There’s lots of talk about the impact of GenAI and LLMs on the practice of law and what they will do to everything from workflows to business models to young lawyer training. But one thing that’s not talked about much is the impact GenAI will have on client relationships.
Clients have always come to their lawyer believing in the rightness of their cause. Now they will come with information from a third player: an LLM tool. Whether that information is right or wrong, it’s going to impact the trust between lawyer and clients. And lawyers better be ready.
To State the Obvious, Clients Are Using GenAI for Legal Questions
The large accounting firm, Deloitte, recently surveyed the top 100 Dutch law firms to determine the state of AI adoption in day-to-day operations. (The fact that the survey was done by an accounting firm that itself offers legal services, at least in some jurisdictions, ought to give legal pause.) The survey looked at a variety of things like strategy, training, and importantly, client expectations, among other things.
Here’s what Deloitte discovered about law firm clients: 60% of the firms report that clients are now using AI tools to perform simple legal tasks. As a result, clients are expecting from their lawyers faster turnaround times, transparency about AI risk, and of course, lower fees. Significantly, only 3% of the firms had seen no change in client expectations.
What this means is not only will clients be using the tools to perform “simple” tasks, but they are also going to use them more and more for pure legal advice and strategy, often even before they see a lawyer. This poses all kinds of practical problems, particularly given the fact that GenAI tools hallucinate, give wrong answers and advice, and often will tell clients what they want to hear.
The Practical Problems
If a client talks to his GenAI and gets bad advice and then acts on it to their detriment, that’s a real problem. By doing so, the client may very well place themselves unknowingly in harm’s way. And by postponing seeking human legal advice, the client may make their position even worse.
There’s also the discoverability problem: what the client tells his favorite bot may itself be discoverable, as I have written before.
So, by the time the client does finally see a lawyer, that lawyer may have to spend time cleaning up a mess. That will likely cost the client more, not less money, in the long run.
But the practical problems may be the least of it.
A Human Relations Problem
The human relations dynamic plays out in concrete ways trial lawyers will recognize immediately. Think of this: there’s a dispute with conflicting testimony. The client thinks his version will prevail in front of a jury and the bot supports him. The lawyer looks at the testimony and knows intuitively that the client’s version will not convince the jury for a whole lot of reasons like body language, jurors’ perception, and bias. How will the lawyer ever persuade the client (and their bot) that the client’s version will not prevail?
Since time immemorial, clients came to a lawyer convinced of the merits of their matter. That their version of the facts is the most convincing. That their strategy of what their lawyer ought to do is the best. It doesn’t matter whether it’s a family law matter or a sophisticated businessperson, most of the time clients think they know more than their lawyers.
Even in the best of times, this always placed the lawyer in a difficult spot. Pointing out to a stubborn client that their theory and strategy is wrong is always dicey. Say too little and the client gets the wrong idea about their case. That wrong idea will only fester and grow over the length of the case and can lead to horrible results and trauma later. I have seen it happen so many times: the lawyer gives the client the idea they are right and a year later when a good settlement offer comes around, the client balks because they think their case is better than it is.
But if the lawyer says too much, it’s also a problem. I’ve heard too many people complain that their lawyer “wasn’t on their side” because they were overly blunt in their assessment. It erodes trust.
But now we have a third player in the mix: a GenAI bot who may just be flat out wrong in its assessment of a case or problem. Moreover, it may be telling the client what they want to hear. And when a client tells their story to their favorite bot, they are going to tell it in the most favorable way.
So, now if a client wasn’t already convinced in the merits of the case before, they now have “evidence” from the bot. The result? It’s going to be harder to disabuse them of what the bot has told them, and the lawyers’ job will get a whole lot harder.
Another problem: if a client listened to a bot before they came to see the lawyer, they are probably going to listen to one throughout the matter. So, every call the lawyer makes, every recommendation they make, might be reviewed by bot.
But the crux of the matter is that the law and legal strategy is always a gray area, even more so than other disciplines like medicine. And when it comes to strategy calls, the lawyer and the client only know the result of the strategy that was adopted, not the ones that weren’t. So, the second guessing never ends.
Add on top of this the fee issue. The client believes based on the bot that the work the lawyer needs to do is not necessary. It’s a simple, slam-dunk case that shouldn’t cost as much as it is.
But the lawyer has to clean up the mess that wrong advice may have caused. The lawyer has to spend time convincing the client of reality and what needs to be done. All of that takes time and increases cost. In the meantime, a case and a relationship turn into a nightmare.
Bottom line: if lawyers aren’t careful, they will face an erosion of trust in the attorney-client relationship as their judgment and advice is substituted for that of AI. That trust has always been the bedrock not only of the relationship but in getting the best result.
It Need Not be Insurmountable
It’s not a hopeless situation. But it does require an understanding of the problem and greater education all the way around.
First and foremost, if there ever was a reason for lawyers to become educated about AI and its risks (and benefits), it is ironically to bolster the level of trust in the human side of things. A lawyer has to be ready to explain to the client not only why the bot is wrong when it is, but also that it’s inherent in the structure of LLMs to make mistakes and try to tell the prompter what they want to hear. And lawyers also need to be ready to tell clients before problems develop about the risks of creating discovery trails.
A lawyer can’t do all that without that knowledge themselves.
On the flip side, lawyers need to realize that GenAI tools often give sound answers. We can’t argue with the result of a prompt if the result is right. That will not breed trust much less yield good outcomes. There is a time and place for GenAI tools and lawyers must use them to their and their clients’ benefit.
All that being said, good lawyers know the law, they understand exposure, and they know how best to navigate the exposure. And now more than ever, they will need to understand their clients and to be adept at explaining all those things to their clients.
And know this: clients will have more information than ever before, so we better be on our toes. Gone are the days where a lawyer can just say this is what we are going to do and expect the client to accept it.
Years ago, I was called upon to explain the intricacies of class actions to a room full of insurance executives. I knew a lot about class actions already, but I spent hours practicing what I was going to say to a group that was a) skeptical and b) had no understanding of class actions and their peculiarities that often seem counterintuitive. At the end of the discussion, there was silence and then one of them said one word: brilliant. That cemented their trust in me.
In the days of GenAI, it is just that kind of trust, earned through preparation, knowledge, and understanding the client, that lawyers will need to earn by doing what GenAI can’t.
Stephen Embry is a lawyer, speaker, blogger, and writer. He publishes TechLaw Crossroads, a blog devoted to the examination of the tension between technology, the law, and the practice of law.