What Anthropic's Product Team Does Differently (And What Small Business Owners Can Steal From It)
By Warren Schuitema, Founder | Matchless Marketing | The AI Dad
Anthropic's head of product for Claude Code, Cat Wu, sat down with Lenny Rachitsky recently and walked through how her team ships faster than almost anyone else in tech. The interview is aimed at product managers at software companies. But I read it and kept thinking: this is actually a masterclass for small business owners trying to get more out of AI without burning out trying.
Here's what stood out, and what I think it means for you.
They Build With the Tool, Not Just For the Tool
One of the most striking things Wu described is how Anthropic's product team actually uses Claude Code to do their own work. They're not just specifying what the AI should do — they're in the tool daily, running into the same friction their users run into, and feeding that back into development.
This sounds obvious. It isn't.
Most small business owners I talk to treat AI tools like a vending machine. You put in a prompt, you get an output, you move on. What Wu's team does is different. They're in a constant feedback loop with the tool — noticing where it breaks, where it surprises them, where it needs a better workflow around it.
For you, this means: stop treating ChatGPT or Claude as a one-and-done machine. Start treating it as a collaborator you're getting to know. The more you use it for real work, the better you'll get at prompting it. The better you'll understand where to trust it and where to double-check it.
That's not a soft insight. That's a practical discipline that separates people who use AI well from people who use it occasionally and wonder why they're not seeing results.
Speed Comes From Eliminating Decision Debt
Wu talks about how fast shipping at Anthropic isn't about working longer hours. It's about removing the drag that slows teams down. Unclear ownership, ambiguous priorities, decisions that never get made.
She calls it "decision debt." Every choice you don't make today becomes friction you have to fight through tomorrow.
I see this constantly with solo operators. You've got five AI tools open, three workflows half-built, and a list of things you meant to automate six weeks ago. Nothing moves fast because nothing is decided.
Pick one workflow to finish this week. One. Not the most exciting one — the one that's half-done and costing you time right now. Finish it. Ship it. Then pick the next one.
Claude is actually useful here. Paste your list of unfinished projects into Claude and ask it to help you identify which one is closest to done and which one is creating the most daily drag. Let it think through the triage with you. That's five minutes well spent.
Small Teams Win By Being Specific About What They Won't Do
Something Wu emphasized that most people miss: Anthropic's product team is disciplined about scope. They decide early what's out of scope for a given build, and they hold that line. That's what lets them move fast on what's in scope.
This is directly relevant to how you use AI in your business.
If you try to automate everything at once, you'll automate nothing well. The businesses I see actually getting time back from AI are the ones who got ruthlessly specific. They picked one repetitive task, built a clean workflow for it, and got it running reliably before touching anything else.
For a coach or consultant, that might be client onboarding emails. Build that one sequence in Claude or ChatGPT, test it with a few clients, refine it. Don't touch lead follow-up or content creation until onboarding is solid.
For a retailer or service business, it might be responding to frequently asked questions. Build a simple response library in Claude. Test it. Own it. Then move.
Scope discipline isn't just for product teams. It's the thing that separates the people who have working AI systems from the people who have a lot of browser tabs open.
The Insight That Actually Matters for You
Here's my honest read on why this interview is worth your attention even though it's aimed at Silicon Valley product managers.
The operating principles Wu describes — use the tool yourself, make decisions fast, be ruthless about scope — are the same principles that make any small business owner's AI adoption actually work.
Most AI content you'll read tells you about features. What model just dropped. What the new API can do. This is about how the people building AI actually work, and it maps almost exactly onto what I watch working (and not working) in the businesses I talk to every day.
The gap between small business owners who are saving 10 hours a week with AI and those who are still experimenting is almost never the tools. It's the operating habits around the tools.
One Thing You Can Do Today
Pull up Claude or ChatGPT right now. Paste in the last thing you worked on manually that you've been meaning to automate. Ask Claude: "What's the simplest version of this I could automate today? What information would you need from me to help me build it?"
Don't plan a full automation sprint. Just have the conversation. See what comes back.
That's how Wu's team moves fast. They're in the tool, running real problems through it, learning as they go. You can do the same thing in the next 15 minutes.
The tool's already on your computer. The only thing left is to use it like you mean it.