{"resourceId":"dhs-fema-ai-use-case-inventory-2026","versions":[{"version":"external-828409e150075cfaa38ca4d17f6ac6c1e3f128aaa5f7e39a7d4fc627c2bafea1","resource":{"id":"dhs-fema-ai-use-case-inventory-2026","title":"FEMA DHS AI inventory: grants/watch copilots mostly pre-deploy; live tools skew admin GPT and workload models","organization":"Federal Emergency Management Agency / U.S. Department of Homeland Security","sector":"Federal emergency management AI inventory","geography":"United States (FEMA enterprise; implications for SLTT partners)","publishedAt":"2026-01-28","publicationDate":"2026-01-28","eventDate":null,"sourceName":"DHS Artificial Intelligence Use Case Inventory (FEMA page)","sourceLabel":"Official DHS simplified AI use-case inventory entries for FEMA","sourceUrl":"https://www.dhs.gov/ai/use-case-inventory/fema","evidenceClass":"government-evaluation","outcomeClass":"emerging","topics":["knowledge-work","governance-procurement","operating-model","developers-agents","accessibility-workforce"],"finding":"The FEMA page of the DHS AI use-case inventory (annual refresh noted Jan 28, 2026) lists numerous EM-related AI items with mixed maturity. Deployed examples include OCFO Response Augmentation Suite and Code Assist GPT (claimed 80–90% time savings on preliminary query drafting), Executive Summary GPT, and supervised IA/PA workload projection models. Pilot examples include Hazard Mitigation Assistance Chatbot and Technical Resource for Mitigation Programs (expected 40–70% document-review time cut; 15–20 analyst hours/week). Pre-deployment includes Grants Manager AI Chatbot, NCR Watch COP semantic search/summarization, and IA document translation (human translation >$1.7M in 2024; 6–10% non-English docs). Several geospatial damage assessment and PARC planning GenAI cases are retired/discontinued.","sledRelevance":"Primary federal inventory for what EM AI is actually fielded vs planned—useful when SLTT buyers assume FEMA already runs agentic disaster AI. Why now: pairs with RAND’s adoption findings to show inventory growth without implying operational autonomy.","evidence":"Opened dhs.gov FEMA inventory page via curl/WebFetch. Status labels, summaries, and quantitative claims transcribed from individual use-case entries and the Jan 28, 2026 update note.","architectureImplications":"Expect Azure OpenAI/RAG admin patterns and classical ML projections before any field geospatial AI; design human refinement hooks and audit logs.","governanceImplications":"Track Deployed/Pilot/Pre-deploy/Retired status; apply high-impact reviews where flagged; do not equate inventory listing with production readiness.","securityPrivacyImplications":"IA translation, fraud, and resume ATS touch sensitive personal data—require strict access and audit.","caveats":"Efficiency figures are often expected/claimed inside use-case narratives, not independent evaluations; inventory updates continuously; retired cases show churn.","streamIds":["emergency-services"],"roles":{"sales":"Use the DHS FEMA AI use-case inventory as a federal reality check: many disaster-facing entries remain pre-deployment/pilot, while deployed tools skew to admin GPT assistants, code assist, and IA/PA workload projections—not autonomous field AI. Discovery should map which FEMA-like workflows (grants, watch COP summarization, IA translation) a SLTT buyer actually needs and what human review remains. Bounded engagement: process mining + chatbot pilot for grants/mitigation docs with citation logging. Unsupported claims include operationalizing every inventory row or equating projected 40–70% review-time cuts with measured production results.","engineering":"Fit varies by status: deployed OCFO/Executive Summary GPTs and IA/PA projection models vs pre-deploy Grants Manager chatbot, NCR Watch COP summarizer, and IA document translation. Architecture typically Azure OpenAI/RAG with audit logging and human refinement. Prerequisites: authoritative document corpora, feedback dashboards, and high-impact reviews where flagged (e.g., AI Resume ATS). Constraints: several geospatial damage and PARC-class cases retired. Proof of value needs before/after handle-time and error metrics, not inventory presence alone.","delivery":"Delivery should phase internal admin and productivity copilots before situational-awareness summarizers, always with analyst-in-the-loop review. Start from Deployed or Pilot inventory patterns, train staff on citation checks, and capture handle-time and error baselines before expansion. Dependencies include cloud hosting/ATO, program SMEs for HMA/PA/IA, records management, and DHS AI governance updates. Ownership: program managers own mission outputs; AI governance owns risk tier; analysts own final decisions. Acceptance criteria: human approval recorded, complete audit logs, and measured—not only expected—efficiency metrics. Risks include inventory confusion between Pre-deploy and Deployed rows and hallucinated policy guidance."},"roleDetails":{"sales":{"customerProblem":"SLTT EM leaders need a clear picture of which FEMA AI tools are live vs aspirational before mirroring them locally.","stakeholders":["FEMA/SLTT program managers","Grants/HMA staff","Watch-center analysts","OCIO/AI governance"],"discoveryQuestions":["Which inventory IDs are Deployed vs Pilot vs Pre-deployment?","What human review remains on GPT outputs?","Are 40–70% review-time claims measured or expected?","Which use cases were retired and why?"],"valueHypothesis":"Citation-backed grants/mitigation copilots and workload projections can save analyst time if audit logs and human approval stay mandatory.","potentialEngagement":"Inventory-to-local workflow mapping and a single pilot with measured handle-time/error baselines.","unsupportedClaims":["All listed FEMA AI use cases are operational","Guaranteed 40–70% review-time reduction in every region","Watch COP AI removes need for analyst verification"]},"engineering":{"fit":"Mostly internal generative/predictive assistants for grants, policy Q&A, and workload forecasting—not autonomous disaster command.","architectureIntegration":"Azure-hosted LLMs/RAG over policy and financial corpora; dashboards for usage/accuracy; human refinement before external use.","prerequisites":["Curated policy corpora","Audit logging","Feedback loops","High-impact assessments where required"],"constraints":["Many pre-deploy entries","Retired geospatial damage / PARC cases","Survivor-facing translation still pre-deploy"],"security":"Keep survivor PII and fraud-detection features under strict access; log prompts/outputs for audit.","proofOfValue":["Handle-time deltas","Citation accuracy samples","Projection calibration","Escalation-rate changes"]},"delivery":{"work":"Pilot one Deployed/Pilot pattern locally, train staff on verification, and track metrics before expanding.","dependencies":["Cloud ATO/hosting","Program SMEs","Records management","DHS/FEMA policy updates"],"ownership":"Program manager owns outputs; AI governance owns risk tier; analysts own final decisions.","skillsAdoption":"Teach staff to treat GPT drafts as starting points with mandatory citation checks.","governanceCheckpoints":["Status verification against inventory","High-impact review","Pilot metrics gate","Retirement watch"],"acceptanceCriteria":["Human approval recorded","Audit log complete","Measured—not only expected—efficiency metric"],"risks":["Inventory confusion","Hallucinated policy","Overstated time-savings"]}},"retrievedAt":"2026-09-25T22:04:34-05:00","enrichedAt":"2026-09-25T22:04:34-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"IA translation and agent-support search aim to cut survivor wait and agent handle time, but remain pre-deploy/retired in places—workforce plans must not assume live AI.","procurementImplications":"Mirror FEMA’s phased approach: start with internal copilots with citations; budget independent measurement of time/error claims.","operatingModelImplications":"Keep analysts/leadership as final decision-makers on GPT drafts and treat projections as planning inputs, not staffing automations.","sourceVerification":{"openedUrl":"https://www.dhs.gov/ai/use-case-inventory/fema","referenceExcerpt":"reduce document review time by 40-70%, saving 15-20 analyst hours per week","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}