The rapid adoption of artificial intelligence tools across the legal profession is forcing clients and firms alike to ask a fundamental question about who is really doing the work.
Jeffrey K. Cassin of Norris McLaughlin P.A. has addressed the question directly, offering what he describes as a lawyer’s answer: it depends.
Horror stories continue to circulate about attorneys at top law firms submitting court documents containing AI-generated “hallucinations” that were not caught before filing.
These failures have eroded trust in AI tools within legal practice and prompted the American Bar Association to issue model rules and ethical guidance on how attorneys should engage with such technology.
Beyond hallucinations, a broader concern has emerged around the risk of legal documents becoming indistinguishable from one another as AI tools homogenise output across the profession.
Kirkland and Ellis Chair Jon Ballis has noted that widely available AI tools are creating a baseline floor for services, adding that his firm does not get hired for the floor.
Despite these concerns, the ABA has pointed to a study suggesting a 69% adoption rate of AI tools by legal professionals, indicating the technology is now firmly embedded in the fabric of legal services.
Cassin draws a clear distinction between what AI tools can and cannot do, describing them as comprehensive data sorting and word-processing tools that create natural language outputs rather than genuine intelligence.
“They do not innovate or exercise judgment, they use complex word-association predictions to determine the most probable (but not necessarily true) response to your query,” Cassin wrote, arguing that innovation, creativity, and judgment remain the lawyer’s responsibility.
AI tools can assist with legal research, due diligence, data organisation, document management, drafting, and comparative analysis, but each of those tasks still requires professional judgment to execute properly.
Cassin also warns that free, open-source versions of these tools may affect attorney-client privilege, making legal-specific AI products a safer choice for practitioners handling sensitive matters.
The term artificial intelligence is itself questioned by Cassin, who argues it is a misnomer given that no actual intelligence is involved, preferring instead to call these systems learning language models, or LLMs.
Clients will ultimately want their attorneys to be efficient and reliable, but the message from practitioners is clear: the attorney’s intelligence and judgment still matter most.