Insight

Caseworkers and call-center specialists, whom we call “navigators,” serve people facing big hurdles like homelessness, health issues, housing insecurity, navigating foster care, and supporting children. Navigators need to help a lot of people quickly, and their clients need clear and accurate information on benefits, housing, food, and more.

Given this need, our project exploring AI tools for public benefits has introduced an assistive AI chatbot to the Benefit Navigator — a tool that helps navigators and their clients wade through benefits and tax credits — to help surface answers on resources more quickly.

Through talking to navigators and looking at data, we now have some specific learnings around their chatbot use:

Why navigators use the chatbot:

• It’s quick. Answers appear in seconds. This is handy when a client is sitting in the office or leaving a clinic visit.

• It’s credible. Our team curated the sources the chatbot uses. Every answer shows its sources, so navigators trust it more than a broad Google search.

• It fills gaps. The chatbot surfaces details that navigators might miss. For example, one navigator used the chatbot to learn about tax credits for foster youth before hosting advising sessions.

• It boosts confidence. They use the chatbot to double-check steps and required documents before recommending actions to clients.

• It’s actionable. Chatbot responses include ready-to-share links, phone numbers, and step-by-step instructions across many benefit programs.

What slows or stops use:

• Forgetting due to infrequent need. Some navigators run benefit checks only once or twice a month and forget the tool exists.

• They feel they already know. Giving a quick phone number or referral can be faster than opening the chatbot.

• Local info is missing. For this pilot, the chatbot couldn’t yet list nearby shelters, hotlines, or agencies, and focused on state and federal programs instead.

• Fear of outdated rules. They want a visible “last updated” date on every answer to avoid passing incorrect info to clients.

• Personal habit of using official sites. Some navigators prefer to check each program’s website directly.

Keep following our project as we release more on what we’re learning, both quantitative and qualitative. We’re excited to share our full pilot evaluation and analysis results soon!

This post is an excerpt from the first of a new Nava Labs Newsletter. If you'd like to sign up for our Nava Labs newsletter, you can do so here.