Resource Referral Experiment
To provide streamlined, holistic support to caregivers, navigators need to be able to make referrals to community-based resources and government assistance programs that meet their clients' unique needs. While conducting a program screening or application, navigators often identify needs that can only be addressed with additional referrals. For instance, if a client is ineligible for SNAP, the navigator will refer them to a food pantry. In other cases, navigators identify needs that are outside of their scope, such as a navigator who specializes in SNAP and needs to refer a client to another resource for help applying for unemployment insurance.
When making referrals, navigators must identify resources that match the need identified, that their client is likely eligible for, and that are accessible (e.g. in-person services are located locally, any contact information provided is current, etc). BDT navigators have access to vetted resource lists that are regularly updated. Despite access to these lists, navigators lose time manually searching vetted lists and run the risk of missing valuable resources through this manual process.
As a navigator helping clients over the phone, I need to provide referrals to local resources and government programs outside of the ones I directly assist with, so that my client can get support with immediate needs.
Question to test: Can AI retrieve, combine, and filter data sources to provide benefit referrals relevant to the applicant?
Hypothesis: A successful AI intervention will be able to identify resources that match the needs identified by the navigator quickly.
Inputs:
- 5 caller scenarios - each representing different household demographics and resource needs. For each scenario, we defined the referrals expected to address client needs
- Resource databases: National 211 API and a spreadsheet of mocked data for NYC.
