Caseworker Empowerment Toolkit
A suite of open source AI-powered tools that help caseworkers at every step of their journey connecting families with the social safety net
Tools for every part of the workflow
The caseworker’s job is to help people. In practice, a lot of their day goes to other things: searching policy manuals while a client waits, re-entering the same data across multiple portals, chasing down documents that don’t meet program requirements. These tools are built to give that time back.
The Caseworker Empowerment Toolkit is a suite of open source AI tools built by Nava Labs, running on Nava Strata, which is open source technology infrastructure proven across government benefits programs.
Piloted with
Evaluation with
Workflow tools — five steps
Connecting clients to services
Clients often need services beyond what a single program offers, but finding current, relevant resources across a fragmented landscape takes time caseworkers don’t have.
The Referral Generator searches state benefits, local resources, and internal programs that fit each client’s situation, then turns those options into step-by-step action plans.
Answering program questions
Caseworkers field eligibility questions all day, often while a client is waiting. The answers live in policy manuals that are dense, inconsistent, and hard to search under pressure.
The Assistive Chatbot gives caseworkers real-time answers in plain language, with direct-quote citations so they can verify before they act.
Verifying documents
Clients applying for benefit programs face compounding documentation requirements. Errors caught late, or not caught at all, cause delays and denials.
DocumentAI identifies document types, extracts required data, and flags issues in real time, giving caseworkers a clearer picture of what’s complete and what needs follow-up.
Completing applications
Applying for benefits often means navigating multiple portals and re-entering the same client information across each one.
The Form-Filling Assistant uses agentic AI to fill out application forms automatically using existing client data. Caseworkers review, fill in any missing information, and submit.
Documenting client conversations
Caseworkers need to stay present with clients while also capturing the details that belong in a case record.
The Voice to Case Note is a prototype being tested with caseworkers that shows how a consented conversation could surface details for a caseworker to check before preparing a draft note.
Our approach
Evidence to inform scale
Every tool we ship includes structured evaluation: usage data, caseworker surveys, and in-depth research into whether the tools improve accuracy and confidence in the field.
Our pilot of the Assistive Chatbot included a randomized evaluation involving 125 caseworkers from six organizations in Los Angeles. Caseworkers used the chatbot for decision support while navigating client scenarios related to SNAP benefits. The results showed that the chatbot improved caseworker accuracy by approximately 40%, with larger gains observed among newer caseworkers and in more complex cases.
Building in the open
Caseworkers need to trust the tools that inform their work. Agencies need to stay in control of systems that shape how families access benefits. Open source addresses both: the models, prompts, and logic are all visible, there’s no vendor to depend on, and the work stays in the hands of the community that needs it.
Philanthropically funded
Nava Labs uses philanthropic funding to take on work that agencies need but can’t always fund through procurement. That means we can move faster, share everything we learn, and draw on Nava PBC’s decade of experience delivering critical government services.
Connect with us
We work with government agencies and community organizations as pilot sites, with funders to support development and scale, and with practitioners and researchers who want to share learnings or explore shared challenges.