Within a month of OpenAI’s release of GPT-4 in early 2023, Ad Idem Network presented a webinar on artificial intelligence that attracted more than 300 lawyers from across the United States and beyond.
The Continuing Legal Education (CLE) program provided practical information for attorneys addressing the impacts of AI in the workplace, according to Canby Wood, a lawyer and co-founder of Ad Idem Network, a global nonprofit networking organization for in-house counsel. This new reality—still uncertain—is being shaped by the emergence of large language models (LLMs), with GPT-4’s generative AI (GenAI) technology leading the way.
According to a statement from the International Monetary Fund, “AI is poised to reshape the global economy” and could eliminate a third of jobs in advanced economies. Legal services top the list of industries with the greatest exposure to AI’s impacts, according to a study by researchers at Princeton University, the University of Pennsylvania, and New York University. Goldman Sachs estimated that 44 percent of legal work could become automated.
Casey Flaherty, chief strategic officer for LexFusion, a New York City-based leader in legal innovation strategies, was the lead presenter during the webinar. In his presentation, Flaherty shared research projecting that, within five years, AI deployment will reshape 80 percent of corporate spending on core legal services.
He noted that, as early as the 1950s, lawyers were warned they would one day be replaced by robots, with waves of excitement about AI that often ended with the technology not meeting expectations.
“This time is different,” Flaherty said in an interview with Business Lexington. “This is not a gimmick.”
The advent of AI is a recent and evolving reality, and businesses of all sizes are working to determine how best to integrate the technology into their workflows without
compromising integrity, privacy, or their final products. Many potential applications for AI in the workplace are emerging as the technology is incorporated into new and existing software. But as with any disruptive technology, it must serve the user, not the other way around.
“We need to understand what [GenAI] can and cannot do today,” Flaherty said.
AI isn’t yet adept at generating precise answers, for example, especially in specialized applications, he said, but it can be a useful tool in research and task workflows. He emphasized that, while much of GenAI’s impact is currently at the individual level, it’s the aggregation of those individual gains that is driving improved organizational performance for organizations.
“If you’re a big firm, you should also be making investments in R&D to see how you can make system-level changes using GenAI,” Flaherty said.
Local legal firms are already looking at how to best integrate AI into their workflow.
Bruce Paul, a partner specializing in intellectual property and litigation at McBrayer law firm in Lexington, is on a Kentucky Bar Association committee preparing a CLE on GenAI for June 2025. “Everyone wants to learn about it,” he said. “The lawyer who is going to be out of work is the one who isn’t willing to understand these technological advancements.”
Richard Dawahare, a Lexington lawyer with a solo practice specializing in elder law, estate planning, and more, attended two sessions on AI at the American Bar Association’s annual convention last summer. “AI was a big topic of conversation there,” he said.
Both Paul and Dawahare have begun using GenAI in their work. McBrayer subscribes to Westlaw, a long-established legal research platform for the profession, while Dawahare subscribes to LexisNexis, a competing platform. Both services have added GenAI legal assistant tools.
“Any time there’s a legal question or issue, we can use AI to quickly get a summary of the law and the latest cases,” Dawahare said. “It’s a shortcut leading to further investigation and analysis. It’s not the answer we should be providing to our clients or the court.”
He also notes that care must be taken when using AI to protect client confidentiality.
“I do a lot of intellectual property work,” Paul said. “In any given intellectual property case, you could get thousands of emails and electronic documents. In the old days, if we asked for those emails, we’d get them as pieces of paper to flip through and read. Artificial intelligence allows you to distill it down with a few keystrokes.”
The accuracy and efficiency of that summary, however, are directly affected by how and where the data is organized. Flaherty emphasizes that a well-organized, consistently used data and communication system is critical to making the most of the technology.
“People are now feeling the pain of their poor data practices,” he said. “They’re seeing more opportunities to leverage data, but they’re not well positioned to do so. [Information is] spread out over all kinds of repositories — in people’s emails, in notebooks, in file folders. It’s a mess.”
Flaherty recommends not starting with the tools but with the problem to be solved. The more important questions are: What am I trying to build? Where can I find greater efficiencies? Rather than, “Hey, someone just invented a chainsaw — what can I cut?”
GenAI has also made its way into public healthcare. Dr. Romil Chandha, chief medical information officer at UK HealthCare, leads a team of informatics physicians studying information and communication systems to strengthen provider-patient relationships and improve health outcomes.
Chandha explained that AI has been used in the medical field for decades, and recent advancements in microprocessors and computing power have increased the accuracy of AI systems. Vendors of electronic health records (EHRs) are preparing to market AI tools based on national and global data sets to healthcare providers.
“If I have a patient, Mr. X,” said Chandha, “and he has hypertension, age of 45, doesn’t have diabetes or kidney disease — if I start him on this one type of medication, what outcome will I see, versus if I start him on a second kind of medication?” He said such predictive AI tools are coming.
In both the medical and legal industries, large language models (LLMs) need training with field-specific data to be more useful and reliable. For example, medical LLMs are
trained with medical journals, EHRs, and textbooks, while legal LLMs are trained with legal documents, case law, and statutes.
Chandha pointed out that it’s important to consider the population on which an LLM is trained. “Eastern Kentucky, Central Kentucky, and Western Kentucky may all have different data sets,” he explained. “And Kentucky as a whole will definitely differ from large metropolitan areas like New York or Los Angeles. Whatever we implement, we need to validate it on our own population.”
In 2019, the American Medical Association established the AMA Joy in Medicine program to help prevent doctor burnout and bring more joy to the practice of medicine. Chandha said ambient voice GenAI holds promise in this area. With patient consent, discussions between physicians and patients can be recorded, freeing doctors from note-taking and data entry and allowing for a more natural conversation.
GenAI can filter out “noise” and generate a summary for the doctor to review and edit. Another feature gaining popularity is the crafting of follow-up messages for patients, which can be reviewed and sent out by doctors or nurses. Several academic medical centers are running pilot programs with this technology. Chandha said UK HealthCare is currently reviewing the technology.
Cornett, a Lexington-based ad agency, has also integrated GenAI into its workflow. “Generative AI tools are like having a really fast intern,” said Peter Cook, director of interactive strategy at Cornett. “They can do quick drafts of things, but it’s not ready [to present to clients].”
It all needs the human touch, he said, however AI “does help me to do some of my more mundane tasks more quickly and efficiently, so I have more time for creative and strategic thinking.”
“We’re not replacing humans with AI in any of the work Cornett’s doing,” Cook said. The agency relies on photographers, writers, and video crews to create its products. AI is mainly useful for creating “mock-ups or early comps” to present project ideas to clients. “We have no desire to replace people. We want people to use their time efficiently, and we want to have plenty of work that only real creative humans can do.”
“We’ve definitely run up against the limitations of AI tools in many ways,” said Jason Majewski, group creative director at Cornett. “Certain nuances just don’t compute yet. We’ve generated a lot of monstrosities. It’s interesting to see what the tools haven’t quite mastered yet.”