AI Data Engineering Use Case Config

Replace {{PLACEHOLDERS}} with your environment values, then deploy to your stack.

# =============================================================================
# AI-ENABLED DATA ENGINEERING USE CASE CONFIG
# Define scope, guardrails, prompts, and success metrics for AI-assisted DE work.
# Plug in values and use to standardize AI tool rollouts across your team.
# =============================================================================

use_case:
  id: "{{USE_CASE_ID}}"                    # e.g. UC-DE-001
  name: "{{USE_CASE_NAME}}"                # e.g. SQL Copilot for Analytics Engineers
  owner: "{{OWNER_TEAM}}"
  sponsor: "{{BUSINESS_SPONSOR}}"
  status: draft                            # draft | pilot | production | retired
  created_at: "{{CREATED_DATE}}"

problem:
  statement: |
    {{PROBLEM_STATEMENT}}
    # e.g. Analytics engineers spend 4+ hours/week writing repetitive SQL
    # for ad-hoc requests that follow known patterns.

stakeholders:
  primary_users:
    - role: "{{PRIMARY_USER_ROLE}}"        # e.g. Analytics Engineer
      count_estimate: {{USER_COUNT}}
  approvers:
    - "{{TECHNICAL_APPROVER}}"
    - "{{SECURITY_APPROVER}}"

data_scope:
  allowed_sources:
    - database: "{{DATABASE}}"
      schemas: ["{{SCHEMA_1}}", "{{SCHEMA_2}}"]
  forbidden_data:
    - pii_columns: ["email", "phone", "ssn"]
    - raw_tables: ["{{RAW_PII_TABLE}}"]
  max_query_cost_credits: {{MAX_QUERY_CREDITS}}   # e.g. 5

ai_capability:
  type: "{{AI_CAPABILITY_TYPE}}"           # sql_generation | doc_generation | anomaly_explanation | pipeline_codegen
  model: "{{MODEL_NAME}}"                    # e.g. gpt-4o, claude-sonnet
  integration_point: "{{TOOL_NAME}}"         # e.g. Cursor, dbt Cloud, custom CLI

prompts:
  system: |
    You are a senior data engineer assistant for {{COMPANY_NAME}}.
    Only generate SQL against approved schemas: {{SCHEMA_1}}, {{SCHEMA_2}}.
    Never expose PII. Always include LIMIT 1000 on exploratory queries.
    Require human review before any DDL or production deployment.

  user_template: |
    Context: {{BUSINESS_CONTEXT}}
    Task: {{TASK_DESCRIPTION}}
    Target table: {{DATABASE}}.{{SCHEMA}}.{{TABLE}}
    Output format: SQL + brief explanation

guardrails:
  require_human_review: true
  block_ddl: true
  block_cross_database: true
  max_tokens: 4096
  log_all_prompts: true
  retention_days: 90

success_metrics:
  - metric: time_saved_hours_per_week
    baseline: {{BASELINE_HOURS}}
    target: {{TARGET_HOURS}}
  - metric: sql_error_rate
    baseline: {{BASELINE_ERROR_RATE}}
    target: {{TARGET_ERROR_RATE}}
  - metric: user_adoption_pct
    baseline: 0
    target: {{ADOPTION_TARGET_PCT}}

risks:
  - id: R1
    description: Hallucinated joins producing incorrect metrics
    mitigation: Mandatory peer review + automated query linting
    severity: high
  - id: R2

// ... download full template for remaining code