Clients are increasingly using AI to inform themselves about their legal cases. It is generally good when clients are actively involved in their cases. Unfortunately, AI is costing clients loads of money and damaging attorney
Leading Questions
People inherently ask AI self interested questions.It’s hard to be objective about our own lives and circumstances. This is the reason law professors routinely advise law students to never represent themselves.
Let’s take a divorce case. If you ask AI, “Do I have to give my wife half my assets in a divorce?” It will tell you “No.” The user then thinks he’ll never give up half the assets. If the user changes the question to, “How are assets divided in divorce?” The answer changes completely.
The same goes for business law. People will ask, “Will I win if I sue my partner?” and then the user proceeds to list all the favorable facts while omitting unfavorable facs or mitigating evidence. In probate, “Will I win if I sue my sister for stealing assets from our mom?” The AI program will always tell the user they are right and find information somewhere on the internet to back it up. The problem is that the questions are not objective and neither are the results AI spits back.
In short, AI tells people what they want to hear. That’s dangerous, especially in law. The most challenging part of being a lawyer is telling people things they don’t want to hear. AI does the opposite, and consequently is advising clients to pursue imprudent aims.
AI is Often Wrong
Most AI is improving, but it’s still wrong more often than it should be. A few specific problems arise with AI being wrong. First, people assume it’s correct and just go with the answers. In the real world, people “kick the tires” on the professionals offering advice – he sounds confident and has information I don’t, I’ll trust him. With AI, people just go with whatever it says because it’s telling them what they want to hear.
The second issue is that it pulls law from the wrong places. The laws in Charleston, SC are very different from L.A., California. It’s not just that it cites law from the wrong states too often, but hallucinations are real. It will just make up cases or facts or law that is not based in reality.
The last reason it’s often wrong is because it takes things out of context. An explanation to provide context in a case can be used by AI as the holding or a legal conclusion. It just searches for answers the user wants, so it does not use judgement to determine if an argument is being used out of context. By the way, lawyers would be disbarred if we mislead a court.
Lawyers Make More When Clients Use AI – A Lot More
The three page AI generated email needs to be read (.5) and researched (.5) and responded to (.5)- that’s 1.5 billable hours at around $400 per hour. The email surely will have accurate information, but it most likely will contain nothing your lawyers didn’t already know. When we receive these emails we have to read, research, and respond. What we find is that AI places too much emphasis on the wrong arguments or there is something totally unimportant or there is factually wrong information.
No matter how accurate AI is, we make more money. And it does not end with the email because the client will respond to our response. And then, if we have to prove our points – I’m right, your bot is wrong – we have to send more emails. Before you know it, over $1,000 goes down the drain. There has been no forward progress, just time billed.
Wrong Emphasis
When AI is asked a direct question, it’s great. “How many miles is a marathon?” A more complex question is how long will it take me to run a marathon? The second question is tricky for AI because there is no way for it to know a particular person, so the results are usually between “2-7 hours.” Even given information about the person, AI can’t predict the results.
One reason AI cannot predict results is because it places too much weight on the wrong ideas. In law it might be failing to understand the opposing side is argumentative or failing to see if someone is telling the truth. These human variances make it difficult for any computer system to understand.
It’s not simply failing to understand human qualities, but also legal priorities. AI might focus on a line in a court decision, but fail to address the jurisdictional issues. Or, it could highlight an argument used by another lawyer in another case with different facts or in front of a different jury pool. In these cases, the advice from AI could end in disaster for the client.
What this means is that AI advice will sometimes miss the most compelling argument, and the user will never know when it’s nailed it or missed it.
Using AI to diagnose medical issues has a similar result. It can’t know the person or how to prioritize factors. Similarly, clients find this when asking AI questions with complex, and highly disputed, facts.
Unnecessary Discord
Clients need to be involved in their case, but it usually works best when it’s attorney lead involvement. We ask questions, review evidence, and ask more questions. We procure evidence from opposing counsel and analyse it. We conduct thorough legal research, interview experts, decide who seems most truthful or sympathetic, apply it to the admissible facts, and provide our clients with options.
When clients depend more on AI than their lawyer’s opinion, then tensions can rise. Once a client loses trust in his or her lawyer, then it’s hard to gain it back. We have not yet had a client lead AI session benefit a client, and increasingly unnecessary discord is arising.
More often, clients insist they did all the work and ask, “why is the bill so high if I did everything?” Well, your misguided “research” took 5 seconds and gave you good answers, bad answers, and a bunch of irrelevant information. We have to read it – that’s the job. Some clients erroneously think AI is saving them money, when it’s actually costing you money and sowing the seeds of discord with their lawyer.
Practicing Law Without a License
AI disclaims all its legal commentary by suggesting users contact a lawyer. This is hardly noticeable, which means AI is practicing law without a license to practice law. It is offering a legal opinion and telling people to follow it. Under South Carolina law, “advising an individual regarding his or her legal rights, remedies, or obligations” is not permitted without a license to practice law in the state.
The primary problem is the number of people acting on the advice of AI. At a juncture in the future, the courts will be filled with frivolous claims because AI lacks human judgement.
There are benefits to AI in the legal realm, but it’s best to let the lawyers use AI to shortcut legal research, scan copious documents, or solicit fine tuning of a document. Lay people using AI is often a costly mistake.
If you want to learn more about these or any topics, visit our blog, YouTube channel, or Instagram.
