Module 7 of 11
Orchestrate configuration of prebuilt agents and apps
Enable, configure, and orchestrate the AI already shipped in Dynamics 365 and Microsoft 365 before building anything new.
Exam questions Q61–Q70 10 units on Learn
Open module 7 on Microsoft Learn ↗ Watch the episode (52 min)
Why this module matters
This module is the practical counterpart to the strategy modules. A great deal of AI capability already exists inside Dynamics 365 and Microsoft 365; the architect's job is to know it exists, confirm the prerequisites, and orchestrate it coherently across apps.
Exam questions here reward restraint. The correct answer is usually to check whether the out-of-the-box capability meets the need before proposing anything custom.
The one big ideaConfigure before you extend, and extend before you build. Most "we need an agent" requests are satisfied by enabling something that already ships.
What Microsoft Learn covers
Units in this module, as published on Microsoft Learn:
- Introduction
- Orchestrate AI features in Dynamics 365 Finance and Supply Chain
- Design AI solutions for Dynamics 365 Customer Service
- Propose Microsoft 365 agents for business scenarios
- Orchestrate and configure Microsoft 365 Copilot for sales and service
- Propose Microsoft Power Platform AI features
- Design interoperable agent experiences for Finance and Operations
- Recommend process knowledge sources for in-app help in Dynamics 365
- Module assessment
- Summary
Watch this module
Microsoft Learn publishes a free video course for AB-100 on YouTube. This module is covered by one 52 min episode. An instructor walking through the module is often faster than reading it twice.
Key concepts and definitions
| Prebuilt / OOB capability | AI shipped inside the product, enabled by an administrator and tuned through supported configuration. |
| Feature enablement gates | What determines whether a user sees a capability: licensing, admin feature enablement, security roles and privileges, and environment/region availability. |
| Process knowledge source | Curated procedural content powering in-app help and guidance inside Dynamics 365. |
| Interoperable agent experience | Agents that work coherently across Finance and Operations and Customer Experience apps rather than in isolation. |
| Power Platform AI features | AI capabilities available across Power Apps, Power Automate, and Copilot Studio that often satisfy a requirement without custom development. |
| Routing ambiguity | When several agents claim overlapping scope, users and orchestrators cannot tell which should answer. Fixed by curating scope, naming, and discovery. |
| Configuration baseline | The documented, intended configuration of a capability across environments, used to detect and correct drift. |
| Configuration drift | Environments diverging over time through manual, undocumented changes. |
Design guidance
Rollout sequence
- Validate prerequisites and licensing → enable in a non-production environment → configure and test → pilot with real users → broad rollout with training and comms.
- Never enable broadly in production first and communicate afterwards.
- Before broad enablement, complete a data-access review, confirm DLP policies, agree the environment strategy, and record a named owner.
Tailoring without building
- Adapt answers to company terminology through grounding sources, custom instructions, and supported configuration — not by fine-tuning a base model.
- Progression when a capability nearly fits: configure → extend → build custom, in that order.
- For in-app help, recommend curated process knowledge sources rather than pointing at unstructured document dumps.
Orchestrating multiple agents
- Curate the catalogue: distinct scopes, clear naming, and controlled discovery so users are not choosing blindly.
- Manage configuration through documented baselines and automated deployment or validation where supported.
- Validate success by adoption plus movement in the targeted business metric, supported by user feedback.
Good practice vs. anti-pattern
Do
- Check the out-of-the-box capability first.
- Test in a non-production environment before pilot.
- Review source permissions before broad enablement.
- Curate agent scope and naming to avoid overlap.
Avoid
- Fine-tune a model to teach it company vocabulary.
- Enable in production and figure out governance later.
- Configure each environment by hand.
- Measure success by licences assigned or settings changed.
Exam traps to watch for
TrapIf a prebuilt agent surfaces data a user should not see, the cause is almost always overly broad source permissions that the agent faithfully honours — not hallucination and not model size.
TrapVisibility of a prebuilt feature is governed by licensing, admin enablement, security roles, and environment/region — not by browser, device, or display language.
TrapPrebuilt does not mean pre-governed. DLP, data-access review, environment strategy, and ownership still apply.
TrapThe near-fit progression is configure → extend → build. Any option that jumps straight to custom development is wrong.
Readiness checklist
You are ready to move on when you can do all of these without notes.
- List the four gates that determine whether a user sees a prebuilt Copilot feature.
- Give the correct rollout sequence for enabling a prebuilt capability.
- Explain how to tailor terminology without touching the model.
- Diagnose why an agent is exposing data it should not.
- Describe how to prevent configuration drift across environments.