At the Inaugural AI for Legal Aid Summit, A Call To Arms for Adopting AI To Tackle the Justice Gap

Generative AI presents a generational opportunity for legal aid organizations to address the justice gap, keynote speaker Laura Safdie told the roughly 350 people who convened in Phoenix on Sunday for the inaugural AI for Legal Aid Summit.

Noting data showing that more than 90% of civil legal needs go unrepresented, Safdie (pictured above), former Casetext cofounder and current head of innovation for legal at Thomson Reuters, which acquired Casetext and its generative AI legal assistant CoCounsel in June 2023, said that she sees a transformative opportunity for legal aid organizations to leverage generative AI to address the justice gap.

She said that the release of GPT-4 marked a significant turning point, enabling the development of AI solutions that can be reliably constrained to specific sources of legal knowledge – which she called a crucial development for professional legal work.

“If the legal aid community is empowered with access to AI, the knowledge of how to use it, the learnings and best practices developed by peers, … then we might be living through a generational opportunity to make significant progress against the justice gap in America,” Safdie said.

Safdie’s kick-off keynote characterized the sentiment of those who attended the six-hour summit, where speakers emphasized the transformative potential of gen AI  for legal services organizations, while stressing the need for careful implementation and quality control standards.

This was the first such summit, jointly organized by Kristen Sonday, founder and CEO of Paladin, and the Legal Services Corporation, and presented on Sunday as a pre-conference adjunct to the LSC’s Innovations in Technology Conference, which began the following day.

The five-hour program was designed to teach attendees the basics of AI technologies and help them understand how AI might amplify legal aid services. It included live demonstrations of AI tools already in use by legal services organizations and even a hands-on training that walked attendees through the steps of building their own GPT.

During her keynote, Safdie outlined three primary ways AI can amplify legal aid’s impact:

• Supporting direct legal work, such as through faster document review and analysis. She said AI is as skilled as a junior associate, but can complete work much faster and with greater thoroughness and accuracy.
• Helping manage organizational operations, including in areas such as grant writing, marketing and human resources.
• Through its collective impact on legal services organizations as it transforms how work gets done and levels the playing field.

Even so, she cautioned organizations against using consumer AI tools such as ChatGPT, urging them instead to choose AI solutions specifically developed for legal professionals that are grounded in authoritative content and that have robust security and privacy protections.

Safdie closed by urging organizations to get started right away. “Begin today,” she said. “AI is too powerful to ignore and its potential impact too great.”

Call for Quality Standards

In a separate talk, Margaret Hagan, executive director of the Stanford Legal Design Lab, addressed the need for developing common quality standards for AI implementation in legal services. She highlighted the resource-intensive nature of AI adoption and maintenance and the fact that AI systems can depreciate over time without proper oversight.

“It’s not like you hit play and then all of a sudden you can be resting assured that the quality will be there,” Hagan said, calling for a coordinated national effort to establish quality evaluation protocols for common legal aid AI use cases, from intake systems to document assembly tools.

Far too commonly in the delivery of legal services, Hagan said, we evaluate quality by the standard, “We know it when we see it.” But that mindset needs to change.

“It’s time to go beyond that, to have common, well-documented protocols about what makes something good or bad when it comes to legal services,” she said. “Right now, those standards are often quite blurry, and it’s time to get concrete, because that infrastructure is going to be the necessary piece to actually making AI work at scale.”

Her team at Stanford has already begun this work, she said. They spent over a year looking at the different use cases for AI in relation to access to justice. Then based on input from end users and justice professionals, they developed evaluation criteria for AI-powered legal question-and-answer systems.

“We need to be doing similar kinds of user-centered and expert-centered research for all of the different tasks nationally,” Hagan urged. “We need to be thinking about this for the most common AI agents that are coming down the pike for access-to-justice use cases, for the intake interviewers, for the document assembly tools, for practice management, for things that draft narratives. How do we have a common way to know if these AI agents are doing a good or bad job?”

AI In Action

Summit organizer Sonday wanted it to have a practical focus, and that was reflected in two o f the day’s sessions, each of which included live demonstrations of AI tools already in use by legal services organizations. These included:

• Scheree Gilchrest, chief innovation officer at Legal Aid of North Carolina, demonstrating LIA, the legal information assistant.
• Conor Malloy, director of innovation at CARPLS, demonstrating Rentervention.
• Sateesh Nori, senior legal innovation strategist at Just-Tech, demonstrating Roxanne (the launch of which I covered last week).
• Adrian Palma, global pro bono manager and digital strategist at Microsoft, demonstrating the DACA CoPilot, a tool that helps with DACA renewals.
• Michael Semanchik, executive director at The Innocence Center, showing how he uses CoCounsel to streamline his work in reviewing innocence cases.
• Zachary Oswald, senior deputy director at the Legal Aid Society of Middle Tennessee and the Cumberlands, showing how they are using ChatGPT to help expunge criminal records.

With a focus on practical skills, the conference also included sessions on developing AI use cases and writing effective prompts, as well as a workshop in which Ransom Wydner of SixFifty walked participants through an exercise in creating a GPT.

While all the day’s speakers agreed that AI offers significant potential benefits for helping legal services organizations better serve their clients, they also delivered the message that success for these organizations will depend on strategic adoption, proper evaluation frameworks, and ongoing diligence.

The bottom line is that Kristen Sonday deserves credit for convening this conference. The fact that 350 people – nearly half of all those who registered for the ITC conference – showed up a day early to attend this summit demonstrates that many are thirsty for more knowledge about how AI can help them help their clients.