10 to the 23 AI logo
Work With 1023AI

Domain Atlas

Benefits navigation & public-facing chat

Conversational systems answering high-stakes questions for the public — where an authoritative wrong answer is indistinguishable, to its victim, from policy.

What AI is doing here

Benefits-navigation chatbots

Generative

Conversational guidance about eligibility and process, for applicants directly or for navigators assisting them.

Retrieval-grounded appeal drafting

Generative

RAG systems drafting determinations or appeal responses from policy corpora for adjudicator review.

What has gone wrong — and right

Documented deployments, presented as stylized model organizations with full citations.

Nava assistive benefits chatbot

United States (pilot deployments)

A retrieval-grounded navigation copilot built on verify-before-use principles — the cautious end of the design spectrum.

Who is in the system

  • Frontline workers. Caseworkers, screeners, eligibility staff — the operator network whose judgment the system augments or erodes.
  • Agency leadership. Owns procurement, policy, and the authority map; answers for the system publicly.
  • Served people & families. Those the decisions land on. Deliberately outside the PAN dynamics — their outcomes are measured, never simulated.
  • Vendors. Build and update the systems; hold the information asymmetry procurement must govern.
  • Advocates & community organizations. Surface harms institutions do not see; historically the earliest accurate signal.

Dominant pressures

  • Caseload surge. Demand outruns staffing; per-case attention shrinks and review becomes triage.
  • Vendor opacity. The deploying institution cannot inspect the model, data, or update pipeline it is accountable for.
  • Data & policy drift. The world, the intake process, and the rules change under a system trained on how things used to be.

Questions leaders should be asking

  1. 1. Is the system answering people directly, or assisting a professional who answers?
  2. 2. What is the retrieval corpus, who curates it, and how fast does it track policy change?
  3. 3. How are confident wrong answers detected — before a journalist detects them?
  4. 4. Does the interface honestly convey uncertainty, or does it perform authority?

For the actions behind these questions, see the Practice Library.