Column

Holding Serve With AI

There’s plenty of evidence to suggest that outsourcing your cognitive processes to AI — whether by letting it do your writing or thinking or analyzing for you — is a bad idea. Partly, of course, this is because the AI might do a poor job of it, leaving you on the hook for badly reasoned arguments or banally worded text.

In the bigger picture, though, the worry is that by delegating your thought processes to a machine, your own cognitive muscles will atrophy, and you’ll be less able to apply acute reasoning when you don’t have the AI to rely on. This is especially a concern for newer lawyers, who are still developing their legal and analytical reasoning skills and are most at risk of losing their mental acuity before they’ve really begun to hone it.

I share these concerns. But I also accept that more and more people routinely turn to AI at least for assistance with thinking things through, and often for more. I also know that thinking things through is hard work — often frustrating, frequently draining — and that hard-working lawyers often feel they have only so much intellectual bandwidth available.

So I don’t think it’s realistic simply to tell young lawyers, “Never use AI to help you with your thinking.” The goal should be to help them do it carefully and wisely.

Here’s what I’ve been telling law students and young lawyers in my talks with them recently: Think of AI as a tennis partner, maybe even an instructor or coach. When you hit the ball over the net, they’re going to hit it back to you — often, especially when you’re relatively inexperienced, more skilfully than you did. Watching and learning from them, your return should be a little better than your first try.

Similarly, with AI, your initial attempt to explore an idea or generate some text should include a request to analyze and assess your position. The AI should then produce a response that highlights what’s strong, what’s merely okay, and what needs work. Your subsequent reply should reflect that feedback and improve on your thinking and writing, inviting further volleys from the AI as you strive to bring your idea or your phrasing to a point where you’re satisfied with it.

In this way, working with AI is an iterative process. Each back-and-forth should advance your progress, as you evaluate the AI’s feedback and decide which of it is helpful and which is not. It’s like engaging with a more experienced colleague who can help you get better each time you ask.

But here’s the critical point about seeing AI as a tennis partner: You have to serve. You have to be the one who comes up with the original idea or the initial draft for the AI to consider. If you let the AI serve first, then you’re just letting the instructor do all the work, and you’re not learning anything useful or producing anything original.

Worse, your finished “work” will be indistinguishable from that of any other lawyer who used AI for the hard stuff. The only truly distinctive feature of your legal service offerings in a post-AI world is your own originality, insight, and humanity. Surrendering your individual creative agency to a machine removes all those features and leaves you in a seriously uncompetitive position.

No good tennis instructor would do all the serving, and no player who’s serious about developing their game to its highest level would ask them to. As tempting as it might be to just give in and let the AI take the lead, it’s imperative that you hold fast and push yourself to come up with your own original insights and ideas at the start.

Let the AI be a helpful instructor. Make sure you remain the rising star.

Comments

  1. This is great! I love the simile and as a prof, I would like to have my students use AI in this way. Currently I tell them not to use it at all because, as you say, they need to learn how to do the work. But that means I’m failing to teach them how to use a tool which will be vitally important in their careers.

    It would be great to have a standardized set of instructions that would put AI into “tennis instructor mode.” Students could be authorized to use the AI only in this mode. Maybe this would work as a Project in Claude?

Leave a Reply

(Your email address will not be published or distributed)