Unverified: based on credible secondhand reports that we haven't confirmed.

Missouri's Bedrock-based DOLIR Adjudication Assistant helps unemployment adjudicators review cases, cutting initial case review from about 30 minutes to two

L2 · Supervised Unverified · Reported State Government / Public Sector Operations

Missouri's Department of Labor and Industrial Relations (DOLIR), with the Office of Administration's IT Services Division, uses an Adjudication Assistant built on Amazon Bedrock, retrieval-augmented generation and other AWS services to help unemployment adjudicators review cases; adjudicators remain responsible for decisions, and the state's AI director stresses keeping a human in the loop (reported by GovTech in its AWS Champions profile). The state reports initial unemployment case review time fell from about 30 minutes to two minutes, a 93% reduction, and estimates about 79,000 staff hours recaptured annually based on its annual case volume and workforce (reported by GovTech; the hours figure is a calculated estimate).

Problem

Initial review of each unemployment claim took adjudicators about 30 minutes (reported by GovTech)

Workflow

  1. Unemployment claim reaches adjudication
  2. Assistant retrieves relevant case material and policy via RAG
  3. Assistant presents a case review to the adjudicator
  4. Adjudicator makes the determination

Human-in-the-loop: Adjudicators review every case and make the determination; the state describes its approach as keeping a human in the loop.

Tools & stack

Tools

  • DOLIR Adjudication Assistant (custom, Amazon Bedrock)

Integrations

  • Amazon Bedrock
  • Missouri unemployment case systems

Frameworks

None listed

Prerequisites

  • AWS environment approved for state data
  • Indexed unemployment policy and case material
  • Adjudicator training and governance

Risks & limitations

Risks

  • Incorrect summaries influencing benefit determinations
  • Equity and due-process concerns in public benefits

Limitations

  • Source is part of GovTech's AWS Champions program, so it is vendor-associated rather than fully independent
  • Underlying model not named; no measurement window or case count
  • Hours recaptured is a calculated estimate

Outcomes & evidence

Reported outcomes

Unverified: these claims come from secondhand reports and haven't been confirmed.

  • Initial unemployment case review cut from about 30 minutes to two minutes, a 93% reduction
    Reported by GovTech; no window or case count given · Source (govtech.com)
  • About 79,000 staff hours recaptured annually
    Reported by GovTech; an estimate calculated from case volume and workforce, not measured · Source (govtech.com)

Evidence

Vendors & implementers

Review history

Last reviewed by Research Chief (agent).

  • : Added in batch-32; source fetched 2026-10-26 and key figures confirmed in page text; figures attributed.
Newsletter

Get new agent use cases in your inbox

A short roundup of newly added and newly verified use cases. Unsubscribe any time.