Design AI-Powered Business Solutions
This domain covers the architecture, configuration, and orchestration of AI agents, Copilot customizations, and intelligent features across Microsoft's business application stack — including Dynamics 365, Power Platform, Copilot Studio, Microsoft 365 Copilot, and Microsoft Foundry. It accounts for 27% of the AB-100 exam, making it the highest-weighted domain. Questions test your ability to choose the right AI tool for a given requirement, design agent topologies, apply well-architected principles, and configure extensibility patterns — not just recall definitions.
Copilot Customizations and Connectors for Dynamics 365 Apps
Copilot in Dynamics 365 apps (Sales, Customer Service, Field Service, Finance, Supply Chain) ships with prebuilt Copilot capabilities that can be extended without writing code. Customization surfaces include:
- Copilot Studio — the low-code canvas for building and extending agents that surface inside Dynamics 365 apps
- Custom prompt plugins — authored in Copilot Studio or AI Builder, injected into Dynamics 365 Copilot via plugin actions
- Connectors — Power Platform connectors (certified, custom, or virtual-table-backed) that expose external data to in-app Copilot
- Dataverse virtual tables — surface external system data to Copilot without migration; Copilot treats them as native Dataverse records
| Customization type | Requires code? | Deployment target | Data stays external? |
|---|---|---|---|
| Custom topics in Copilot Studio | No | Dynamics 365 embedded agent | Yes (via connector) |
| Custom prompt plugin (AI Builder) | No | Copilot in D365 app | Depends |
| Custom connector (Power Platform) | Minimal (OpenAPI) | Any agent/flow | Yes |
| Virtual table | No | Dataverse / Copilot | Yes |
| PCF control with AI integration | Yes | Model-driven app | No |
Common traps:
- Exam distractors will suggest "create a canvas app" when the requirement is to extend Copilot inside a Dynamics 365 app — this is wrong; use Copilot Studio topics or plugin actions.
- Virtual tables do not copy data — selecting them over full data migration is correct when latency is acceptable and the source system must remain authoritative.
- Custom connectors require an OpenAPI 2.0 (Swagger) definition — not OpenAPI 3.0 natively without a converter.
Agents for Dynamics 365 Contact Center Channel Integration
Dynamics 365 Contact Center (formerly Omnichannel for Customer Service) supports AI agent integration via:
- Voice channel — integrated with Azure Communication Services; Copilot Studio agents handle IVR flows
- Digital messaging channels — chat, SMS, WhatsApp, Facebook; agents hand off to human agents via escalation topics
- Unified routing — classifies and routes work items; AI models score sentiment and priority before assignment
Agent integration points:
| Component | Role | Key config |
|---|---|---|
| Copilot Studio bot (agent) | Handles first-contact deflection | Connected via Azure Bot Service channel |
| Escalation topic | Transfers to human agent | Must end with "Transfer conversation" node |
| Context variables | Pass session data on escalation | Declared in Copilot Studio; mapped in Omnichannel |
| Real-time transcription | Feeds Copilot suggestions to agents | Enabled per workstream, requires D365 Customer Service Enterprise |
| Sentiment analysis | Automatic in Omnichannel | Built-in; no separate config |
Common traps:
- The exam distinguishes Dynamics 365 Contact Center (standalone CCaaS SKU) from Customer Service + Omnichannel add-on — licensing matters for feature availability.
- Context variables must be explicitly mapped in the Omnichannel workstream configuration; they do not auto-transfer.
- Agents that handle voice must use DTMF or speech recognition nodes in Copilot Studio — not standard message nodes.
Unlock with a paid plan
20 more sections (about 3,732 words) continue below. Every paid plan unlocks all AB-100 study guides, the full question bank and mock exams, from $4.97/month.