# AI Agents for Government Workflows Course (Self-Paced) Canonical URL: ## Overview AI agents represent a new class of artificial intelligence tools capable of planning, acting, and adapting across multiple steps to complete tasks — and they're now shipping in commercial products that government agencies are already evaluating and deploying. This course provides a practical, no-code introduction to agentic AI for the government workforce. Participants learn what AI agents are, how they differ from traditional automation and chatbots, how agentic tools work at a technical level, and how to evaluate whether a specific tool is appropriate for a given government workflow. The course covers the current agentic AI landscape, permission models and security considerations — including prompt injection — and a structured evaluation framework participants can apply immediately. Governance, lifecycle management, and workforce readiness round out the program, aligned with the White House's 2026 National Policy Framework for AI. This course focuses on evaluation, governance, and informed decision-making rather than technical development or coding. ## What you'll learn - Explain what AI agents are and how they differ from automation, chatbots, and decision-support tools - Describe how agentic tools work at a high level — the AI model, planning loop, tool access, and containment boundary - Evaluate whether a specific agentic tool is appropriate for a given government workflow using a structured framework - Identify risks, limitations, and operational warning signs specific to agentic AI systems - Apply human oversight models (HITL, HOTL, HIC) and governance frameworks to agentic tool deployments - Assess organizational readiness for agentic tool deployment using a practical checklist - Identify security considerations specific to agentic tools, including prompt injection and permission models ## Curriculum #### Module 1: Understanding AI Agents - What AI agents are and how they differ from automation, chatbots, and decision-support tools - Core characteristics of AI agents: goal-driven behavior, multi-step execution, autonomy, and context-awareness - The agentic AI landscape today: desktop agent tools, enterprise agent platforms, developer tools, and AI assistants with tool access - How agentic tools work under the hood: the AI model, planning loop, tool access, and containment boundary - Live demonstration of an agentic tool performing a multi-step government workflow task - Human oversight roles: human-in-the-loop, human-on-the-loop, and human-in-command - Common misconceptions and key risks, including prompt injection - Current federal AI policy direction and its implications for government agencies #### Module 2: Evaluating Agentic Tools for Government Work - The evaluation mindset: understanding the workflow before evaluating the tool - Identifying agent roles in a workflow: intake, analysis, recommendation, and escalation - The Agent Evaluation Blueprint: goal, trigger, inputs, actions, boundaries, and oversight - Permission models and data access: folder-level vs. organization-wide, network access, and least privilege - Prompt injection: what it is, how it works, and why agentic tools are uniquely vulnerable - Questions to ask before saying yes: a practical pre-approval checklist - Hands-on activity: evaluate a realistic agentic tool proposal for a government workflow #### Module 3: Where Agentic Tools Fit — and Where They Don't - Appropriate use cases: case triage and routing, document analysis and extraction, internal coordination, and monitoring and alerts - High-risk or inappropriate uses of agentic tools in government - Operational risks: over-automation, over-reliance, and rubber-stamping - Drift: data drift, concept drift, and objective drift — and how to detect them - Early operational warning signs that an agentic tool may be failing - Hands-on activity: red team a deployed agentic tool scenario to identify risks and recommend action #### Module 4: Governing Agentic Tools in Your Organization - Why governance is essential — and why no centralized federal AI regulator is coming - Governance vs. technical controls: policies, oversight bodies, and accountability structures - Legal, ethical, and procurement considerations: FedRAMP, ATO, vendor data handling, and records retention - Security for agentic tools: access controls, prompt injection defenses, anomaly monitoring, and incident response - Lifecycle management: design, pilot, deploy, monitor, update or retire — including regulatory sandbox alignment - Performance monitoring and metrics: accuracy, override rates, equity indicators, and user satisfaction - Workforce readiness and change management: role clarity, training on tool limitations, and avoiding fear and over-trust - Deployment readiness checklist: a practical gate review before any agentic tool goes live - Hands-on activity: conduct a readiness gate review for a proposed agentic tool deployment ## Instructors ### Steve Pesklo — Instructor Steve is an energetic trainer who focuses on applying technical concepts to everyday work practices. He is the founder and president of SoftLake Solutions, a company that specializes in providing data and AI applications to identify fraud for Internal Audit, Criminal Investigations, Forensic Accounting, Privacy, and Compliance. Steve brings a large amount of experience across multiple industries and government agencies. He is an expert in implementing large data analysis projects across the world, including Inland Revenue in the UK and Argentina, New Zealand, Africa and across Europe. Previously, he was the manager of Data Architecture and Data Services for a large mortgage company. He is a frequent speaker on data analytics and project management topics and speaks fluent German. He has been teaching at the Graduate School for over 10 years. Steve has an M.B.A. from the University of St. Thomas and a B.S. in Computer Science from California Lutheran University and the Universität Salzburg in Austria. He is certified as a Certified Fraud Examiner (CFE), Project Management Professional (PMP), and a Certified ScrumMaster (CSM). ## Pricing **Tuition:** $675