More and more of my clients come to our meetings having researched the relevant legal concepts or processes on ChatGPT, Claude, or another general-purpose AI tool. (I mean general-purpose chatbots here, not legal-specific AI platforms.) This in itself is positive, since a client who understands the terminology or issues at hand typically brings more targeted questions to the conversation. I can spend less time on the basics and more time discussing how the matter relates to their business’s priorities and risks, and we can often get to a decision faster. Whether AI ends up costing a client more comes down to how they use it, not whether they use it. Used judiciously, it benefits both the client and our working relationship. Used without judgment, it does the opposite. Some clients view the chatbot’s output as a finished analysis rather than a starting point. They come to our meeting already having formed conclusions based on information that has not been validated. When this happens, meetings often become less productive and more costly for clients, since I must undo their thinking before our real legal work can begin.
Informed vs. Pre-Loaded Clients
In my experience, there is a meaningful difference between informed clients and pre-loaded clients. An informed client is one who uses AI to understand legal concepts, terms, or processes well enough to formulate good questions and get more out of the conversation. That kind of preparation makes meetings more efficient and allows me to spend more time on judgment, strategy, and advice rather than getting clients up to speed on basic concepts.
Pre-loaded clients are a different story. They use AI not simply to understand the issue, but to work through the issue and reach a conclusion prior to our meeting. By the time we talk, the client may have formed an assumption about how the law applies, a position on a contract provision, or a conclusion about the strength of their case. The problem is not that my client has an opinion. The problem is that the conclusion may be based on a set of assumptions that the chatbot made without knowing the full circumstances.
A general-purpose AI tool does not know what it does not know. For instance, it does not know everything about a client’s business, risks, or objectives. It may miss important context or make assumptions about facts that were never provided. It may also apply too much weight to certain details, not understanding that they are in fact less significant under the particular set of circumstances. All too often, AI tools reach conclusions that sound authoritative without being right. A client’s instinct to get ahead of the problem is a good one; it is simply aimed at a tool that is not equipped to finish the job it started.
When AI Research Creates More Legal Work
Clients who use AI for legal analysis often become strongly attached to the chatbot’s conclusion. When this happens, I cannot simply set aside the client’s beliefs and seamlessly move on to the legal work at hand. I learned this the hard way. Early on, when a client arrived with AI-generated analysis that was off the mark, my instinct was to engage with it head-on and rebut it point by point, the way I would argue against opposing counsel. It took time and energy, along with a particular kind of lawyer angst that comes from arguing against a machine rather than a person or a position. And it didn’t work. It took me a while to see why: the client wasn’t anchored to the AI’s reasoning, but to its conclusion. Picking apart the reasoning left the conclusion standing, and we were no better off for the effort.
What works better is to reframe rather than rebut. I now treat the AI’s output as one input among several and redirect the conversation to the actual facts, the missing context, and the real question at hand, rather than relitigating the chatbot’s homework. Even so, correcting assumptions, surfacing missing facts, and explaining what the law actually requires is real work, and none of it is optional if the advice that follows is going to be right.
All of that takes legal work, which is how a client’s injudicious AI use can directly increase legal fees. It’s not AI use itself that increases legal fees; it’s misuse. A client who uses AI to get oriented adds no cost. A client who treats its output as a finished, reliable answer turns a straightforward engagement into a longer, more complex, and more expensive one, because so much ground must be re-covered before the real legal work can begin. That is where and how AI use increases what a client pays, and it is avoidable.
Confidentiality Issues
Clients also need to be cautious with what they input into an AI tool. Unlike communications when seeking legal advice from counsel, a client’s conversations with a general-purpose chatbot are not protected by attorney-client privilege. When clients use the chatbot to work through sensitive facts, they may be unintentionally creating a record that later becomes relevant to discovery, potentially adding cost and complexity to their matter.
In general, clients should avoid inputting confidential company information or personally identifiable information into a general-purpose AI tool. That is true whether or not the client is paying for the service. A subscription may change how the product operates or handles data, but it does not give the client the same confidentiality protections as communications with counsel. Both cautions are further ways that careless AI use can add cost and complexity to a matter.
Going Forward
This is not to suggest that clients should stop using AI to prepare for conversations with their attorney. Keep using it to get smart before the conversation. Used judiciously, AI can help clients understand legal concepts, ask better questions, and make meetings more productive. The key is to use AI to prepare, not to pre-decide. Getting oriented is different from arriving with the analysis already settled in your head. A chatbot’s plausible-sounding conclusion does not equate to sound legal advice, much less advice tailored to a client’s specific facts, risks, and objectives.
The fee point is simple. Judicious use costs nothing extra and helps; injudicious use is a direct, avoidable driver of higher fees. A practical tip: bring the AI’s output to your attorney as a starting question, not a finished position, and never enter confidential facts, personally identifiable information, or matter-specific details into a general-purpose AI tool, subscription or not.
The smartest clients I work with aren’t the ones who never use AI. They’re the ones who know what it’s for.
Mendi Sossamon, a Partner at Potomac Law (PLG), crafts legal solutions to help businesses grow. She leverages more than 27 years of legal experience, which includes a stint in Big Law, running her own law firm, and serving as a Deputy General Counsel for a billion-dollar company. Based in Austin, Texas, Mendi focuses her practice on outside general counsel services, complex corporate and commercial transactions, intellectual property strategy, and commercial optimization.
