A couple of characteristics stand out when two William & Mary computer science professors are asked to describe Lingfei “Teddy” Wu Ph.D. ’16. • “I remember Lingfei as tenacious,” says Evgenia Smirni, computer science department chair and Sidney P. Chockley Professor. • “He will work day and night to succeed in any task,” agrees Andreas Stathopoulos, Wu’s doctoral program advisor. “And while scientists are skeptics by training — or by nature — Lingfei brings a rare optimism to the table. This powers him through any adversity and he always succeeds.” • Wu draws on both qualities as the cofounder and CEO of Anytime AI, a new startup that ventures into the developing frontier of artificial intelligence (AI) in the legal field.
Channeling the Force of AI
Lingfei ‘Teddy’ Wu Ph.D. ’16 sees the potential for artificial intelligence to transform the practice of law
January 30, 2025
By
Tina Eshleman
Illustrations By
Katie Thomas
Hearing the professors’ descriptions, Wu laughs. “I would use a different word. I call it persistence when you feel your inner calling about something and you have a passion to do that,” he says. “If you want to be a top expert and do it well, you need a lot of persistence and hard work.”
‘A Present Necessity’
After spending the early part of his career leading a research team at IBM’s Thomas J. Watson Research Center focused on machine-learning tasks such as speech recognition, Wu moved to more directly consumer-oriented companies such as Pinterest.
Feeling the pull of entrepreneurship, he decided to apply his knowledge of machine learning to create a platform that uses artificial intelligence to help law firms work more efficiently. He launched Anytime AI, marketed as “the premier AI legal assistant for plaintiff lawyers,” in October 2023.
Defined as technology allowing computers to perform complex tasks that typically have been done by humans, AI is both celebrated and feared as it increasingly permeates 21st-century life. In a TEDx presentation last July, Wu described AI as the most recent of two dozen technologies that have revolutionized human society over the past 10,000 years — among them wheels, printing, railways, electricity and the internet.
“We are standing at the threshold of another transformative era,” he told the audience at California Science and Technology University. “AI will bring unprecedented efficiency and capabilities to every industry, including the legal industry. Adoption of AI is no longer a future consideration but a present necessity.”
Wu sees AI as increasing efficiency for law firms, meaning they can do more work without adding staff members. He is optimistic that this will make legal services accessible to more people by reducing the costs for clients.
As an example, he says a legal service that typically would require four lawyers working 10 hours a day for six weeks at $400 per hour could be done using AI with three lawyers in three weeks working eight hours a day. Even if the firm increases its hourly rate from $400 to $500, the overall cost for a client would drop from $480,000 to $180,000.
While it might take a lawyer four hours to summarize 100 pages of depositions, this task can be done in one minute using AI, including highlighting key facts and citing original page numbers, he says. His company, Anytime AI, works with many personal injury firms, and he says that AI can analyze 1,000 pages of medical records and extract key information in 15 minutes — something that would take a person at least 15 hours.
According to an October 2024 Wolters Kluwer survey, 76% of legal professionals in corporate legal departments use generative AI at least once a week, as do 68% of their counterparts in law firms. Looking ahead, 58% of law firms and 73% of legal departments plan to increase their investment in AI over the next three years.
Some legal experts have warned that AI can be unreliable, noting a tendency of large language models to fabricate or “hallucinate” information. In one notorious case, a New York personal injury lawyer was fined after he submitted a legal brief using ChatGPT that contained false citations. A Stanford University study released in 2024 found that legal hallucination rates range from 69% to 88% in response to specific queries. In his 2023 Year-End Report on the Federal Judiciary, Chief Justice John Roberts emphasized that “any use of AI requires caution and humility.”
To ensure that Anytime AI is providing trustworthy results, Wu says the company has developed its own technology for retrieval-augmented generation, or RAG.
“Every time we return an answer back to a client, they not only have an answer from AI, but also a citation link,” he says. “If you link to the citation, it will automatically open the corresponding page in the document and highlight the text block for you, so you can verify the answer.”
Fredric Lederer, Chancellor Professor of Law and director of William & Mary’s Center for Legal & Court Technology (CLCT), sees AI as presenting tools — such as summarizing lengthy depositions — that are of great value for lawyers and law students. But he says AI also means that some areas of legal practice are likely to disappear.
For example, JPMorgan began using AI to review commercial loan agreements, “completing in seconds what used to take 360,000 hours of lawyers’ time over the course of a year,” Time magazine reported in 2020.
“Jobs all over the world, legal and nonlegal, are going to be affected,” Lederer says. “When you go to law school now, you go with the knowledge that a whole lot of things that people used to do, often as their primary legal tasks, will no longer exist. That’s going to change the nature of legal work, the nature of the legal profession and ultimately the nature of legal education.”
Repetitive tasks such as preparing an uncomplicated last will and testament, for example, could be done using AI.
“At our law school, we are proud that we are taking steps to prepare our students for the future,” he says. One way of doing that is through teaching how to use generative AI accurately and ethically for legal writing. Laura Killinger, director of the law school’s Legal Practice Program and clinical professor of legal writing, addresses topics such as how to ensure that data included in a generative AI prompt does not inadvertently expose confidential information. Other courses cover the interaction of law and technology.
The CLCT and its McGlothlin Courtroom allow for testing of new legal technology, including virtual reality evidence, holographic evidence and 3D evidence. Begun in 1993 as the Courtroom 21 Project, CLCT is a joint initiative of W&M Law School and the National Center for State Courts. Originally, CLCT’s emphasis was on using technology to improve dispute resolution, but during the last decade, the center has expanded into the impact of artificial intelligence and cybersecurity on the legal system and legal professionals. Via the CLCT Court Affiliates, the center supports about 44 courts in the United States and Canada. In addition, CLCT designs facilities all over the world.
“Our courtroom is usually considered the world’s most technologically advanced, with technology loaned from companies all over the world at no cost,” Lederer says.
Working with CLCT, Anytime AI has offered W&M law students and faculty — most immediately the 30 law student fellows who work with CLCT — free access to the company’s platform.
Asked what the future holds for law students, given the pace of technological advances, Lederer reflects on what human beings can do that artificial intelligence cannot.
“AI is based on data. One of the things that AI can’t do particularly well at the moment is to offer human insight,” he says. “The legal system changes constantly, not just based on what laws have been created or on past court decisions, but also because people come up with new ideas based on human values. For example, it’s hard to see AI in the near future determining what would constitute justice from a human perspective.”
Solving Problems
A native of China, Wu is the son of a bus driver and a homemaker who had training as a community health care provider — referred to as a “countryside doctor,” he says — but stepped away from that role to take care of him and his two siblings.
Though they did not attend college themselves, Wu’s parents emphasized the importance of education and made sure their children dedicated time to their studies. Wu became fascinated with physics at age 7. He recalls poring over a textbook and writing notes and questions. For example: Why do you see a shadow after looking at a bright light?
“I would try to describe that as a science problem,” he says. “That’s how I started my journey. I loved to study phenomena and ask, ‘Why is that?’ and then try to find an explanation.”
He learned about W&M while attending the University of Science and Technology of China, after discovering research by Class of 2027 Professor in computer science Gang Zhou, who is also from China, on wireless sensor networks and how electronic devices communicate through low-power networks such as Bluetooth and WiFi.
“While in China, I had done intensive research on wireless sensor networks,” Wu says. When he learned that Zhou was a professor at William & Mary, Wu looked up information about the university.
“People in China generally want to go to bigger schools in the United States,” he says. “But I looked at the William & Mary website and thought, ‘Wow, this is a beautiful school with a very long history.’ I was very attracted to William & Mary and thought, ‘Why not submit an application?’ I’m very lucky that I got admitted with a full scholarship.”
He and his wife, Fangli Xu M.S. ’18, met at Anhui University in Hefei, China, where he studied electrical engineering before receiving a master of science in automation from the University of Science and Technology of China. They married in 2012, about two years after he began his Ph.D. program at William & Mary. She followed him to the university, where she also pursued graduate studies in computer science. Xu now works with her husband at Anytime AI.
At William & Mary, Wu shifted his academic focus to computational mathematics and scientific computing. His dissertation was on extracting knowledge from large-scale data applications.
Wu held leadership positions as vice president of the Computer Science Graduate Student Association and vice president of the Chinese Students and Scholars Association at W&M.
He says his interactions with William & Mary professors prepared him well for a career in computer science.
“The faculty in the department are top researchers, so they have very high standards,” he says. “To solve a problem, you have to put a lot of effort into studying, doing the background research and then thinking about what a potential solution is and how to implement that. They provided a model for how to truly understand a subject.”
Wu particularly credits guidance he received from his advisor, Stathopoulos.
“Every time that I wanted to have a conversation with him, I would simply knock on the door,” Wu says. “He would stop his work and have a meeting with me to discuss the thing that I needed help with. And then, gradually, I was able to propose some solutions that truly impressed him.”
The process of working through problems and finding solutions that he learned at William & Mary is one he carried forward after graduating.
“I have problems every day,” he says. “I have to think about what the existing solution is and what is a potential solution I can come up with.”
Knowing he wanted to run his own company someday, Wu began a career in industry rather than pursuing a path as an academic researcher. He took a job at IBM, where he had done an internship during his Ph.D. program.
Wu routinely sets five-year goals for himself, and one of those was to obtain patents.
“I had a very good idea for how to improve one of the fundamental algorithm methods in machine learning, and I felt excited about that,” Wu says. He asked colleagues who had received patents for advice on how to pursue one.
“I realized that to qualify for a patent, you need not only an innovative idea, but you also need to have an innovative system that can employ your idea and then make it practical,” he says.
Wu had many more good ideas, and he ultimately received 65 patents for his inventions while working as a researcher at IBM. Four of those received recognition as a “high value patent,” meaning it has earned over $1 million for the company. One of those translates a database language, MySQL, to natural language and vice versa, making it easier for people to work with computer systems.
In 2021, Wu set a new five-year goal for himself: Build something truly useful to society. “I wanted to see technology having an impact,” he says.
Feeling he had accomplished what he set out to do at IBM, he transitioned to working as a principal scientist for JD.com, a consumer products company similar to Amazon that is popular in China. There, he developed a virtual shopping assistant to guide customers toward products.
“It’s very similar to ChatGPT, but it was two years earlier,” he says. “That made me see that this is the future: In each domain, we will have a virtual assistant to help us do things beyond shopping. Can I have a virtual assistant help me book a flight ticket? Can a virtual assistant give me financial advice?”
When JD.com closed its office in the United States, Wu moved to Pinterest, where he led a team of applied scientists, software engineers and product managers to leverage large language models and generative AI technologies to recommend content for users who engage with the social media platform. At Pinterest, Wu began using ChatGPT to more quickly generate computer codes and to create LinkedIn posts and emails.
Meanwhile, he assembled a small group of advisors from his professional contacts to discuss potential business ventures.
“I could see that this technology will fundamentally change human daily life in the next 10 years,” he says. “I told my team we have to act now.”
Because review of documents is such a large part of legal practice, the field of law seemed to present fertile ground for the application of AI technology, Wu says. His team at Anytime AI includes several former IBM colleagues as well as William & Mary Law School alumna Xue Chang J.D. ’14 and James Knicely, a Williamsburg-based attorney.
“It’s really about how we can customize a large language model into the legal workflow so that the lawyer and paralegal can greatly accelerate their tasks,” Wu says. “AI is able to extract information in a few minutes, and then you just need to review to validate the results.”
Making an Impact
When Wu decided to start Anytime AI, he first pitched his services to about 200 law firms around Armonk, New York, where he lives.
Among the law firm executives he met with were several William & Mary alumni, including Wenjie Sun LL.M. ’12 and her husband, Shengyang Wu LL.M. ’12. Sun is the co-founder and partner of Sun & Ren PLLC, which primarily focuses on business immigration and represents artists and musicians with obtaining visas. Her husband, who is not related to Lingfei Wu, is a founding partner of Alpha Law whose practice focuses on personal injury litigation and investigations.
“He contacted us for brunch, and he told us he’d started his own company, and he was making AI software for lawyers,” Sun says. “We were very early users of his software, and we provided feedback for him.”
Over the past year, Sun has been using Anytime AI software for translating and summarizing lengthy documents. Her husband finds it helpful in comparing two documents and pointing out the differences when, for example, the opposing party’s lawyer wants to make a change. For Sun, Anytime AI has increased efficiency, eliminating the need for another part-time bookkeeper.
She and her husband also became investors in Anytime AI.
“We actually invested in the person, not the company,” she says. “We believe in Lingfei not only as an alumnus, but as a professional in this area.”
Wu is always looking for ways to improve the business. Ultimately, he envisions making legal services more affordable and helping connect people to lawyers who can meet their needs.
He often reminds himself of a quote from author Roy T. Bennett: “Success is not how high you have climbed, but how you make a positive difference to the world.”