Templates to Infrastructure Agent

Written By chip baker

From template ecosystems to agent-driven infrastructure orchestration

Project Overview

This project explored the evolution of Azure's developer tooling from curated CLI templates to agent-driven infrastructure orchestration. Templates (Bicep and Terraform) provided standardized security, performance, and compliance baselines, reducing onboarding friction for developers.

Building on this foundation, the project envisioned a future where AI agents could dynamically generate, edit, and deploy infrastructure plans, bridging gaps left by static templates.

My Role

I served as a cross-organizational connector and strategic advisor, bridging product silos to create a cohesive developer experience. My contributions included:

  • Breaking down silos: Identified that templates living at https://azure.github.io/ai-app-templates/ had limited discoverability for developers working in the Azure portal, and advocated for tighter integration between the portal and IDE experiences

  • Building cross-divisional alignment: Sought out designers working on templates across different divisions and established regular cross-divisional team meetings to coordinate efforts and share learnings

  • Strategic influence: Championed thinking beyond product boundaries toward holistic customer journeys, connecting teams that hadn't previously collaborated

  • Agentic vision: Leveraged expertise from previous AI/agents work to guide the evolution from static templates to intelligent orchestration

The Challenge

Even with curated templates, developers faced:

  • Discovery gap: Valuable templates existed at https://azure.github.io/ai-app-templates/ but were invisible to Azure portal users (Typically IT but increasingly Developers)

  • Siloed thinking: Teams operated in product silos rather than considering the full customer journey across portal, IDE, and CLI

  • Manual effort: Developers still needed to compare architectures and configure deployments manually

  • Lack of dynamic guidance: No intelligent support for complex workload decisions

  • Fragmented experiences: Disconnected onboarding flows across infrastructure provisioning touchpoints

The core challenge: How could we connect existing template assets across Azure's ecosystem and evolve them into dynamic, AI-driven workflows that accelerate deployment while reducing manual toil?

Vision / Concept Summary

The concept envisioned agent-driven infrastructure orchestration built on cross-product integration:

  • Connected ecosystem: Templates discoverable and accessible across portal, IDE, and CLI—not siloed by product boundaries

  • AI-powered discovery: Agents help users find and select the right templates through natural language

  • Dynamic plan generation: AI agents generate infrastructure plans from prompts, using templates as intelligent starting points

  • Interactive refinement: Users preview, edit, and approve plans through conversational or visual interfaces

  • Intelligent comparisons: Side-by-side workload configuration comparisons highlight tradeoffs and ensure compliance

  • Flexible orchestration: Templates provide the foundation, but agents add intelligence and adaptability

This created a seamless evolution from static, siloed template experiences to autonomous, agentic workflows spanning the Azure ecosystem.

Key Capabilities Explored

  • Cross-Product Template Discovery — Making https://azure.github.io/ai-app-templates/ templates accessible in portal and IDE

  • AI-Powered Template Matching — Agents help users discover relevant templates through natural language

  • Dynamic Infrastructure Plan Generation — Agents generate plans based on user prompts

  • Plan Previews and Edits — Users refine plans via chat or UI before deployment

  • Artifact Generation — Automatic creation of Terraform/Bicep code

  • Configuration Comparison — Side-by-side comparison of workload setups

  • Seamless Integration — Templates and agents work together across touchpoints to reduce cognitive load

Research & Validation

Validation included:

  • Participating in usability tests for plan generation, editing, and deployment

  • Conducting bug bashes to uncover flow issues and friction points

  • Joining team standups to guide decisions and maintain momentum

  • Iterative prototyping of entry points, interaction models, and artifact previews

  • Incorporating feedback from both internal developers and stakeholders

Key insights:

  • Developers valued comparisons and previews as essential for trust

  • Chat-driven edits made AI-driven planning more approachable

  • Cross-product visibility dramatically improved template utilization

  • Entry point and flow design was critical for adoption

  • Breaking down silos between portal/IDE/CLI improved the overall customer journey

Design Process

Phase 1: Template Ecosystem Foundation

  • Identified the discovery gap for templates outside the portal

Phase 2: Cross-Product Integration

  • Advocated for bringing https://azure.github.io/ai-app-templates/ templates into portal and IDE

  • Sought out designers working on templates across divisions and established cross-divisional team meetings

  • Connected teams working in silos to think about holistic customer journeys

  • Discovered hackathon project exploring AI-powered template discovery

Phase 3: Hackathon to Production

  • Joined forces with hackathon team exploring AI template discovery

  • Guided project from proof-of-concept through production launch

  • Participated in standups, research, bug bashes, and decision-making

  • Connected stakeholders and removed blockers

Phase 4: Agentic Orchestration

  • Applied learnings from previous AI/agents work to guide UX vision

  • Ensured consistency with Agentic UX Playbook ( which I also help to author)

Phase 5: Testing & Iteration

  • Validated concepts through bug bashes and usability sessions

Impact & Outcomes

  • Broke down product silos: Connected portal, IDE, and CLI experiences around a unified template strategy

  • Built design community: Created cross-divisional meetings that brought together designers working on templates, establishing coordination where none existed

  • Increased template discoverability: Made valuable templates accessible where developers actually work

  • Positioned Azure for AI-driven infrastructure: Established foundation for agent-driven orchestration

  • Influenced strategic direction: Shaped roadmap decisions for Copilot and future agentic experiences in cloud infrastructure

  • Created cross-functional momentum: Connected dots across teams, enabling collaboration that hadn't existed before

What I Delivered

  • Cross-organizational influence and strategic guidance connecting siloed teams

  • Cross-divisional designer coordination through regular team meetings

  • End-to-end UX flows for infrastructure agents

  • Entry point designs for agent interaction across portal and IDE

  • Template standardization and guided onboarding workflows

  • Advocacy and execution plan for cross-product template integration

  • Hands-on participation in standups and bug bashes from hackathon through production

  • Usability testing insights and recommendations for roadmap integration

  • Customer journey thinking that bridged product boundaries

Visual Artifacts

  • Template ecosystem dashboards showing cross-product integration

  • Before/after flows showing silo vs. connected experiences

  • Infrastructure plan previews and comparison mockups

  • Agentic workflow flows (discovery, generation, edit, deploy)

  • Prototype screenshots and Figma designs

  • Customer journey maps spanning portal, IDE, and CLI

  • Usability testing feedback visualizations

  • Hackathon-to-production evolution timeline

chip baker
Previous

Guided Stack Planning (Concept)
Next

Blog Post Title Four