AI in government is not one thing. It is a federal analyst asking an agency-approved assistant to cluster 400 public comments, a SNAP eligibility worker clearing an exception queue of pay stubs the extraction tool could not read, a FOIA analyst accepting or rejecting a suggested (b)(6) redaction, a patrol officer checking every sentence of an AI-drafted vandalism narrative against the body-camera audio before signing it, a 911 telecommunicator taking over from a non-emergency virtual agent, a detective holding a facial-recognition candidate list that is a lead and not an identification, a court research attorney checking every citation in an AI bench memo, a staff officer drafting a CUI information paper on GenAI.mil, an aircraft maintenance planner deciding whether a predictive alert earns an inspection, and a grant writer tracing every number in a drafted need statement back to program data. This free hour walks through those jobs one at a time, as the person who does them sees them.
For each job you see how the work is done today, what the AI actually does and with which data, the tool categories and some named products as of October 2026 (GSA USAi, OPEXUS FOIAXpress AI Assistant, Veritone Redact, Axon Draft One, Axon 911 with Prepared and Carbyne, Learned Hand, GenAI.mil with Google Gemini for Government, the Air Force's PANDA predictive-maintenance system built on C3 AI, Babel Street, Dataminr, Grantable, Instrumentl and Candid), where the human decision must stay human, and the specific control that applies: OMB M-25-21's high-impact categories and minimum practices, NARA's August 2026 guidance on AI prompts and outputs as federal records, the FNS advanced-automation memo for SNAP, FOIA exemptions (b)(6) and (b)(7)(C), Utah Code 53-25-602 and California Penal Code 13663 on AI-drafted police reports, the CJIS Security Policy, Maryland's facial-recognition statute (with the Detroit-only Williams settlement as a reference point), California Rules of Court rule 10.430, the DoD AI Ethical Principles and the CUI program, ICD 203 analytic standards, federal grant-application certifications and VAWA confidentiality. EU AI Act Article 5 prohibitions and Annex III appear only where they reach US organizations, labeled as such. Products are named to show what exists, not to endorse them, and vendor performance claims are labeled as claims or left out. Defense content is limited to administration, maintenance, intelligence-analysis support and policy; weapons targeting and engagement are out of scope. Every lesson ends with a short exercise on a fixed, fully synthetic packet with an answer key, such as finding the planted defects in a drafted police narrative or collapsing syndicated duplicates in a fictional port-strike news feed. This course is educational and is not legal advice. A 10-question exam earns a free Wirewalk Institute certificate of completion and assessment, and the last lesson maps each role to the paid course that goes deeper. The certificate attests to completion of this course and its automated assessment only. It is not a government, CJIS, DoD or agency training credential and does not satisfy any legally required training.