The AI-powered assistive chatbot can answer a range of questions about public benefits programs.
We’re excited to announce that Nava Labs has partnered with Imagine LA — a 501(c)(3) nonprofit dedicated to ending family poverty and homelessness in Los Angeles — to pilot an assistive chatbot powered by a large language model (LLM). The chatbot is designed to connect case managers to trusted information about benefits and tax credits that could support the people they serve.
This work builds on over a year of research and feedback from diverse stakeholders, a range of experiments, and Imagine LA’s 18-plus years of providing direct social services. The pilot is part of a broader Nava Labs initiative to explore how generative AI can enhance access to public benefits and reduce administrative burdens.
What we’re doing
Imagine LA manages the Benefit Navigator, a powerful tool that assists case managers (“benefit navigators”) in guiding clients through public benefits and tax credits. It includes a benefits screener, application tips and links, tools to help people avoid benefit cliffs, and program-specific eligibility information across many local, state, and federal programs available to Los Angeles County residents.
A screenshot of Imagine LA’s Benefit Navigator website.
We’ve helped Imagine LA add a new layer to their tool: an assistive chatbot powered by an LLM, designed to support these navigators in real time. It can answer questions about eligibility, application steps, documentation, and program interactions — drawing from curated sources and responding with direct-quote source citations. It’s also multilingual and designed to complement the Benefit Navigator’s existing workflow.
Examples of the capabilities of the chatbot: the range of topics it can assist with, what a typical response looks like, and what source citations look like.
Why it matters
Public benefits programs are complex. The Benefit Information Hub in the Benefit Navigator contains highly vetted, trusted information about nearly 40 benefit programs and tax credits, each with its own rules and application processes. In addition to their ongoing work supporting families, navigators need to understand these programs, interpret conflicting rules, manage edge cases, and find the right information under pressure. It can take years to learn. Our chatbot aims to reduce that cognitive load, speed up responses, and help navigators feel confident when assisting clients.
Imagine LA focuses on providing clear, understandable benefit information through the Benefit Information Hub. Early uses of the chatbot have demonstrated that it builds on the foundation of the Hub and provides navigators with quick, accurate program guidance, facilitates better planning discussions with families, and reduces the time required to interpret complex rules.
Why a pilot
A pilot allows us to evaluate the tool’s usefulness and accuracy before scaling and launching to more users.
In partnership with researchers at Cornell University and Georgetown University’s Better Government Lab, we’ll collect detailed feedback to refine the chatbot. We’re testing:
What kinds of questions navigators ask
How useful and accurate the answers are
When the chatbot is helpful — and when it is not
How to ensure navigators feel confident using the tool
Whether the chatbot helps reduce or add to workflow administrative burden
What’s next
We’re running this pilot through the spring, with plans for two public Demo Days to share out early learnings and final results.
Want to see a demo of our chatbot? Register for our Demo Day on April 17! We’ll showcase what we’ve built, what we’ve learned so far, and what’s coming next.
