Businesses are using AI in a way that could create ‘catastrophic’ problems (Courtesy of the Jacksonville Business Journal) — Small-business owners and other professionals are saving thousands of dollars a year by using artificial-intelligence tools for legal work, but they could be setting themselves up for catastrophic trouble based on how they’re using the technology.
A new survey from tech venture Smallpdf found that 55% of professionals surveyed have used AI tools for drafting, editing or reviewing contracts. The average time savings was about four hours per week, or 26 workdays a year.
Ten percent of the respondents said they saved $10,000 in the past year by using AI instead of hiring legal counsel, according to the survey, while 47% said they closed a deal faster by using AI instead of waiting for legal review.
Those are the positives.
“The downsides of misusing or heavily relying on AI in legal work can be catastrophic,” said Star Kashman, technology attorney and founding partner of Cyber Law Firm, in an email.
Kashman said AI can be a helpful starting point — such as helping small-business owners with reviewing contracts or drafting basic legal documents — but it’s not a viable replacement for an attorney. Up-front savings might mean mistakes and costly errors that business owners have to pay for down the road.
Why? Because public-facing tools cannot access real-time case law or conduct legal research, and the propensity of AI tools to “hallucinate” can lead to made-up legal precedents or contract clauses — things that have already happened to attorneys in real life.
Even the AI tools available in the legal world have flaws, Kashman said.
“These AI tools may have access to case law, but they frequently misinterpret, misunderstand and misstate both the law, elements of laws, cases and case precedent, often wholly misinterpreting cases as supporting one idea when the case actually goes against that idea,” Kashman said.
More-robust AI tools may help organize or format more efficiently and could come up with bare-bones outlines and simple contract drafts that can save money, especially for hourly-bills cases, but they lack the ability to analyze legal nuances, tailor language to specific jurisdictions, or anticipate how some terms may be interpreted in court or play out in real life.
“For example, if a business uses AI to generate an operating agreement without attorney reviewal and guidance, it may unknowingly waive key rights or protections, opening the door to disputes that could and would jeopardize the entire company if disputes arise,” Kashman said. “In those cases, there was no point in having any operating agreement in the first place at all, given it did not protect a partner the way it is intended to.”
For business owners and professionals who don’t use AI for legal issues, according to the Smallpdf survey results, the biggest reasons are a lack of trust and digital and privacy concerns. Among those respondents, 68% said they would trust a contract written with AI tools, while 20% said they would not.
The respondents are mindful of how the technology is changing the workplace, though, as 74% said they would consider using AI in the future.
Complexities challenge AI systems
Among the survey’s additional findings, 36% of the professionals surveyed said they’ve had to revise or backtrack on a contract because of AI-made mistakes, including on the scope of work, definitions and terminology, and payment terms.
Wayne Unger, associate professor of law at Quinnipiac University, said one big risk with AI is that the business owners who use it don’t know what kind of data it was trained on. The better the data, the better the tool; the worse the data, the worse the tool.
“A small-business owner runs the risk of using AI that is trained using crap data and, as a result, the output of that AI is going to be crap,” Unger said.
The United States is also a complex legal system, and lawyers are paid to understand those complexities, such as a small business being subject to various federal regulations while also being subject to state and city ordinances. But if a contract dispute arises, the contract might be subject to common law and the Uniform Commercial Code, Unger said.
“While it is possible that an AI solution is trained to understand the complexities of the legal system in the United States, it is also possible that the AI solution is trained using data from Arizona that would not be binding on a small business in Massachusetts,” Unger said. “A lawyer can help a business owner sort through the mesh of legal requirements and obligations.”
Unger added that AI could be a good tool for small-business owners to learn more about various legal issues. For example, a business owner could ask AI to better understand indemnification.
Even with the growth of AI tools, Kirk Sigmon, a partner at law firm Banner Witcoff Ltd. who has worked in the area of cutting-edge AI technologies, said there are plenty of good “old-fashioned” templates that exist online that can be edited by anyone to use. There also are many affordable attorneys willing to take a quick look over a draft contract.
“In other words, that process can be extremely cheap and surprisingly reliable,” Sigmon said. “This is always the approach I recommend anyone take: Start yourself, and get one of the many hyper-talented but hyper-affordable startup-focused lawyers to get you across the proverbial finish line.”
AI-enabled contract-drafting systems, on the other hand, are almost inherently designed to introduce problems, Sigmon said. They cannot think through the logistics of contractual terms, and you never truly know if the output is appropriate or up-to-date.
“Simply put, people that exclusively rely on ChatGPT or similar tools to draft legal contracts are asking for trouble,” Sigmon said.
How companies are using AI
The proliferation of AI tools has come with the promise of increased productivity and often game-changing technology, but it also has come with significant issues, including “workslop,” the rapid spread of AI-generated content that looks good at first glance but ends up generating more work for others down the line.
The cost of the time that it takes fix “workslop” can add up, with a $186 monthly cost per employee on average, according to a survey of desk workers by BetterUp in partnership with the Stanford Social Media Lab. About 40% of the workers surveyed said they received “workslop” in the last month and that it took an average of two hours to resolve each incident.
There also is evidence that small businesses collectively are not racing to adopt AI tools. According to a recent Census Bureau Business Trends and Outlook Survey, which is conducted once every two weeks surveying 200,000 small businesses, overall adoption appears to have slowed in some cases and stopped in others. The percentage of small-business owners who report using AI to produce goods or services has grown since the survey launched in 2023, from an initial 3.7% to 9.7% in September, according to the survey, first reported by the Apollo Academy. The companies that are adopting the technology vary substantially by size, though.
Additionally, a recent report by the Massachusetts Institute of Technology’s NANDA initiative found that 95% of AI pilot programs fail to generate any measurable impact, whereas only 5% achieve rapid revenue acceleration.
Experts and industry watchers also are bracing for what they see as inevitable price hikes and cost increases as AI companies seek to cover losses and generate profit even as companies spend billions on AI infrastructure and systems.