Meet Generative AI, Your New Summer Associate

Think back to the halcyon days of yore, when we were all wide-eyed, optimistic (and not yet jaded) 1Ls and 2Ls vying for one thing: a summer clerkship. Becoming a summer associate often meant extra cash, and for the lucky few, a job offer post-graduation. But it wasn’t without toil on our parts.  

Suddenly, we had to draft real memos outside of our Legal Research and Writing classes. We were expected to be available at all hours for any question a partner might have. You encountered things like the Rule Against Perpetuities in the wild (this is very real and happened during my 1L clerkship at an oil and gas firm in Houston). Really, you spend the vast majority of your summer hoping to high heaven that you aren’t making mistakes or embarrassing the firm.

Summer associates are untrained, untested, and a tad unpredictable (we all knew that 2L who over-imbibed at a firm happy hour never to be heard from again). However, they are also smart, eager to learn, and desperate to prove themselves. After all, they want a job with your firm.

But now comes the newest generation of summer help: generative AI. For firms who cannot afford the cost or devote the time to hire and train up a human summer clerk, but would still appreciate the help, generative AI is evolving to fill that void. Generative AI can’t get your coffee or carry a redwell into court for you, but it can draft memos and perform research on your behalf. As time goes on, I imagine that generative AI will grow into being able to help with most aspects of summer associate life.

A word of caution about your new robot associates.

Remember, generative AI is fairly new to the field of law. There are still a lot of kinks to work out, and many state bars are scrambling to get new rules on the books about AI. Some state bars have been more proactive about the governance of AI. See the Florida Bar’s proposed amendments incorporating AI, the California Bar’s guidelines, and the New York State Bar’s report on AI. States like Texas have formed workgroups or coalitions on artificial intelligence. All this to say – AI isn’t going anywhere, and we need to start getting comfortable with its application in law.  

No matter the state bar, pretty much every AI guideline coming out echoes the same core tenant: treat AI like it’s your employee. The bars aren’t really adding brand new rules into their ethical guidelines, so don’t worry about having to memorize a new set of regulations. Rather, they want to ensure that the existing rules now apply to the use of generative AI. For instance, just as we have to adhere to confidentiality requirements, so too does our use of generative AI.

Guidelines for managing summer associates

Keeping the aforementioned in mind, I thought it might be beneficial to have some guidelines in place for managing your summer associates, be they human or artificial:

1. Never fully believe them. Remember, these clerks don’t have law degrees (although ChatGPT did pass the bar exam and MPRE). They are desperate for your approval, but also can be arrogant. They are green and susceptible to going down rabbit holes because they don’t know any better, which means sometimes you’ll get an answer to a query that is only half-true or not relevant at all.
   
   What’s to be done? Pertaining to physical being and robot alike, the more specific you can be in your requests and instructions to the associate, the better work product you will get. For humans, this means telling them exactly what you are looking for (e.g., a memo on establishing paternity) and where to find the information. For generative AI, effective prompt-writing will be your ticket to solid results. How you ask the question has a direct effect on the quality of work product you will get back from generative AI, so make sure you know how to ask the right prompts. The same could be said of human associates.  

2. They just want to make you happy. Summer associates hope to do well and get a job offer. Generative AI hopes to do well to get you the exact answers that you want. Both have similar motivations that result in them sometimes bending the truth or lying altogether. I remember being in court during my 2L summer, and opposing counsel completely lied about the holding of a case he had cited in his brief. Turns out, his summer associate had bungled that draft and lied to make his boss happy (we won that motion, by the way).
   
   For generative AI, making you happy sometimes manifests as hallucinations. In its zeal to please you, AI can and will make stuff up. This includes case law. And, the case law will look convincing on its face, so be careful. Don’t ask ChatGPT for a brief on confirming separate property in Texas and expect the cases cited to be legitimate. Ever.
   
   What’s to be done? Use legal-specific AI software that has the proper guardrails in place to minimize the risk of hallucinations. And always doublecheck the work product of your summer associates. Even the smartest minds mess up (and sometimes they are too scared to admit defeat). Part of managing a law clerk is supervising their output, which leads me to the next guideline:

3. They require constant supervision. As many state bars have advised (see above links), you are ultimately responsible for the work submitted by your staff, which now includes generative AI. Articles abound about attorneys who relied on ChatGPT to help draft a brief and got the bench slapping of a lifetime from the judge (because turns out, dear reader, the cases in said brief were fictional). Similarly, human law clerks must be heavily monitored to ensure that their output is credible and reliable.
   
   What’s to be done? Use legal technology that cites its sources. This enables you to verify and doublecheck the work accomplished by AI. The same applies to human associates. If they cite a case, they better be printing out that case and highlighting the relevant parts for your perusal. Also, in a universal application, no document should leave your office without you having blessed it. Remember, human and generative AI associates do not possess a license to practice law (yet), and it’s your bar card on the line, not theirs. Conduct yourself accordingly.

4. They can’t keep a secret. Associates like to gossip with their compatriots (it’s how we get through our clerkships), but sometimes they overshare information about the firm and its clients. This is especially true if your firm handles celebrity or high-profile cases – many want to brag and one-up each other when it comes to their summer clerkships.  This in turn can embarrass the firm and can embarrass you. This is why their behavior at networking events should be monitored and taken into consideration when making an offer.
   
   Similarly, your secrets are not safe with generative AI. If you put sensitive or confidential client information into a program like ChatGPT, that information is now available for public consumption and data scraping. Every state bar has ethical rules that require strict adherence to client confidentiality, which AI use is not immune from. It’s also not just lawyers being affected by this.
   
   What’s to be done? Your human associates have no real idea what client confidentiality is. You need to clearly explain it to them before they start at your firm. This includes divulging any client name or sensitive information to others. For generative AI, as mentioned supra, use a legal-specific technology that has the proper safety measures in place to ensure that your firm’s data is never shared with a third party.

Implementing AI might sound overly burdensome or a pain, but generative AI is only getting hotter and its use in the legal field is growing exponentially by the day. Attorneys are figuring out how to use it to get more free time back in their days, and they are also monetizing it with amazing results (check out my discussion of AI and fixed fee billing here). Staying ahead of the curve with generative AI will yield dividends for your law practice, just remember to treat it like a bumbling summer associate and you’ll be fine.