There’s a chasm between understanding how to ask the free version of ChatGPT to punch up your emails and building a useful AI workflow. A couple weeks ago, I went to a conference with a panel titled “The Attorney’s Guide to AI” that was supposed to open lawyers’ eyes to building skills and deploying agents, but had to pivot to “so, you type in ‘c-l-a-u-d-e-DOT-a-i’ and enroll in a paid account.” For a change, this isn’t just a case of lawyers being hopelessly behind the curve on technology. The AI industry that needs to produce $6 trillion in revenue by 2030 to stay afloat only musters about $200 billion right now, mostly because the 98 percent of the public doesn’t see anything about AI to justify paying more than… nothing.
But for lawyers who understand how to elevate AI from the free sample platter to something that can provide actual value, they need time to experiment. The message from that same panel mentioned above was that every lawyer needs to start playing around, building skills, running tests, and figuring out what produces quality output.
Ropes & Gray understands this, and will now let all of its associates count up to 100 hours of AI experimentation toward their billable targets. It expands last year’s pilot, which allowed first-years to put up to 400 of their 1,900 hours toward playing with the robots.
This is the right call whether you expect generative AI to become the future of the profession or an ultimately minor productivity lift. Whatever the merits of the technology, moving into the world of skills, agents, and vibecoding is disorienting and the only cure is setting aside time to experiment with it. Whatever the plan, it’s going to fail in some spectacularly unexpected way on the first attempt. But if users keep iterating with the tool, it really does get better.
Just like a first-year associate.
One Ropes associate told Reuters, “you’re investing a lot of time in it and you’re not seeing a ton of efficiencies at first.” That might be an understatement. Horror stories of hallucinations and sloppy drafting tend to arise from practitioners trying to use cheap AI as an “Easy” button. Vendors building professional legal tools have done a lot of work to tailor it to a lawyer’s workflow. And it really is about “workflow.” Even if the technology provides accurate results out of the box, that doesn’t matter much if the way it does this robs the lawyer of meaningful judgment. The human brain can’t move faster and if the AI process isn’t giving the lawyer the necessary time to get the optimal as opposed to bare minimum result, that’s a workflow problem better addressed before it becomes the client’s problem. With individual tweaking, it can get even better.
As long as attorneys don’t become stereotypical “screamer” partners.
This policy generated some snide criticism, with many tying the move to recent reports that the big AI providers — especially Anthropic — have leaned on religious leaders to accept their belief that AI may be a conscious life form. Which is of course goofy.
That said, people who thrill in abusing AI have more than a little of that “serial killers always start out abusing animals” energy. Along with terms like “clanker,” which seem to flirt with people seeking the dopamine hit of using racial slurs without actually using any, there may be reason to crack down on treating AI as a sandbox for humanity’s worst impulses beyond whether it’s sentient or a fancy word calculator. And for the record, it’s the latter.
Look at me going off on tangents again.
Unfortunately, the billable hour treats experimentation as a personal failing. Any associate who wants to commit a Tuesday afternoon to building skills knows they’re losing time. Like sands through the hourglass so are our bonus targets. Giving associates time to learn AI without sacrificing their compensation will pay dividends down the road. It’s a mechanism that Ropes has deployed in the past, like when it let associates bill up to 100 hours for diversity and inclusion work. Of course, that was in 2021… before they deleted all the DEI info from their website to avoid offending the Trump administration.
Akerman has a more formalized process for giving credit for AI experimentation. At Akerman, playing with AI is another assignment that associates can pick up.
Either way, these firms are making an investment. While the AI cult wants you to believe the technology will soon replace 100 percent of attorney hours, the reality is probably somewhere between 5 and 20. Using the Ropes model, 100 hours is about 5 percent of a 1,900-hour target, so if the firm believes that well-oiled AI can boost productivity even 6 percent it’s worth it.
Put aside building and customizing with AI, the proliferation of AI tools requires some experimentation. We’re entering the kitchen sink era when it comes to AI tools. Firms are now willing to buy competing products and let teams and individual lawyers sort out their preferences. Beyond competition, products are announcing MCP connections to complementary products every day, creating a network of warp tubing like a Super Mario Bros. level. Figuring out which user interface the lawyer prefers to do all their work in is an experiment in itself.
For now, these programs represent an exception rather than the rule. But every firm planning to get serious about AI has to come to grips with the fact that this tech isn’t going to optimize itself. Lawyers need time to make it suit their style.
Earlier: The Grace To Dabble: Two Biglaw Firms Look To An AI-First Future
Agentic AI And The QWERTY Problem
Joe Patrice is a senior editor at Above the Law and co-host of Thinking Like A Lawyer. Feel free to email any tips, questions, or comments. Follow him on Twitter or Bluesky if you’re interested in law, politics, and a healthy dose of college sports news.
