Certified Agentic AI Professional

Part A — knowledge and scenarios (70% of the score)

1. 1Advanced prompting and AI-powered analysis 15%

   Getting consistent, high-quality results at scale and turning information into decisions. Candidates design system instructions and structured outputs, and use AI to analyze documents and data for decision-ready reporting.

   • System instructions, personas and guardrails for consistent behaviour
   • Structured outputs: tables, JSON and templates that other tools can use
   • Chained and multi-step prompting for complex analysis
   • Analyzing long documents, multiple sources and datasets with AI
   • Turning findings into decision-ready reports, briefs and presentations
   • Building and maintaining a shared team prompt library
2. 2AI use-case design and business value 15%

   Finding, assessing and prioritizing the AI opportunities that matter. Candidates map processes, evaluate value, risk and feasibility, and define how success will be measured.

   • Mapping a business process and locating its pain points
   • Assessing use cases for value, risk, feasibility and data readiness
   • Prioritization: quick wins versus strategic projects
   • Defining success metrics, baselines and expected benefits
   • Writing a simple business case and pilot plan
   • Stakeholder engagement and change management for AI adoption
3. 3AI agents: concepts, design and building 25%

   The core of the Professional Exam. Candidates explain how AI agents work and design agents that are grounded in organizational knowledge, use tools safely and operate at the right level of autonomy.

   • Agent components: goals, instructions, knowledge, tools, memory and planning
   • Levels of autonomy and choosing the right level for a task
   • Grounding agents in policies and documents (retrieval-augmented generation)
   • Writing agent instructions, boundaries and escalation rules
   • Connecting tools and actions, and limiting what an agent may do
   • Testing and evaluating agents: accuracy, consistency and failure cases
4. 4Workflow automation and integration 20%

   Moving from single tasks to end-to-end automated processes. Candidates design multi-step workflows that combine AI with business rules, human approvals and the applications people already use.

   • Triggers, actions, conditions and data flow in automated workflows
   • Combining AI steps with business rules and human approvals
   • Integrating with email, chat, document libraries, spreadsheets and forms
   • Handling errors, exceptions and hand-offs to people
   • Monitoring, logging and improving workflows over time
   • Choosing between an AI agent, an automated workflow or both
5. 5Security, governance and responsible AI 25%

   Making sure what is built is safe, compliant and trustworthy. Candidates apply privacy, security and governance controls across the full life cycle of AI agents and automations.

   • Access control, least privilege and protection of sensitive data
   • AI-specific threats: prompt injection, data leakage and misuse of tools
   • Human oversight, approval points and named business owners
   • Audit trails, documentation and version control
   • Bias, fairness and transparency in automated decisions
   • AI policies, risk assessment and regulatory considerations