Most advice is a Rorschach test - it means different things to different people. Take the classic: “Do one thing well.”

This is clear if you’re building a payments API or email delivery. But for in-house patent attorneys? It’s messy.

Their job isn’t one thing - it’s ten: prefiling, prosecution, licensing, divestiture, strategy, and more. Each could be its own startup. So when we build product for them, the question becomes: at what layer do you “do one thing well”?
Patent workflows are bespoke. Everyone uses different pieces. Some features are daily-use. Others, once a year.

I previously built products focused on maximizing daily use. But in patents, counsel aren’t chasing engagement, but clarity. They need dozens of small, situational insights to stitch together the right strategic call for the company. Something they saw 6 mo ago combined with something from yesterday. The workflow is research, synthesis, and judgment.

So we’re building features that optimize for decision quality.

My favorite example that we shipped this month: reports. Ask a question via natural language over your portfolio, and the AI compiles data from the right tables into an answer.

It’s not something people use every day; you may only have a specific question once a year. But, if it can answer that question, it can impact the decisions you make for the next few months or more.

Turns out "doing one thing well" for bespoke workflows means building the full toolkit. Because the one thing is the decision, and good decisions need all the pieces.

The advice is great. We just need to squint at it in our own way.