{
  "id": "workflow:llm-cost-optimization",
  "_kind": "Workflow",
  "_file": "workflows/workflows/workflows-ai-era.yaml",
  "_cluster": "workflows",
  "attributes": {
    "displayName": "LLM Cost Optimization",
    "workflowKind": "operational",
    "triggerType": "scheduled",
    "typicalCadence": "weekly",
    "complexity": "cross-team",
    "description": "Reviews and optimises spend across LLM API providers — analysing per-model\ntoken consumption, identifying prompt-length bloat, evaluating cache-hit\nrates, testing cheaper model substitutions for low-criticality tasks,\nauditing retry and fallback policies that inflate costs, and projecting\nbudget burn-rate against forecasts. Produces a cost breakdown dashboard\nand actionable savings plan. Excludes model fine-tuning work.\n"
  },
  "outgoingEdges": [
    {
      "from": "workflow:llm-cost-optimization",
      "to": "role:ml-engineer",
      "kind": "involves_role",
      "attributes": {}
    },
    {
      "from": "workflow:llm-cost-optimization",
      "to": "role:cost-tracker",
      "kind": "involves_role",
      "attributes": {}
    },
    {
      "from": "workflow:llm-cost-optimization",
      "to": "role:staff-engineer",
      "kind": "involves_role",
      "attributes": {}
    },
    {
      "from": "workflow:llm-cost-optimization",
      "to": "skill-area:prompt-engineering",
      "kind": "requires_skill_area",
      "attributes": {}
    },
    {
      "from": "workflow:llm-cost-optimization",
      "to": "skill-area:context-management",
      "kind": "requires_skill_area",
      "attributes": {}
    },
    {
      "from": "workflow:llm-cost-optimization",
      "to": "domain:ml-ops",
      "kind": "applies_to_domain",
      "attributes": {}
    },
    {
      "from": "workflow:llm-cost-optimization",
      "to": "domain:platform-engineering",
      "kind": "applies_to_domain",
      "attributes": {}
    },
    {
      "from": "workflow:llm-cost-optimization",
      "to": "responsibility:cost-optimization",
      "kind": "triggers_responsibility",
      "attributes": {}
    },
    {
      "from": "workflow:llm-cost-optimization",
      "to": "responsibility:capacity-planning",
      "kind": "triggers_responsibility",
      "attributes": {}
    },
    {
      "from": "workflow:llm-cost-optimization",
      "to": "org-unit:ml-platform-team",
      "kind": "performed_by_org_unit",
      "attributes": {}
    },
    {
      "from": "workflow:llm-cost-optimization",
      "to": "org-unit:infra-engineering",
      "kind": "performed_by_org_unit",
      "attributes": {}
    }
  ],
  "incomingEdges": [
    {
      "from": "tool:fireworks-ai",
      "to": "workflow:llm-cost-optimization",
      "kind": "supports_work",
      "attributes": {
        "confidence": "medium",
        "evidence": "Inference provider choice affects cost, latency, and throughput trade-offs."
      }
    },
    {
      "from": "tool-server:mcp-fireworks-ai-candidate",
      "to": "workflow:llm-cost-optimization",
      "kind": "supports_work",
      "attributes": {
        "confidence": "medium",
        "evidence": "Provider and model telemetry supports cost and latency optimization."
      }
    }
  ]
}