Key Takeaways

  • AI dramatically increases individual output, but many founders fall short by not seeing automation through to completion.
  • Anthropic's Head of Product for Claude Code, Cat Wu, says achieving 90-95% accuracy with AI automation is “not really an automation” because the remaining manual intervention negates any efficiency gain.
  • The real value in AI comes from committing to the "last 5 to 10%" of effort required to make an automation 100% reliable, even if it feels slower initially.
  • This means investing the time to iteratively teach AI tools your specific preferences and giving continuous feedback until they perform perfectly.
  • This commitment to total reliability is the core of Cat Wu's powerful "100% Automation Rule."

The 100% Automation Rule

Cat Wu warns against the seductive trap of partial automation. Many founders get AI tools to a "good enough" state—say, 90% or 95% accurate—and then stop. But, as Wu explains on Lenny's Podcast, that remaining 5-10% of manual cleanup makes the "automation" nearly worthless. It just shifts the work, often creating new cognitive load.

The real breakthrough comes from pushing through that final, often tedious, phase of refinement. This is where the 100% Automation Rule comes in:

Rule: If an automation doesn't work 100% of the time, it's not really an automation.

Action: Put in the elbow grease to teach Claude your preferences to like give it feedback so that it can improve its skill so that it can get to that 100%.

Context/Warning: Building the automation is often a lot slower than you doing it yourself. You have to know to define a skill. You have to know to like use this skill and give it feedback. And then you have to know to tell co-work to update the skill based on all the feedback that you gave. And then you also have to know where to read the skill to like make sure that the feedback was incorporated the way that you want.

When This Works (and When It Doesn't)

This rule applies perfectly to any manual, repetitive task you find yourself doing multiple times. Cat Wu argues that by committing to 100% automation, you can truly offload 'grunt work' and free up your time for creative, high-impact activities that only a human can do. It works best when you invest the initial effort to thoroughly train the AI and refine the automation process, understanding that the upfront cost of time will pay dividends.

However, this approach isn't a silver bullet for every task. It shines for high-volume, highly structured tasks with predictable inputs—think data synthesis, report generation from templates, or initial drafts of communications following a clear format. It falters when tasks demand subjective judgment, deep creative insight without specific guardrails, or when the input data changes too rapidly or unpredictably. Don't try to automate complex strategic decisions; focus on the repetitive actions that drain your clock every week.

What to Do With This

Stop admiring the concept of AI automation and start applying the 100% rule this week. First, pick one recurring manual task that consumes 1-2 hours of your time. Maybe it's drafting weekly investor updates, summarizing internal meeting notes into a standardized format, or generating a specific type of market research report.

Now, apply the 100% Automation Rule. Recognize that if your AI drafts a 95% complete investor update, you're still spending time editing, verifying facts, and polishing tone. That's not automation; it's a co-pilot that still needs constant supervision.

Next, take action. Use a tool like Anthropic's Claude Code or similar AI to perform that task. After the first draft, provide highly specific feedback. Don't just say "make it better." Instead, try: "Always include a paragraph on key hiring updates here," or "Shorten the intro and make sure to bold the ARR numbers for quick scanning." If using co-work, define a 'skill' for this specific task (e.g., 'Draft Investor Update'). Give it feedback, then ensure co-work updates that skill. If you can read the updated skill definition, do so to ensure the AI learned correctly.

Cat Wu cautions that the initial setup will take longer than doing the task yourself. That's the "elbow grease" she talks about. Don't quit after the first few tries. Push through. Keep iterating and refining until the AI delivers a draft that needs zero manual edits. Only then have you truly automated the task and freed up your valuable time for creative, founder-level work.