The executive briefing
From inventory to insight
A concise reading of federal AI portfolios, deployment and the evidence agencies report.
Read the briefingFederal AI landscape
Federal AI at a glance
14 agencies · 445 reported use cases · OMB 2025 Inventory. Portfolio size reflects reporting, not performance.
Scale & composition
Select an agency to exploreAgencies behind federal AI
In development/acquisition140Pilot23Deployed227Retired55Largest portfolio: 215 · 14 agencies
Conceptual agency landscape · positions do not represent deployment locations. Proximity does not imply a relationship.
AI technology mix
Reported classification · click to cross-filter18% generative or agentic. Each record carries one classification; not reported is shown hatched.
Mission areas
Topic area as reportedLifecycle composition
445 recordsExplore lifecycle stages and impact
Within each stageHigh-impact · Gen./agentic
Pre-deployment100%39%
Pilot100%48%
Deployed100%7%
Retired100%0%
Impact and technology shares use each stage's records as the denominator.
From development and acquisition to retirement. Select a stage to explore its reported use cases.
Reading the inventory
Evidence, with its limits visible
VA, DOJ, DHS report 384 of 445 use cases (86%).
Understand the methodology →Evolution of reported AI use cases
By agency-reported start year · click a year to filterGenerative or agenticOther or unclassified148 of 445 records report a parseable start date
Start dates are agency-reported operational or pilot dates, not inventory submission dates. 2026 values are planned or early-year. Undated records are excluded from this chart only.
Sourcing
Reported for 56% of recordsEach square is 1% of filtered records. Where reported, 31% involve in-house development.
Minimum practices · 445 high-impactin place
Pre-deployment testing10%
AI impact assessment8%
Independent review8%
Ongoing monitoring9%
Operator training9%
Fail-safe8%
Appeal process4%
User and public consultation5%
Reported in placeReported in progressNot applicableOther or ambiguous answerNo answer reported
High-impact is the agency's own classification, not an independent risk rating. No answer reported is missing evidence, not evidence that a practice is absent. 29% of high-impact records answer any practice question.
Analytical observations
Computed from the current filter; reproducible- FindingVA, DOJ, DHS report 384 of 445 use cases (86%).
- Finding7% of reported deployed use cases (17 of 227) are classified as generative or agentic AI. View records
- FindingOf 227 deployed high-impact use cases, 36 report a completed AI impact assessment and 101 leave every minimum-practice field blank. View records
- InterpretationBlank practice fields are missing evidence, not evidence that a safeguard is absent. They mark where follow-up questions to the agency are most useful.
- FindingSourcing is reported for 56% of use cases. Where reported, 31% involve in-house development.
AI classification reported88%
Stage not reported0
Sourcing reported56%