[EPIC] Copilot expansion inside azd

[kristenwomack]

Problem statement

azd has shipped its first agentic flows — AI-assisted init, hosted agent management — and they're already helping developers get to Azure faster. Now we're looking at the next set of opportunities: tasks where a developer currently has to read docs, copy-paste configurations, and manually wire up environment variables, connection strings, or observability settings across services. These are tasks where Copilot inside azd can add real value — not by replacing developer judgment, but by handling the mechanical work so developers can focus on what matters.

The next step: defining the specific scenarios that will deliver the most value and then building them.

Vision

Developers working in azd encounter moments where Copilot steps in and handles complexity — not as a gimmick, but as genuinely useful assistance on real tasks. "I didn't know azd could do that" becomes a common reaction, and the scenarios we ship demonstrate that azd is the AI-powered front door to Azure.

Who this helps

• New Azure developers: May not yet know how to configure OTel, wire up environment variables, or set up connections between services. Copilot guidance inside azd bridges the knowledge gap with in-context assistance.
• Experienced Azure developers: Know how to do the configuration work but would rather not. A Copilot-powered command that handles environment and connection management saves time on every project setup.
• Copilot CLI users: Already use AI-assisted workflows and expect the experience to get richer over time. New Copilot-powered commands deliver on that promise.
• Platform engineers: AI-assisted configuration helps close the gap between "the recommended way to set things up" and how setups actually happen in practice, leading to more consistency across teams.

Goals (in scope)

• Define 1-2 concrete AI-assisted scenarios (April); candidates include:
  • Environment and connection management: AI-assisted wiring of environment variables, connection strings, and service bindings
  • OpenTelemetry configuration: Copilot-guided OTel setup for observability across services
  • Context management: AI-assisted understanding and management of application context (what does my app need?)
• Ship at least one new Copilot-powered command beyond the agentic flows that exist today
• Validate the scenario with real developer usage and feedback

Non-goals (out of scope)

• "Making everything AI": this is scoped to 1-2 high-value scenarios, not a broad AI transformation
• Owning org-wide Copilot strategy: azd owns its own Copilot integration, not the company-wide approach
• Replacing existing non-AI commands with AI versions: new capabilities, not rewrites
• Shipping scenarios without clear alignment on priorities; if scenarios aren't defined by end of April, the timeline shifts rather than shipping something unvalidated

Success criteria

• 1-2 concrete Copilot-powered scenarios defined and prioritized (end of April)
• At least one new Copilot-powered command implemented and shipping
• Developer feedback collected on shipped scenario: qualitative signal that it's useful, not just novel
• Scenario success metric has a measurable data point from this work
• Scenario documented: what it does, how to use it, and how to extend the pattern

Dependencies

• Scenario definition: This epic cannot proceed to implementation without concrete scenario commitments. April is the decision window.
• Copilot CLI platform capabilities: Depending on the chosen scenarios, we may need specific Copilot CLI platform features (context passing, tool calling, etc.)
• Cross-pillar: If the chosen scenario involves azd ai flows, coordination with the Hosted Agents pillar is required. If it involves init flows, coordination with the UX Revamp pillar is required.
• Evals epic: Once a Copilot scenario ships, it should be added to the eval framework (see Evals for azd Core) to measure ongoing quality.