Legal Aid of North Carolina deploys AI virtual assistant to address civil legal needs gap

Legal Aid of North Carolina (LANC) operates at the frontline of the United States' persistent justice gap — the disparity between civil legal needs among low-income Americans and the resources available to address them. Nationally, 92% of substantial civil legal needs go unmet, a figure that defines the scale of the challenge. LANC's existing website offered limited self-service capability, forcing clients to navigate complex legal situations without guidance. Demand concentrated heavily in high-stakes areas — domestic violence, child custody, landlord-tenant disputes, and consumer law — routinely exceeded organizational capacity. The result: vulnerable populations left without access to basic legal information at moments of critical need.

In July 2024, LANC launched LANC-LIA (Legal Information Assistant, pronounced 'Leah'), an AI-powered virtual assistant built within LANC's Innovation Lab in partnership with LawDroid. The assistant is powered by GPT-4 and BERT models, employing advanced natural language processing to understand and respond accurately to civil legal queries. LawDroid provided the underlying technical infrastructure, while LANC's team focused the assistant on high-demand practice areas. LANC-LIA handles multilingual inquiries, automates routine communications, and surfaces self-service options for straightforward legal matters — integrating directly into legalaidnc.org to extend the organization's reach without expanding headcount. The tool explicitly does not provide legal advice, maintaining a clear boundary between information and representation.

LANC-LIA extended LANC's service reach by making legal information available on-demand, at scale, to communities that previously had no self-service path. Key outcomes include:

• Multilingual access: users can submit queries in multiple languages, addressing a longstanding barrier for non-English-speaking low-income populations
• Improved site usability: routine inquiries are now handled without staff intervention, freeing attorneys for complex casework
• Pre-launch validation: comprehensive user testing with clients, attorneys, and law students from Vanderbilt Law School's AI Lab (VAILL) confirmed both accuracy and accessibility before public release

While LANC has not published volume or deflection metrics, the deployment represents a structural shift in how the organization scales access to justice.

• AI chatbots can extend legal aid capacity without replacing attorneys — the key is scoping the tool to information delivery, not legal advice
• Partnering with a legal-tech vendor (LawDroid) rather than building from scratch allowed LANC to deploy faster with a robust, maintained infrastructure
• Multilingual NLP support is not optional when serving underserved communities — it should be a baseline requirement, not an enhancement
• Pre-launch testing with real end users — including clients, not just legal professionals — is essential for accuracy and trust in high-stakes domains
• Housing AI development inside a dedicated Innovation Lab provides the organizational structure and mandate to iterate responsibly