AI Risk Management, Including Agentic AI
Identify, assess, and manage AI risks, including the added risks of agents that use tools and take actions.
Course overview
This one-day instructor-led course introduces practical AI risk management for business and technical teams. Explore risks across predictive AI, generative AI, and agentic workflows, from inaccurate outputs and privacy exposure to unauthorized actions and unsafe tool use. In a fictional workplace case, participants assess risks, select proportionate controls, and plan oversight and incident response. No coding or prior AI risk management experience is required.
Course coverage
Move from identifying AI risks to planning controls and oversight for a fictional agentic workflow.
- 01 AI risk basics and the NIST AI Risk Management Framework
- 02 Predictive and generative AI risks and risk assessment
- 03 Agentic AI, tools, delegated actions, and trust boundaries
- 04 Prompt injection, data exposure, and unauthorized actions
- 05 Proportionate controls, approvals, and accountability
- 06 Testing, monitoring, incidents, and escalation
Participants will learn to
- Describe common AI risks and the people they may affect.
- Distinguish AI-generated advice from agentic tool use and actions.
- Map an agent workflow, its data, permissions, and decision boundaries.
- Prioritize risks using likelihood and impact, noting assumptions.
- Recognize prompt injection and unsafe instructions in untrusted content.
- Recommend least-privilege access, approval, and other controls.
- Plan testing, monitoring, incident response, and human oversight.
- Document owners, remaining risks, and when to seek specialist review.
One-day course outline
Morning - AI risk foundations and assessment: Introduce predictive AI, generative AI, and agentic systems, their lifecycle, and the difference between producing an answer and taking an action through a tool. Identify stakeholders and discuss inaccurate or biased outcomes, privacy and security exposure, third-party dependencies, and operational disruption. Introduce the voluntary NIST AI Risk Management Framework (AI RMF) and its four functions: Govern, Map, Measure, and Manage. Use them to frame accountability, understand context, assess risks, and plan responses; the framework is a reference, not a compliance checklist. Map a fictional service-desk agent that reads incoming requests, retrieves guidance, drafts replies, and proposes ticket updates. Identify data sources, tools, users, permissions, and affected people. Workshop: Build a risk register for the case, estimate likelihood and impact, record assumptions and potential harms, and prioritize risks for treatment.
Afternoon - Agentic controls and ongoing oversight: Examine how untrusted request text or retrieved content can contain instructions intended to redirect an agent, and how excessive permissions, connected tools, repeated actions, or external destinations can magnify harm. Discuss least-privilege access, separation of untrusted content from instructions, controlled data access and outbound destinations, action limits, human approval before consequential actions, and safeguards enforced outside the model. Plan representative tests, logging appropriate to data sensitivity, monitoring, review triggers, and how to pause an agent. Workshop: Assign controls and owners to the morning risks, identify residual risks and approval points, and walk through a fictional attempted prompt injection or unauthorized ticket update: pause or contain, notify the owner, preserve appropriate evidence, and escalate. Written prompts or approvals alone do not eliminate risk; controls must be tested and reviewed.
Format and audience
For: Business leaders, managers, project and product teams, analysts, and technical, risk, privacy, security, or compliance professionals responsible for AI-enabled work.
Prerequisites: Basic computer and workplace experience. No coding, prior AI knowledge, or risk management experience is required.
Workshop setup: A computer with a modern browser and an application for viewing and editing documents or worksheets. The instructor supplies fictional scenarios and templates; no AI tool account is required.
How it works: One day of instructor-led explanations, discussions, case studies, and guided workshops. Contact us to discuss dates and virtual or onsite delivery.
Explore the course
Request a quote for AI Risk Management, Including Agentic AI and discuss your team's learning goals.