AI Enablement Engagement

Session 5 — Azure DevOps Integration (Detailed Agenda)

Total Duration: 2.5 hours (150 minutes) Phase / Week: Phase 2, Week 4 Format: Hands-on workshop with live pipeline changes Required attendees: Turner developers; DevOps/platform engineer with permissions to modify a pilot pipeline

Pre-session prep (Turner)

  • Confirm pipeline changes to the learning bed will be made in a non-production branch
  • Provision a service account / token for AI API access from pipeline agents
  • Confirm the DevOps engineer authorized to commit pipeline changes will be present

Agenda

0:00 – 0:15 (15 min) — Where AI fits in the Azure DevOps lifecycle

  • Lead: Nubitz
  • Format: Briefing
  • Map AI's role across the lifecycle: backlog refinement, PR workflow, automated checks, test scaffolding, release notes. Frame the pipeline patterns we'll build today.

0:15 – 0:40 (25 min) — Backlog and story workflow with AI

  • Lead: Nubitz, with Turner participants
  • Format: Demo + light exercise
  • Use AI to refine a real story, draft acceptance criteria, and decompose an epic. Discuss when to trust the output vs. revise heavily.

0:40 – 1:15 (35 min) — Exercise A: Wire AI into the learning bed's PR workflow

  • Lead: Developers and DevOps engineer; Nubitz floating
  • Format: Hands-on
  • Add an AI-assisted PR description generator or change summarizer to the learning bed's pipeline. Open a PR and observe results. Iterate on the prompt and the trigger conditions.

1:15 – 1:25 (10 min) — Break

1:25 – 2:00 (35 min) — Exercise B: Automated code vetting or test scaffolding step

  • Lead: Developers and DevOps engineer; Nubitz floating
  • Format: Hands-on
  • Add one of: AI-assisted code review comment, AI-driven test gap detection, or AI-assisted change-risk summary to the learning bed's pipeline. Choose based on what the team will use. Wire it as a non-blocking pipeline step first.

2:00 – 2:20 (20 min) — Pipeline guardrails and failure modes

  • Lead: Nubitz, facilitating
  • Format: Discussion
  • Permission scoping for AI service accounts. Deterministic vs. non-deterministic gates. Cost and rate-limit awareness. Prompt leakage in logs. Capture the agreed guardrails as an addition to the governance framework.

2:20 – 2:30 (10 min) — Synthesis and rollout pattern

  • Lead: Nubitz
  • Format: Synthesis
  • Document the pattern Turner will replicate to additional pipelines. Preview Session 6 capstone.

Materials

  • Reference YAML for AI-augmented pipeline steps (sanitized; Nubitz)
  • Guardrails one-pager (Nubitz)
  • Cost/rate-limit estimation worksheet (Nubitz)

What Nubitz produces after this session

  • At least one AI-augmented pipeline step running in the learning bed's pipeline
  • Reusable pipeline pattern documentation that Turner can apply to other products
  • Pipeline-integration content for the Phase 2 deliverable

Risks to manage

  • Permission delays: without pipeline edit permission the session degrades to a demo — confirm 72 hours in advance.
  • *Production risk: