AIAI Agent Dev Labaiagentdev-lab.com
Sign in to save progressSign in
Menu

STEP 1 · Study

AB-100 StudyAgentic AI Business Solutions Architect (Expert)

Read the key points for each exam domain, then use “Practice this domain” to answer questions from that domain only. Domains with a larger weight appear more often on the exam.

01

Plan AI-powered business solutions

25–30%

Learning goals

  • Evaluate where to use agents among task automation, data analysis and decision support
  • Inspect the quality of grounding data (accuracy, relevance, freshness, cleanliness, availability)
  • Outline a company-wide strategy based on the Cloud Adoption Framework AI adoption process and an AI CoE
  • Perform ROI analysis including TCO and decide between build, buy and extend

Key points

  • First classify “what the agent will do in that work”: task automation, data analysis or decision support. Decision support assumes a design that leaves the final decision to people.
  • Inspect grounding data for accuracy, relevance, freshness, cleanliness and availability. If the data is not in order, agent quality will not improve.
  • Follow the Cloud Adoption Framework AI adoption process to progress through strategy, planning, readiness and governance in stages.
  • Decide by requirements whether extending Microsoft 365 Copilot is enough, building a custom agent in Copilot Studio is needed, or creating custom models in Foundry is needed.
  • Establish a prompt library and prompt writing guidelines to align quality across departments. An AI CoE (Center of Excellence) is the core of promotion and governance.
  • Evaluate ROI including TCO (licenses, development, operations, human review effort). Comparing build, buy and extend, and cost optimization with a model router, are also options.

Workplace uses

  • SalesSeparate what Dynamics 365 Sales Copilot features cover from where a custom quoting agent is needed, and compare build costs.
  • General AffairsInventory freshness and owners of internal policy data before using it for grounding. Clean up outdated policies first.
  • AccountingEstimate invoice processing ROI by TCO including review effort and operating costs, not just time saved, and use it in the approval request.

Exam objectives (from the exam guide)

  • Analyze requirements for AI-powered business solutions
  • Design overall AI strategy for business solutions
  • Evaluate the costs and benefits of an AI-powered business solution

Learning resources

Learn in detail (3 topics with mini quizzes) →Practice this domain (89 questions) →Next domain ↓
02

Design AI-powered business solutions

25–30%

Learning goals

  • Design using task agents, autonomous agents and prompt-and-response agents appropriately
  • Design Copilot Studio topics (including fallback), agent flows and prompt actions
  • Design extensions with MCP, computer use and Microsoft 365 agents
  • Orchestrate configuration of prebuilt AI features and agents in Dynamics 365 and Power Platform

Key points

  • Distinguish agent types: task agents that perform requested tasks, autonomous agents that act on their own from triggers, and prompt-and-response agents that answer questions.
  • The more autonomous the agent, the heavier the design of permissions, approvals and monitoring. Choose the type by scope of impact.
  • In Copilot Studio, choose by requirements among standard natural language processing, conversational language understanding (CLU) and generative AI orchestration. Prepare a fallback topic for unexpected utterances.
  • Combine extensions such as MCP (tool connections), computer use (operating UIs without APIs) and agents on Microsoft 365 (Teams, SharePoint).
  • Consider prebuilt AI features first (Dynamics 365 Copilot, Microsoft 365 Copilot for Sales / Service, Power Platform AI features and so on) and build only what is missing.
  • Apply the Power Platform Well-Architected Framework to intelligent application workloads to ensure reliability, security and operational excellence.

Workplace uses

  • SalesUse prebuilt Copilot features to summarize opportunity notes, and design credit checks and discount approvals as a custom task agent plus approval flow.
  • General AffairsPlace a prompt-and-response agent on the internal portal in Teams and SharePoint; unexpected questions go to the general affairs desk via fallback.
  • AccountingStart month-end close reminders with an autonomous agent, but send them only after staff approve the recipients and wording.

Exam objectives (from the exam guide)

  • Design AI and agents for business solutions
  • Design extensibility of AI solutions
  • Orchestrate configuration for prebuilt agents and apps

Learning resources

Learn in detail (3 topics with mini quizzes) →Practice this domain (89 questions) →Next domain ↓
03

Deploy AI-powered business solutions

40–45%

Learning goals

  • Design agent monitoring processes, metrics and telemetry interpretation and run an improvement cycle
  • Build testing strategies spanning agents, custom models and multiple Dynamics 365 apps
  • Design ALM for data, Copilot Studio, Foundry Agent Service, custom models and Dynamics 365
  • Design security, governance, mitigations for prompt manipulation, data residency, access control and audit trails

Key points

  • Monitoring is a cycle of “metrics → telemetry interpretation → tuning.” User feedback and the backlog are also inputs for improvement.
  • Decide testing processes and metrics first. Prepare validation criteria for custom models and end-to-end scenarios for work spanning multiple apps.
  • Design ALM per target: data used by AI, Copilot Studio agents, connectors and actions, Foundry Agent Service, custom models and Dynamics 365 AI features.
  • Security and governance: design agent permissions, model protection and access control to grounding and tuning data.
  • Analyze prompt manipulation (such as prompt injection) as a vulnerability and mitigate in layers with input handling, tool permissions and approvals.
  • Verify compliance of data residency and movement, keep audit trails of model and data changes, and review adherence to responsible AI principles.

Workplace uses

  • SalesTurn quoting agent usage and lost-deal feedback into a backlog and tune monthly.
  • General AffairsManage the internal portal agent with ALM using Copilot Studio solutions, and deploy policy revisions after end-to-end tests.
  • AccountingStore change history of data and models used for journal suggestions as an audit trail, ready to present in financial audits.

Exam objectives (from the exam guide)

  • Analyze, monitor, and tune AI-powered business solutions
  • Manage the testing of AI-powered business solutions
  • Design the ALM process for AI-powered business solutions
  • Design responsible AI, security, governance, risk management, and compliance

Learning resources

Learn in detail (4 topics with mini quizzes) →Practice this domain (137 questions) →Try the mock exam →

When you finish reading

Practice all domains (315 questions)Mock exam (50 questions · 75 min)AB-100 home
Notes (unofficial, original questions, no pass guarantee)