthropic Head of Product