Commentary: This article presents the author’s views and arguments. It is not advice about a particular case.
A lawyer has only so many hours in a week. Hearings, client calls, research, drafting, and running an office all compete for that time.
Law firms have always dealt with this limit by sharing the work. Associates research cases, paralegals prepare drafts, and assistants organize records. The attorney reviews the work, makes decisions, and takes responsibility for what goes out under the attorney’s name.
Artificial intelligence changes how quickly some of that work can be done. The duty to check it remains.
Used well, AI gives a lawyer a much stronger starting point. Without careful review, it can make a bad legal argument sound convincing.
What legal AI offers
There is a difference between AI built for legal work and a general tool such as ChatGPT.
Lexis+ with Protégé combines research and drafting tools with LexisNexis legal sources. Its Shepard’s tools help lawyers check citations and see how courts have treated a case. That matters because a real case may have been reversed, limited, or rejected on the point being argued. LexisNexis product information
Thomson Reuters CoCounsel Legal brings together AI tools with Westlaw and Practical Law content. Its features support legal research, document review, and drafting. The exact tools depend on the product and subscription. Thomson Reuters product announcement
ChatGPT is a general tool. It can help organize facts, compare documents, explain difficult language, and prepare drafts. With web search, it can also find current sources. But a basic chat should not be treated as a complete legal database or a service that checks whether a case remains good law.
The choice should depend on the task. For finding and checking case law, legal sources and citation tools offer a major benefit. For turning rough notes into a clear letter, a general tool may be enough.
Research can move much faster
Before online research, lawyers worked through books, indexes, and printed case reports. Online databases made that process faster. AI adds another step: it can help search, group results, summarize opinions, and compare arguments.
A research task that once took hours may now produce a useful starting memo in minutes.
Still, searching a database containing thousands of cases is different from carefully reading thousands of opinions. Speed alone does not show that the research is complete.
The lawyer must open the cases, check the cited passages, and decide whether the law fits the facts. Does the case come from the right court? Does it address the same issue? Is it still good law? Are there cases pointing the other way?
Compiling data in minutes
A recent client brought me almost 500 pages of raw data. In under two minutes, AI produced an initial breakdown by category, calculations, and a summary.
Instead of starting with hundreds of pages, I had an organized starting point for review. That let me focus sooner on what the records meant for the client’s case.
Checking the source records remains important. Missing pages, repeated entries, or a misread number can change the totals. But the time saved in preparing that first breakdown was substantial.
Drafting is easier than choosing what belongs
AI can prepare first drafts of many legal documents. Letters, agreements, discovery requests, and motions can all benefit from that help. But a document can sound right and still change the deal.
Imagine a homeowner who wants to let a neighbor use a driveway for six months. Asked to draft a simple agreement, AI adds a permanent easement—a lasting right to use the property that may continue after either home is sold.
The draft sounds professional, but it grants far more than the owner intended. The lawyer must catch the difference between short-term permission and a lasting property right.
Extra provisions can look thorough while creating duties the client never agreed to. A good review asks more than, “What is missing?” It also asks, “What should be removed?”
Finding the missing support
Suppose a business receives a demand for payment under a service contract. The letter sounds firm and cites several contract terms, but it never explains how the amount was calculated.
AI can help compare the demand with the contract and invoices, flagging missing support or charges that need a closer look. That gives the lawyer a focused question: Which charges do the records actually support?
A missing explanation does not prove that nothing is owed. It shows where to ask for details. AI helps organize the comparison; the lawyer evaluates the claim and decides how to respond.
Clients are using AI too
Clients now arrive at initial consultations with AI printouts explaining why they should win. These can help organize their concerns and questions. But AI may build a convincing argument around a law that does not apply to their situation.
The answer may rely on another state’s law, a different kind of claim, or facts the client has not established. If the client’s question leaves out an important fact, the answer may rest on a mistaken view of the dispute.
The lawyer must sort out what applies, what is missing, and what the evidence supports. Sometimes that means explaining why the client’s research does not lead to the result the client expected.
That work has always been part of a consultation. AI gives clients another way to prepare and lawyers another set of assumptions to examine.
Law schools are adapting
Today’s law students will enter a profession where AI is already part of daily work. Learning to use it is becoming part of their training.
A 2026 Thomson Reuters survey of more than 1,800 law students found that 72 percent considered AI literacy essential. At the same time, 74 percent recognized the risk that relying too heavily on AI could weaken their own legal skills. 2026 Law Student Pulse Survey
At the University of Missouri, an AI chatbot plays a difficult opposing lawyer so students can practice negotiation. The exercise gives them a chance to test their judgment against a demanding adversary. Missouri’s negotiation chatbot
Students need practice using these tools while developing the knowledge to recognize when an answer does not fit.
Negotiation requires judgment in the room
Imagine a lawyer who enters a settlement negotiation with an AI printout that says, “Do not accept less than this amount.” The lawyer sticks to that number even as new information changes the risks and value of the case.
In this example, talks fail. After more expense and delay, the dispute settles for less than the earlier offer. Both clients have spent time and money that a more flexible discussion might have saved.
A prepared position matters. But negotiation requires listening, asking questions, and responding to what happens in the room. Payment terms, timing, legal risks, and the cost of continuing the fight can change the value of an offer.
An AI-generated number may be a useful starting point. It does not become a sound settlement position simply because it appears on a printed page.
AI can help compare offers, estimate costs, and prepare questions. The lawyer weighs that advice against the facts and the client’s goals. The client decides whether to settle.
Professional duties still apply
The American Bar Association addressed generative AI in Formal Opinion 512, issued in July 2024. It explains how existing duties apply, including competence, confidentiality, communication, supervision, and reasonable fees. Lawyers must understand enough about the tools they use to evaluate the risks and check the work. ABA Formal Opinion 512
Florida’s Ethics Opinion 24-1 also permits lawyers to use generative AI while requiring safeguards. Lawyers must protect client information, oversee the work, and charge reasonable fees. Before entering private client material, the lawyer needs to understand how the service handles that information and whether client consent is required. Florida Bar Opinion 24-1
Courts expect lawyers to check the work
In Akerlund v. Atlas Air, Inc., decided July 10, 2026, the Eleventh Circuit addressed briefs containing nonexistent cases. The court stressed the need for lawyers to verify AI output before it reaches a court filing. Court opinion
A lawyer’s signature is not a promise that no error could ever occur. It does carry professional duties to investigate, review, and support the document.
An invented citation is an obvious failure. A real case used for the wrong legal point can be just as damaging. Review must go beyond checking that the case exists.
The effect on staff and fees
AI puts pressure on work long handled by legal assistants and paralegals. Routine summaries, document comparisons, and early drafts may take fewer hours. That could allow a small firm to handle more work without adding as many staff.
The Bureau of Labor Statistics projects little or no employment growth for paralegals and legal assistants from 2025 to 2035. It identifies AI as one factor that may limit demand. Skilled staff still bring value through client contact, case management, factual review, and careful oversight. BLS employment outlook
Faster work also affects billing. Under the ABA’s guidance, a lawyer billing by the hour must charge for actual time worked. A task that once took four hours does not justify billing four hours when it now takes one. Reasonable time spent checking and revising the result remains part of the work. Flat fees must also remain reasonable. ABA Formal Opinion 512
The lawyer remains responsible
The profession has long accepted that lawyers can rely on help while remaining responsible for the result. AI extends that practice at a speed that changes how much work a lawyer can handle.
It can help find useful law sooner, organize hundreds of pages, prepare a draft, and spot a weakness that deserves attention.
The lawyer must understand the client’s goal, choose the right legal approach, and recognize what does not belong. That judgment matters most when facts change or a conversation takes an unexpected turn.
When my name goes on the document, the responsibility remains mine.