Technical requirementsProvisioning an AI engine for pipeline integrationChoosing the AI engineProvisioning Azure OpenAI for the pipelineDesigning deterministic AI workflow stepsStep 1: Creating the new AI release workflowStep 2: Collecting pull request metadataStep 3: Generating release notes with AIStep 4: Validating the generated release notesStep 5: Publishing the draft releaseStep 6: Testing the workflow with a pull requestStep 7: Reviewing the generated releaseCreating a reusable Azure OpenAI composite actionStep 1: Creating the composite action folderStep 2: Defining the action inputsStep 3: Defining the action outputsStep 4: Validating inputs and dependenciesStep 5: Calling Azure OpenAI from the composite actionStep 6: Updating the release workflowStep 7: Fixing the composite action pathReducing prompt and input complexityStep 1: Updating the PR metadata collection branchStep 2: Adding truncation and sanitization helpersStep 3: Updating the PR information fieldsStep 4: Remove unnecessary metadata fieldsStep 5: Adding API call guardrailsStep 6: Strengthening the prompt instructionsStep 7: Testing the updated workflowManaging AI costs and performanceStep 1: Skipping AI calls when there are no PRsStep 2: Adding retry and backoff logicStep 3: Handling success, failure, and retry attemptsStep 4: Tracking token consumptionStep 5: Creating the telemetry artifactStep 6: Adding conditions to validation and publishing jobsStep 7: Testing the full flow with a pull requestAdding observability and feedbackSecure AI integration patterns in CI/CDExtending the workflow with AI-augmented ChatOpsStep 1: Creating the ChatOps workflowStep 2: Creating the ChatOps promptStep 3: Creating the GitHub tokenStep 4: Updating the workflow with the tokenStep 5: Adding and testing the workflowSummary