AB-100 exam prep: Agentic AI Business Solutions Architect

Sign into Microsoft Copilot Studio to create agents that handle input and output with natural language and generative AI. Learn pricing, trial access, and work or school email requirements.

Explore why you need a work or school email to access Power Platform and how to obtain a free work email through Microsoft 365 trials and domain setup.

Explore what agents are and how Copilot Studio provides prebuilt agents for task automation, data analytics, and decision making, including catalog enrichment, document processing, and store operations.

Create a new environment and a new solution to organize agents and their objects. Select a region, set a publisher name, and apply a unique prefix for object identifiers.

Provision your first agent in Copilot Studio, define clear measurable goals, and set orchestrations while ensuring data security, privacy, and alignment with company goals.

Create a leave allowance topic in Copilot Studio with trigger phrases for vacation and sick leave. Test using new sessions and send messages and cards to the user.

Build an HR bot that asks which employee and returns their leave allowance, using a name variable, multiple choice, entities, and conditions with Power Fx.

Add and design adaptive cards to collect multiple data points, using text blocks, input numbers, and variables; manage topics, end conversations, and deploy tools and agents for advanced workflows.

Learn Copilot Studio design topics by creating a second topic with a global employee name variable, testing topic transitions, and reviewing custom topics, system topics, and fallback and escalate behaviors.

Create a new solution and set it as preferred. Build a hotel classic agent with orchestration, define pricing topics, and test double and family room costs.

Design and test generative orchestrated agents that draw from uploaded spreadsheets or web sources, with controlled knowledge, avoiding hard-coded data.

Explore building a generative ai agent using a spreadsheet as the knowledge source, avoid assuming the logged-in employee, and craft clear, active instructions with markup.

Discusses Microsoft responsible AI principles: fairness, reliability and safety, privacy and security, inclusiveness, transparency, accountability—and how governance and policies ensure compliant AI solutions.

Explore how generative AI and diverse knowledge sources power Copilot Studio agents, comparing classic and generative orchestration, and configuring web search, documents, and databases.

Publish and configure Copilot Studio agents across channels with authentication settings and suggested prompts. Manage voice and text configurations, telephony options, and channel-specific publishing.

Create a generative AI orchestration agent that returns room rates for a room type from the attached room rates spreadsheet, grounding it with this data source.

Follow the ALM data governance cycle to ensure fresh, clean data from appropriate sources for AI models and grounding, including pre-processing, de-duplication, anonymization, OCR indexing, and standardized terminology.

Review and organize structured and unstructured data for grounding AI solutions, ensure quality, accuracy, completeness, timeliness, and availability, and manage data sources and permissions for reliable AI.

Identify data residency rules and cross-region movement for CoPilot Studio by reviewing environment regions, Bing search data boundaries, and configurable options across EU, UK, Australia, and US.

Learn to switch the agent model in Copilot Studio, use Microsoft Foundry custom models, and add tools and prompts while understanding model categories like deep, auto, and general.

design autonomous agents that trigger on external events like emails, look up leave allowances from spreadsheets, compose and send replies, and follow strict scope, data quality, and auditing practices.

Add an email trigger to make the agent autonomous, configure Outlook inbox filtering, test the trigger, review permissions and Power Automate flow, and publish.

Add a tool to your autonomous agent to send emails using Office 365 Outlook, configure recipients, subject and body, and test triggers.

Create an autonomous agent that triggers on emails containing the word room and uses a room rates spreadsheet to determine room type and price, then publish and test.

Create a custom prompt action in Copilot Studio to retrieve an employee's job role, test and refine prompts, and save a dedicated employee role tool for integration.

Integrate the custom prompt into the autonomous agent, publish changes, and test by looking up employee job roles to compose and send email responses in Copilot Studio.

Expand an agent by building a custom prompt tool to count hotel rooms by type, test with single and family rooms, and automate email replies with room counts and rate.

Create a second agent to track leave by leveraging a leave requests spreadsheet stored on OneDrive for Business, using it as a knowledge source and testing multi-agent design patterns.

Compare inline child agents and connected agents, weighing simplicity and scalability; learn when to use each, how they share parent topics, manage inputs and triggers, and address publishing and security considerations.

Explore multi-agent designs, from sequential orchestration to concurrent, group chat orchestration. Learn magentic orchestration, handoff orchestration, and agent handoff pattern, plus maker checker loops for specialised, time-sensitive tasks.

Create a third agent called the leave balance aggregator that uses agents B and D to compute remaining vacation and sick leave by subtracting taken from the allowance.

Create a new agent that writes a leave request to a OneDrive spreadsheet, including name, date from, date to, vacation days, and leave type; enforce field validation and confirm writing.

Explore designing agent extensibility in Copilot Studio with agent flows, using tools, topics, inputs, and conditions to create predictable leave-request automation.

Explore designing agent extensibility in Copilot Studio using the MCP model context protocol to surface Microsoft 365 data and tools with enterprise-grade security and authentication.

Create and publish an agent that calls a previous agent to fetch the room rate for a specified room type and email the results using a configured email tool.

Explore how a copilot studio agent uses the computer use tool with a Windows computer. Configure templates and inputs to retrieve years in service and automate web tasks.

Test and integrate the computer use tool into our agent, demonstrating automated web navigation, data retrieval of years of service, and trigger-based execution for autonomous workflows.

Develop and test a computer use tool to retrieve the average room size for a room type in square meters and report the size, type, and rate in email.

Learn to monitor and analyze Copilot Studio agent performance using published agents, with configurable date ranges, engagement, CSAT, tool usage, and outcomes across main and child agents.

Access telemetry data from your copilot studio agent with azure application insights, log custom events, query with KQL, and monitor performance metrics to tune models.

Learn how to test agents in Copilot Studio by creating test sets in the evaluation tab, uploading CSV test cases, and applying quality metrics like relevance and completeness.

Explore application lifecycle management in Power Platform environments, creating development, test, and production solutions (unmanaged vs managed), exporting and importing changes to safely test before production.

Design the ALM process to move solutions between development, test, and production using export, import, and pipelines, while using git, azure devops, or github for version control.

Secure agents in Co-Pilot Studio by monitoring protection status, configuring authentication, and applying content moderation and threat detection; govern data with Purview, sensitivity labels, and customer managed encryption keys.

Analyze common artificial intelligence vulnerabilities and mitigations, such as restricting agent sharing, enforcing authentication, and hardening tools and prompts to prevent data leakage.

Explore governance for agents by configuring data policies, retention schedules, and advanced connector policies to prevent data leakage, classify and retain data, and enforce privacy requirements.

Create and customize agents in Microsoft Foundry, compare it with CoPilot Studio, set up a project, add tools and data sources, and test agent performance for enterprise AI.

compare models and tools in Microsoft Foundry, review pricing, technical specs, deployments, and the model leaderboard to select the right model for your agent.

Deploy a GPT nano model in Microsoft Foundry, create a room rates agent, upload the pdf, set temperature to 0.2, enable voice mode, and test the outputs.

Learn to create and deploy fine-tuned AI models in Microsoft Foundry, using base models, distillation, and three tuning methods. Understand when to fine-tune, training types, data, and deployment workflow.

Learn to create custom models with Fireworks inference runtime, configure with a config.json for architecture and hyperparameters, import weights, and scale throughput to deploy models in regions for exam prep.

Explore the use cases for customized small language models, and contrast them with large language models, highlighting PHI and Mistral SLMs, MAAS deployment, and fine-tuning benefits.

Explore the model router in Microsoft Foundry to route prompts to the most suitable models (DeepSeek, Claude, GPT Nano) for optimal quality, cost, and latency.