Summary
Throughout this module, you learned how to configure the generative AI capabilities that power an agent like the Woodgrove Bank client advisor, from knowledge retrieval through response shaping and structured reasoning tasks.
What you learned
- Configure a generative answers node: how to add a generative answers node to a topic, connect it to targeted knowledge sources, and apply node-level source priority to scope retrieval precisely.
- Connect custom knowledge sources: how to supply data from a proprietary API and Power Automate flows as custom data, using the required Power Fx Table format so the AI can synthesize grounded responses from sources outside standard connectors.
- Shape responses with custom instructions: how to write custom instructions to control the tone, format, and scope of AI output within a generative answers node, turning vague quality goals into specific, observable behaviors.
- Author custom prompts: how to build a reusable custom prompt in Prompt builder, define input variables for runtime data, select the right model for the task, and connect a Microsoft Foundry model for scenarios where managed models aren't sufficient.