AB-100 practice test

Assessing where agents help in automation, analytics and decisions, reviewing grounding data for accuracy, relevance, timeliness and availability, the Cloud Adoption Framework's AI adoption path, multi-agent designs across Microsoft 365 Copilot, Copilot Studio and Foundry, prebuilt agents and their use cases, deciding between extending Microsoft 365 Copilot and building custom agents, when a custom or small language model is warranted, prompt libraries and prompt engineering guidelines, the AI Center of Excellence, solutions spanning several Dynamics 365 apps, and the economics: return on investment criteria, total cost of ownership, build against buy against extend, and model routing.

Designing Copilot in the Dynamics 365 customer experience and service apps including business terms, customisations and connectors, agents for contact centre channels, task, autonomous and prompt-and-response agents, Foundry Tools, generative pages and agent feeds, Copilot Studio topics and fallback, grounding and data processing, AI components in canvas apps, the Power Platform Well-Architected Framework for intelligent workloads, choosing between classic understanding and generative orchestration, agent flows and prompt actions, extensibility through custom Foundry models, Microsoft 365 Copilot agents, Model Context Protocol and Computer Use, agent reasoning and voice, and orchestrating the AI features already present in the finance, supply chain, customer experience and service apps.

Monitoring agents and interpreting telemetry for tuning, analysing backlog and user feedback, testing: process, metrics, validation criteria for custom models, prompt practice and end-to-end scenarios across several apps, lifecycle management for agent data, for Copilot Studio agents, connectors and actions, for the Foundry Agents service, for custom models and for the AI inside the Dynamics 365 apps, and the whole of responsible AI, security, governance, risk and compliance: agent and model security, vulnerability analysis including prompt manipulation, responsible AI review, data residency and movement, access control on grounding data and tuning, and audit trails for changes to models and data.