No Country for Old Men
AI is going to change how lawyers work – and there’s not a damn thing they can do about it.
The calls have already started. ChatGPT has been out one month, and the ban hammer is already manifesting in schools and articles are already written about how terrible it is. All these writers and administrators should give up now before they are outpaced by a whole bunch of people smarter than them – and who won’t try to hang onto the past. The clever ones are already trading prompts, trying to figure out how far they can push it. It may be a bit clunky, but it will get better. Or maybe a more refined product will come along and ChatGPT will reside in the Internet Hall of Fame along with Netscape, AOL Online, and Napster. But just like those programs, ChatGPT is going to change the world again in ways that are so profound we won’t be able to anticipate it.
It will be interesting to see how things change. A lawsuit against Midjourney AI over copyright could profoundly change how the concept of copyright is understood, a concept that hasn’t changed since the 1920s. I can, however, make a few predictions based on what I know right now. Here are nine ways that ChatGPT is going to change the practice and the process of law:
1. It will proofread.
This is such a pet peeve of mine.
Anyone who says that they can just type up a perfect document with no typos quickly is just delusional about their own ability. They probably think they are “above average” drivers too. I suggest they examine how many times they’ve broken the law and not gotten caught, much less how many times they have. Yet, look at any job listing, and proofing is somewhere on there. Sometimes it’s explicit, sometimes it’s hiding under the skill “attention to detail.” Employers get caught up in this, expecting both speed and perfection every time and wondering why it never works. I knew a woman who went through ten assistants in three years because of this. She would dress them down for tiny typos until they were in tears. Eventually, the firm got rid of her because the turnover costs were becoming too much, and they were afraid of being sued.
If you want perfect text, hire a proof-reader. Anyone who works in publishing will tell you that. They will say, “We’ll check it and check it and check it again, and you still don’t catch everything until you are in the blues.” (Old timey term for the last proof before the first print run of a book.) Partner two people with express authority and responsibility to proof each other’s work, like an old-fashioned typing pool. Or, if you want something done fast, do it for yourself and stop haranguing your assistant. It will be so much quicker.
After staring at text so long, anyone with any real intelligence is going to see a pattern that will blind them to errors. That’s what intelligence is. MS Word does a pretty good job, but it doesn’t always understand the context of the word. ChatGPT does and that is a level up from your typical grammar check. Of course, there are already services like “Grammarly” but those services work on the cloud, which is a privacy risk from a legal standpoint. Having a service like ChatGPT built into your native wordprocessing app is going to be a whole different ball of wax.
What is more, it’s ableist. For dyslexics to have to proofread their own writing is to set them up for complete failure every single time. It’s tremendously cruel. If your workers are living and dying by their ability to typeset, you are just burning money because there are probably a whole lot of other things, they do a whole lot better, and you are just demoralizing them. Stop it. Get a life. You hopefully didn’t hire them just for their typesetting skills. From now on, everything I do is going to be processed through AI, including this post.
2. It will change exams and maybe render them obsolete.
I hate, HATE, closed book tests. I guess they are useful when you must remember basic things like how to spell or do math. But law is always changing; the interpretation of law is always changing. You go to bed one day and by the next morning, fifty years of black letter law are tossed out the window and abortion is illegal in Mississippi. Good lawyers know they should always note up cases. Much mischief has been done by well-meaning lawyers practicing outside of their core expertise. Self-doubt makes fewer mistakes. Legal professionals are notoriously risk adverse. And in law, you aren’t expected to “remember” anything. Transcripts are made, notices are personally served, appointments are faxed. The beautiful thing about the procedural law is that it is specifically designed to make sure you don’t forget a thing.
What do closed book tests test anyway? Not how resourceful you are, not how well you approach and solve problems. Not even how well you remember things because people usually perform much more poorly in memory tasks under pressure. Recently, ChatGPT passed Wharton’s B school exam. You have to ask, what was it really doing for people that a computer could pass it? Was it predicting how well a person would perform in a c-suite or was it just spitting out information acquired, which in 15 minutes could be wrong? What was that test actually measuring, anyway? Maybe the next time you have a major business decision to make, instead of hiring a VP, you can just ask ChatGPT.
There are a lot of educators out there wringing their hands because of the fear that the next time they assign an essay, how will they know if that essay is written by ChatGPT or by the student? Honestly, it’s not so hard. Teachers know when a student is ripping off something else because they know just from talking to them, they didn’t write it. Most people write like they speak. If something doesn’t line up, they probably had some help. No, what they need to be concerned with is how they will know who is cheating on a test where answers are uniform and simple.
***Quick digression for context***
It appears that, despite the complaints of bankers and judges and anyone over 55, employees and students are going to demand to be able to do certain things remotely. After a year of hectoring the public, JP Morgan is still threatening the 20% of employees who aren’t showing up. Apple had to eat a bunch of free-range crow. Desperate HR departments are listing open jobs as “remote” when what they really mean is “maybe you can work one day from home.” Employees would rather quit than go back into the office. I mean, it’s pretty awful right? I just want my own private toilet, thank you.
Anyone who is even slightly introverted, has ADHD, is on the Autism spectrum is sighing with relief. No more pretending to like happy hours after work or someone you don’t work for walking by and dumping random documents on your (very messy but you know where it all is) desk. Do you really need employees to work in a team? What if they unionize? What if they fraternize? What if they go out drinking after work, get way too drunk, accuse each other of stealing their boyfriends, and a fight breaks out and you fire them, and then they sue you for $100,000? And win, and then win more on appeal? (You know who you are). Not everyone needs to work together to be valuable. Sometimes it’s better if they don’t.
Parents are doing the math to compare driving and daycare and lunches and work clothes and sick time and overtime and finding it doesn’t add up. Single parents were crushed over the pandemic. People are more productive at home by the account of most studies. Actual productivity can be more easily tracked if necessary. Just recently some shiftless employee who sued for wrongful dismissal wound up having to pay her company back for her time theft. I wonder if her co-workers knew how much less work she was doing and getting paid for it? Elon proved that there were a lot of people working for Twitter who were doing yoga at the office and not much else. In such a toxic political environment, no one of any party wants to be out. And god forbid your kid gets a messy cold, and you are stuck at home for days at a time, and even though they are fine, you know if you send them to school you are going to get a call, and you are a single parent with no back up, and now you are on the “bad parent” list and everyone is giving you that withering stare when you finally show up to get your kid and you have to use up all your vacation days on this while your boss gets annoyed because he just wants to be able to talk to himself in his office and things just magically happen….Fine. FINE.
***Back now***
It was one thing when you could bring students into the classroom and test them there, but if they are remote, who knows if they are checking their cellphones, or a second monitor? There’s really no way to ensure that people don’t cheat, or even if they paid someone else to cheat for them. Or went to the bathroom with a cellphone.
Are schools really going to invest in classrooms and real estate for those two or three times a year that students need to be in the room to take a test? Really? Most of them don’t want to pay for teachers, who are now actually valuable because **humans** can tell if a student has the skills to do whatever it is. And are schools really going to give up all the revenue they can get from teaching students far away? Are students going to want to move to a city where the real estate is miserably expensive, and crime is high, and they can’t work enough to pay the rent on a studio? Are they going to want to bunk up six at a time to learn how to do online research on LexisNexis?
In a world of online, remote learning and working, if you cheat on a test, you are going to have an edge. Opaque, classist, possibly useless and maybe generally unfair, testing as we know it is going to have to change or go away. Maybe we can bring back apprenticeships.
3. Standard Forms
I have three Alexa devices in my house. (Well apartment, let’s be real here, I’m under 50.) I love them. They tell me when to stop writing and make dinner. They turn off the lights. If I’m cold I go “Computer, I’m cold” and they turn on the heat fans. If I say “Computer, Red Alert” I’m suddenly on the deck of the Enterprise as the lights turn red and that classic alarm goes off. It’s my favourite thing. I don’t even mind the spying.
There’s absolutely no reason in this world I couldn’t look at my computer and say “computer, using the most recent record in DivorceMate, make me a Notice of Family Claim, asking for shared guardianship, shared parenting, primary residence with mom, child support based on the guidelines, spousal on the high side of support, and 50% property division of all family property. Use standard law and cases under the section “legal basis”. Rely on affidavit number 1 of the claimant.”
The technology already exists to have a working, mostly completed form in less time than it would take for a legal assistant to return to her office. How to set that up however, takes a little sophistication and some basic programming. I’m sure there is an equivalent in real estate. Or in insurance. Or in negligence. Any form where there is a software package that merges information to templates would be eligible to take the next step to becoming an AI application. Would that mean that paralegals would be out of a job? Well, maybe, but I’m pretty sure they will still be needed to do the things lawyers don’t want to do, (which is everything.) Which leads us to the next thing:
4. Basic drafting
Every area of litigation has its standard cases and law. You know the ones. They go in every notice and every brief because you have done this same divorce, this same auto accident, this same insurance claim over and over and over again. You always have to put in those same standard cases to make your claim, followed by a string of citations. New lawyers may know them or have heard of them, but you still send them to dig through an old file to find the motion that has the case and the argument and get it plugged into a new one. These cases and law, of course, have to be followed by their proper citation etc, and maybe your paralegal has set up autocorrect so the citation is automatically added in Word…you know the drill. You might have even saved entire paragraphs in Quickparts under the title “Insurance - duty to defend.” Hopefully, you don’t forget to wipe the precedent clean of your previous client’s names.
There is literally no reason why anyone should have to do that by hand again. AI can do that. Sure, you might have to fine tune it after, and maybe make a cursory check to see if the treatment has changed (it hasn’t under the principle of “Heuristics that are almost always right,” See Astral Codex Ten, Substack.com), but the writing is good enough for government work and you don’t need to reinstall cases and their arguments over and over again. This cuts your billable hours in half, of course, or more importantly your paralegal’s, but you are the better lawyer with now lower prices so I’m sure you will have more clients than that other guy, and hey, didn’t the Law Society complain that lawyers were too expensive and are going to do something awful to your area of the law if you don’t lower the costs? How lucky do you feel?
5. Timekeeping and billing/ non-billable work
Up until now, most law firms I have worked for utilized a stand-alone software package to track their billable hours and do billing on paper, even if they send out email bills. These packages usually are full-service, but sometimes only a handful of the functionality is actually used by the firm. There is virtually no training for support staff because training is sold as a service and the firm won’t pay for it. Sometimes the law firm only purchased the “bare-bones” version, and only key functionality is operational or key upgrades haven’t been installed for years.
Think of a “theoretical” law firm where bills, after being physically printed and signed, are then scanned, attached one-by one to an individually written email and sent to the recipient, whose correct email address is not always the one in the database, so you have a second, more accurate Outlook address book. This email would then be printed out and attached to said signed bill and filed in a physical file. Such a law firm would have support staff working for hours on tasks that a properly programmed computer could accomplish in minutes, but for some reason the senior partners can’t make this connection, so they waste tons of money on support staff to do it. This is after having to “gently” remind the lawyers to make sure their time was put in. This firm is theoretical and I’m not pointing fingers, but I suspect this scenario is far more common than many people realize. To any other industry that sounds completely archaic because, unlike law, there’s no big barrier to plucky startups funded by Peter Thiel that prevents them from eating a law firm’s lunch. Yet.
There will come a day, in the very near future, where a lawyer or a paralegal will just say “Computer, add the following time to this client’s bill” and voila, it will be added. The technology already exists to totally automate the process above. AI will make it even smarter, and easier, and, more importantly, cheaper. Firms that say “we’ve always done it this way” probably don’t care. They should start caring, because the old way is expensive, wasteful and an aggravating process. AI will make it all go away for the smart firms who can capitalize on it first. Ditto for everything else that’s being done like it’s still 1995. Maybe that can subsidize your pro bono work, which you do because you are not going to be the lawyer that doesn’t adhere to the unspoken social contract that you entered when you got your law license, amiright?
6. Complicated Math
Right now, in legal firms everywhere, are software packages that calculate damages. They calculate fees, they calculate child support and spousal support and property division, they calculate taxes. All of these kinds of calculations must be vetted by a person apprised of the individual situations of each client. These calculations act as a starting point to figuring out who is going to pay what. I have known some truly gifted lawyers in my life, who seemed to know, down to the penny how much tax anyone was going to pay on a theoretical gift, and how to argue through some arcane tax loophole to get out of it.
The beauty of AI is it learns, it’s flexible, it can take diverse information and apply it. There’s only so much range in how much money litigants will get. Certainly, a lawyer can ask, argue, pound on the desk for more, but with such things like the implementation of no-fault car insurance, that math can be done by a computer. Even I, mathematically challenged, could do a typical conveyance if the computer asked me the right questions. Now imagine if I just need to talk to it like a person.
7. Self-serve legal advice.
Right now, there’s a section on reddit called “legaladvice” and people with legal questions go online and ask questions with the hope that they can get some direction as to how to deal with their problems. It’s clear that a lot of people giving advice aren’t really legal professionals, so it’s hit or miss.
But the people who are “real” generally start asking questions and take those questions and turn them into real advice. Often people just need direction like, “do I need a lawyer?” “Can I appeal this judgment?” “I’m getting married, if I get divorced, do I get the house?” Often these questions don’t have definitive answers. I mean, have you ever gotten a yes or no answer from a lawyer? But they can get a good idea of the lay of the land and know what kind of questions they are going to have to answer.
I’m sure at some point, some clever lawyer is going to figure out they can use AI to triage cold calls. We are already at a point where bots can answer the phone, ask for addresses, point to solutions to problems that a customer has. It’s only a matter of time. And if by triaging cases, frivolous ones get eliminated, well, some kind hearted soul is going to put up a chat self serve so clients don’t waste a whole lot of time and money. Everyone will be able to get a second opinion. Lawyers will fume. The Law Society will rejoice. Access to justice problem SOLVED.
8. Dr. Google
It used to be that you had to be in medical school to get “medical school syndrome.” This is the tendency that medical students have to convince themselves that they have whatever malady they were reading about that day. Minor headaches became meningitis, some tiredness was cancer. We all knew that one girl in college studying psychology who annoyingly would try to psychoanalyze everyone she came across. No, Stacey, I do not have schizoid personality disorder. (It’s a totally different one, thank you very much.)
Nowawdays, anyone with WebMD can pester their doctors with claims of Lupus or Fibromyalgia. Various mental disorders become the fads of the time and tweens and teenagers suffer a rash of ADHD/Aspie/BPD/Bipolar/PPMD/DPD and gender dysphoria, depending on the year. Surely some of them are the real deal, but like the new “pretendians” some are probably just trying to curry social currency. Well get ready, lawyers, for the wave of vexatious litigants, and advocates who have graduated from the School of ChatGPT.
Don’t get me wrong, I’m a bit of a Google Scholar myself. And PLENTY of important, smart, and successful people in history had no formal schooling at all. It should be noted that there’s no Constitutional requirement that a judge on the Supreme Court of the United States attend law school or even be a lawyer at all. But knowing when to bring a case is knowing what is in scale. How much harm is “harm” for the purposes of a personal liability claim? Is it really worth it to argue over who get the kids on Thanksgiving?
It’s still relatively hard to sue someone if you don’t know what you are doing, what forms to fill out, what cases to cite. There are court costs to be paid. And the ability to go to the courthouse library when it’s open during business hours is still the luxury of people with a lot of anger and nothing better to do. With AI, the “Hail Mary” cases might seem a lot more possible than they ever were before.
9. Gaging Judges
One of the advantages of getting a more senior lawyer is that after practicing for years and years, they get to know the judges. This can be helpful because if you know a certain judge has a philosophy of your issues, you can gage maybe before you get into the courtroom how successful you will be. Now, I’m not saying that judges aren’t objective or making decisions not based in the law, but judges are humans with opinions about what the law means or is designed to do and under that law, you can sometimes guess what they will decide. This is especially true if the law gives them significant latitude in their discretion.
A few years back I took an online test as part of a “What color is your parachute” kind of program for unemployed people. The program was designed to help people who had been laid off through no fault of their own find new employment and part of the program was exploring other careers. They had this sort of early AI tool that you answered a bunch of questions online and it spat out jobs that it thought you would enjoy. Curiously, mine questions resulted in the typical art gallery owner, surgeon, and lawyer (I’m not saying they were practical.) and then “auto mechanic.” I was flabbergasted. I had told NO ONE that when I was child that I wanted to be car mechanic, a job with a very high barrier for entry for girls who couldn’t drive. But I loved cars, and for not an insignificant time in my life I wanted to be an auto mechanic. My family would have never let me go to the trade school to be one, I was college bound, like it or not, and maybe I would have gotten bullied out for being a girl. But I wanted it, and this computer knew it.
Here we are, at least 20 years later, and the technology has only gotten better. Pump that tech with every opinion and decision that a judge has written or concurred with and I’m sure it could pump out a fairly accurate picture of how that judge may rule. Which is the point of the exercise, isn’t it?
Not everything – but an awful lot
I’m not suggesting that lawyers or even legal support staff can be totally replaced by machines. Eventually human beings are going to need other, real human beings on the other end of the phone. Having a legal problem is scary and confusing and having another person talk you down, (or give you a lecture) is often what clients really need. Lawyering is not a technical business, it’s a people business, and for that you need people. And one thing I did not include, which I could, was “justice by algorithm” where cases are judged by the machine, a Judge Dredd if you like. I don’t see that happening. Mercy is not quality computers have. There is no ghost in that machine.
But there are a lot of technical aspects to lawyering. After the initial interview right up to the hearing, a lot of work is plug and play. Calculating numbers, drafting documents, docketing time, routing calls and emails, filing, sorting, tagging and even scheduling are all things that could wind up being either partially or completely managed by AI. And now anyone, with no training, can do it. Anyone.
The true barrier to AI doing any “real” lawyering is lawyers themselves. No one wants to see their job replaced by a robot. Especially a robot with infinite knowledge of every single case and treatment and statute and regulation ever entered in a database. Yet, every day, other industries are installing kiosks and robotic arms as fast as they can. Surgeons are doing robot assisted surgeries. Radiology was found to be more accurate at diagnosing cancer with a computer. FINALLY, I don’t have to check my bag every time I leave McDonalds to make sure there’s enough ketchups in my bag. Why are lawyers so special anyway? Why should they get a pass?
Lawyers may still be better, but you have to ask, “compared to what?” Everywhere in Canada and the USA, there are litigants walking into courtrooms with no lawyers at all, muddling through applications and forms, scared witless because they think they are going to lose visitation with their kids or go bankrupt. In the USA, it is estimated that anywhere from 80 to 90% of litigants are “pro se” depending on the court. And in Canada it was an estimated 40% who were unrepresented in family courts (where you almost always really do need a lawyer), and upwards of about 70% in civil. Almost everywhere, it’s agreed that the general public’s access to justice is close to “dumpster fire” level. Yet, despite this, any innovations, like allowing skilled paraprofessionals to speak in court, or AI programs like “Donotpay.com,” the New Thing is fiercely resisted. And think of all the clients that could afford lawyers if law firms could manage large volumes and lower their prices accordingly? It appears a miserable status quo is preferable to a mediocre middle. That will only last so long. Whether it is allowing non-legal professionals to own law firms, or subsuming auto insurance claims under a restricted tribunal system, lawyers should hesitate to relax behind their hand-carved, oak desks, and rage against the machine.
A long time ago I worked for a telephone company. It was right around the time they were replacing their old mechanical switches with computers. The old switches take up a lot of space, and so, dotted across the state, were ugly, gigantic, windowless buildings where the sole occupants were these switch machines. From the invention of the telephone to the end of the mechanical switch era, AT&T was the largest real estate holder in the United States. As computers replaced them, the old machines were not needed anymore, nor were the buildings that housed them. So, the department I was in was responsible for selling off this real estate to developers that could do something with them. (Although they were probably sold to John Malone.)
Machines are not people, but the automation would continue at a pace, wiping out the yeoman’s work of industry just like those old switch machines. Following the telephone operators would be telephone books, maps, little advertising firms, secretarial pools, newspapers, telecoms, cable, and record stores- the MAIL, it would all disappear one-by-one. This was my first job out of college, and it taught me something:
The future comes for everyone, whether you want it to or not.