Purpose Before Prompts

As we near the three-year anniversary of ChatGPT’s release, it’s time to return to first principles.

The tool is not the point. The work is the point. And in the justice system, excellence is not aspirational. It is the difference between a decision that people can accept even when they lose, and a decision that ends the lawsuit but fails to resolve the problem. When we compromise excellence, we risk compromising lives.

I do not fault the technology. The trouble begins when process crowds out purpose and we start with the prompt instead of the principle. That shift pushes us to value speed over judgment and efficiency over fairness. It tempts us to automate steps without asking what those steps were protecting. The harm is not abstract. It shows up in a missed fact, a bad citation, or a summary that flattens what matters, and real people live with that result.

This is how I think about GenAI in chambers. First decide what must remain human, which steps require independent judgment, and what cannot be delegated because it carries the weight of the robe.

In practice, that means no AI assistance begins until a human has reviewed the filings and the record. Only then do we invite the tool to help by drafting a neutral bench memo, summarizing transcripts, building a timeline of key dates for a human to confirm, verifying authorities, and checking page limits and deadlines so human time goes to the merits.

Speed is not the standard. Accuracy and fairness are. The questions that matter are whether the decision rests on the record and the law, whether the parties felt heard, and whether the community can rely on the process. If a tool helps us meet those standards, we use it. If it tempts us to skip the hard part, we pass.

Purpose before prompts. Standards before shortcuts. People before tools.​​​​​​​​​​​​​​​​

Happy Thanksgiving everyone!