Artificial Sweetener: How Law Firms Should Engage AI

By Mary Warner posted 05-30-2024 02:50 PM


Artificial Sweetener: How Law Firms Should Engage AI 

By Jared Correia, Red Cave Law Firm Consulting

Artificial intelligence has finally captured the popular zeitgeist via ChatGPT – a generative AI program, that can provide near-instantaneous responses to almost any query. Before we get too far along here, note that artificial intelligence is, essentially, the mimicking of human intelligence by a computer system that has learned via a ‘large language model’, i.e. lots of varied data. Generative AI is simply an AI program that offers data output, whether it’s text (like ChatGPT and Google Gemini), or images (like Midjourney or DALL-E). 

You can ask these tools just about anything and they will offer an answerHey, Google Gemini, I have these 5 ingredients in my refrigerator. Generate a recipe for me.  Midjourney, I want you to create an image of Joe Biden and Donald Trump holding hands and skipping through a field of daisies.  ChatGPT, draft me a contract clause for default on a commercial real estate transaction. (Yup, that too.) 

The more effectively you craft these ‘prompts,’ the better the response will be(And, if you’re a really good prompt engineer, you might have another highly compensated career opportunity unfolding before you.)  AI is an impressive tool that is a testament to human ingenuity (and perhaps, overreach), which is currently generating a massive effect in every industry and everyone’s personal lives. That’s all well and good, but it also begs the question: How should lawyers use ChatGPT and tools like it? 

AI Is Currently an ‘Assistive’ Technology 

If you’ve seen any science fiction movie of the last . . . well, ever, you’ll know that AI is almost always portrayed as a rogue force that goes off the rails and eventually attacks its creator, along the themes of the classic ‘Frankenstein’ story. And we’ll likley get there, honestly, just probably not in our lifetimes. That means that AI is at an earlier, mostly benevolent, stage. It’s currently an assistive technology. That means its sole goal is to help humans with their queries – sort of like a virtual butler or home health aide (theoretical concepts, currently, that will also come to pass). 

The fact that you can ask a generative AI tool almost anything and get a pretty coherent response is a testament to the technology, but also to the breadth and variance of the data fed into the model it learns from. It’s also learning each time it interacts with you. So, the technology continues to improve, to pick up speed, and to learn from every interaction it entertains. 

And you’d be foolish not to use it, as a human, as an attorney, and as a business owner – since it is a viable tool for providing shortcuts to a final product. In the end, AI, as it’s presently constituted, is a time-saving device, just like a second computer monitor, but with far more horsepower. If you are a lawyer (especially an attorney who owns a law firm), you should always endeavor to use time-saving tools. But, when you’re using any new technology, you should be aware that it has limitations and approach it with some level of caution. That doesn’t mean you shouldn’t use it. The early adopter catches the worm, after all. You should just utilize it with an appropriate level of discretion. Here are some tips for doing just that: 

Oversee AI Like a Staff Person 

In many ways, generative AI sits on the razor’s edge between being a useful technology and becoming a staff person, which is something that legal ethics attorneys are currently grappling with. An AI tool is effectively software, but it’s software with which you can have a conversation. You can ask it a question or make a request of it. You can even iterate in the tool by asking it to refine its response in a way that will revise the answer, e.g. be less formal, add bullet points, summarize this section, cite your sources.  AI tools will endeavor to do everything you need them to do to the best of their abilities. 

So will your staff person. If you ask a paralegal or an associate attorney to draft you a motion or a contract, they will perform that task to the best of their abilities, too. Does that mean that you would turn that document over to a client, sight unseen, without reviewing it? Of course not. And, you wouldn’t do that for a whole host of reasons: it’s not good business sense, as the boss you should be approving everything that goes out the door under your letterhead, and, as an attorney, you have a stone-cold ethical obligation to oversee your staff, including their performance, but especially their work product. So, although AI is functionally just another technology in your toolbelt, it’s really more than that. It’s actually another staff person. Therefore, treat AI work product as you would a staff person’s work product. Review it with a fine-tooth comb, both for quality and accuracy, and make appropriate edits (which may be heavy), before anything leaves the physical or virtual confines of your office. 

Where lawyers have gotten over their skis is by being too trustworthy of AI, assuming it is always correct without checking the final work product. As you would never do that with a staff person, don’t do it with a technology product that emulates a staff person. 

Talk to AI Like a Kid 

When I ask my kids to do anything, I have to be extremely specific. I can’t just say, “Clean up after dinner.” I have to relay a set of granular instructions, like: Take your plate from the table. Use the fork to scrape any stray food into the garbage. Wash off any remaining residue in the sink. Place the plate inside the dishwasher tray. Drop any utensils in the basket.”  You get it.  

You have to treat AI the same way. Don’t presume it knows anything at all if you want to get more effective results. 

This is why prompt engineering is so important. To get the most out of any AI tool, you have to ask the right questions. You can start by priming the pump. Tell the AI tool about the persona you want to use: Draft this as if you were a real estate attorney in Edina, Minnesota, with 20 years experience. Then, perhaps, feed it some supporting documentation: I’m going to now input several emails I’ve drafted, so you can try to copy my style.After that, edit the output until it’s more refined in terms of quality, tone, style, information sets included, etc. 

Once you think the output has a fine point on it, review it again.  Measure nine times, cut once. Just like you would check over your child’s homework or review the work product of a staff person. 

Don’t Believe AI (The ‘Hallucination’ Problem) 

The biggest news stories in the legal vertical related to AI involve ‘hallucinations’ – which is a fancy way of saying that AI makes things up. In at least a couple of cases, AI has hallucinated or made up case cites that weren’t real; and attorneys have been disciplined because they submitted these documents to courts, including the false case citations. This should not be surprising, because that’s how AI works. It’s programmed to provide an answer, no matter what – and, like a child – if it can’t provide an answer, it panics and makes something up. But this isn’t a reason to avoid using AI – same as it’s not a reason to avoid talking to your lying kids! You simply need to check your sources and ensure that the information the AI is providing is accurate – the same way you would in reviewing a staff person’s work. You also don’t want to ask AI if its sources are real and leave it at that because, if it’s already making things up, its programming will force it to compound the lie. Check your sources and cites independently as part of your review process. 

Don’t Trust AI (The Confidential Data Problem) 

It’s still somewhat murky in the AI landscape when it comes to how the data in the learning models is collected and used. Several prominent authors recently filed a lawsuit against Open AI, which operates ChatGPT, for copyright infringement, since they claim that their works have been used to train the bot – which is probably true. As alluded to previously, these systems are learning from you via every interaction you have with them, including with respect to the data you input. In that environment, it makes sense not to input any sensitive or confidential data into a publicly available generative AI tool until these issues are clarified to a greater degree. At the very least, don’t include actual client names or identifying information. 

Note that this problem will likely be obviated as legal products, behind subscriptions and logins, have begun to utilize AI technology in their closed systems. 

. . . 

Artificial intelligence is far from perfect, but it is useful, even as-is. As a lawyer, you’ll want to access it sooner rather than later to familiarize yourself with how AI works. In fact, as you’re reading this, legaltech software vendors around the world are building out extensive new AI features. Oh, and the technology products you use every day are probably already implementing AI at some level – even if you’re not aware of it. These features will surface in a more transparent way in the near term.   

If technology awareness and utilization is an important part of the modern attorney’s competency obligation, then you must begin to understand the AI tools you’re already using, as well as those you will come to use. Lawyers are often late adopters, lagging in technology upgrades and efficiencies, but that approach has never been a wise one. It’s also a dangerous strategy when it comes to AI, which your competitors will use to outpace you on a massive scale. It’s time to enter the race – before it’s too late to catch up. 


Jared D. Correia, Esq.

Jared D. Correia, Esq. is the Founder and CEO of Red Cave Law Firm Consulting , which offers subscription-based law firm business management consulting and technology services for solo and small law firms.  A former practicing attorney, Jared has been advising lawyers and law firms for over a decade.  He is a frequent presenter at local, regional and national events, including ABA TECHSHOW.  He regularly contributes to legal publications, including his column, ‘Managing,’ for Attorney at Work, and his ‘Law Practice Confidential’ advice column for Lawyerist.