{
  "source": "Cooked Index — occupational AI risk register",
  "page": "https://cookedindex.com/jobs/metal-workers-and-plastic-workers-all-other/",
  "methodology": "https://cookedindex.com/methodology",
  "notice": "Verdicts are re-examined as evidence accumulates. Re-fetch before relying on this; the page above always carries the current score.",
  "scored_at": "2026-08-11",
  "model": "claude-opus-5",
  "occupation": {
    "title": "Metal Workers and Plastic Workers, All Other",
    "soc_code": "51-4199",
    "category": "Production",
    "us_employment": 15900,
    "median_annual_wage": 45950
  },
  "verdict": "EXPOSED",
  "risk_resistance": 43,
  "contested": false,
  "near_boundary": false,
  "dimensions": {
    "task_resistance": 14,
    "embodiment": 16,
    "liability_shield": 2,
    "trust_premium": 4,
    "judgment_accountability": 7
  },
  "reasoning": {
    "task_resistance": "At 14 the work sits just above the automation line because the tasks that land in this catch-all bucket — hand-deburring odd castings, shimming a fixture that won't seat, straightening warped plastic trim, feeding short-run parts nobody wrote a program for — are the leftovers precisely because they resisted the cells that already took the high-volume work, but they are physical routines a robot arm with decent vision will eventually reach, which is why this isn't 18.",
    "embodiment": "16 reflects standing at a press, oven, or bench for a full shift handling hot stock, sharp flash and slippery coolant, where grip force, part orientation by feel, and reacting to a jam are the job — not 19 or 20 only because the shop floor is an indoor, fixtured environment with known machines rather than a field site or crawlspace.",
    "liability_shield": "A 2 is honest: there is no state license to trim plastic or tend a heat-treat furnace, and while a shop may carry AWS or ISO/AS9100 requirements, those certify the employer's process and the QC signoff, not you personally — the 2 rather than 0 acknowledges forklift or crane certs and some welding-ticket holders in the bucket.",
    "trust_premium": "4 is where an anonymous part number puts you — the customer receives a bracket, not you, and the only relationship that shields the job is the shift lead who knows you can eyeball a bad shot; that internal familiarity is worth 4, not 0, but it doesn't survive a plant relocation.",
    "judgment_accountability": "7 puts you at the bottom edge of real discretion: scrap-versus-rework calls, deciding when a die needs pulling, and catching drift before the inspection sample does are judgment with money attached, but the print, the spec, and the router define the acceptable answer, and a supervisor or QC signs off on anything expensive."
  },
  "rationale": "This is a catch-all bucket for shop-floor metal and plastic workers who don't fit the named machine-operator titles — hand-forming, trimming, fixture loading, deburring, heat-treat tending, inspecting and reworking parts. Language AI barely touches this work, but purpose-built industrial automation does: CNC cells, robotic tending, and vision inspection have been eating exactly these tasks for two decades and continue to. The moat is physical, not legal or relational — no license, no client relationship, and little decision authority beyond scrap/rework calls, so the job is protected by capital cost and part variety rather than by anything AI can't learn.",
  "outlook": "Headcount in this residual category keeps shrinking as high-volume tasks migrate to automated cells, while workers who can set up, troubleshoot, and salvage low-volume custom work stay in demand and out-earn the loaders.",
  "what_would_raise_it": {
    "levers": [
      {
        "dimension": "task_resistance",
        "change": "Task-mix shift within the bucket: as robotic tending and vision inspection absorb the high-volume repeat work, what remains is low-volume/one-off fixturing, hand-forming of non-repeating geometries, and rework of parts that failed automated inspection — work where each piece is dimensionally unique. Watch for shops reclassifying these workers as 'rework/first-article technicians' tied to prototype and MRO lines rather than production runs.",
        "plausibility": "already happening",
        "would_add": 3
      },
      {
        "dimension": "embodiment",
        "change": "Growth of the defense/aerospace MRO and repair segment, where the workpiece is a used, deformed, corroded part with no as-built CAD model — Navy shipyard and depot-level repair hiring (e.g., NAVSEA public shipyard trades expansion) puts this work in unfixtured, in-situ positions robots cannot reach or model.",
        "plausibility": "plausible",
        "would_add": 2
      },
      {
        "dimension": "liability_shield",
        "change": "Named-operator traceability requirements in safety-critical part chains: AS9100/NADCAP heat-treat and special-process rules already require certified operator sign-off on process runs, and FAA Part 21 production approval requires identified personnel for inspection buy-off. If primes push per-part named-certifier records (as some already do for weld and heat-treat lots) rather than shop-level certification, the individual operator becomes the required signature.",
        "plausibility": "plausible",
        "would_add": 4
      },
      {
        "dimension": "judgment_accountability",
        "change": "Ownership of the disposition call on nonconforming material — if MRB (material review board) authority for scrap/rework/use-as-is is formally delegated to floor certifiers rather than engineers, as happens in lean shops, the role owns a consequential ambiguous call with cost and airworthiness consequences.",
        "plausibility": "plausible",
        "would_add": 3
      }
    ],
    "ceiling_note": "No realistic route to a trust premium: buyers purchase parts to spec and never learn who touched them. Even with every lever, the moat stays capital-cost-and-part-variety, which erodes as cell setup cost falls; ceiling is roughly the mid-50s."
  },
  "adjudication": null,
  "employment_history": {
    "points": [
      {
        "y": 2017,
        "emp": 22470,
        "wage": 35820
      },
      {
        "y": 2018,
        "emp": 25470,
        "wage": 35920
      },
      {
        "y": 2019,
        "emp": 24340,
        "wage": 34830
      },
      {
        "y": 2020,
        "emp": 21140,
        "wage": 35520
      },
      {
        "y": 2021,
        "emp": 18740,
        "wage": 36990
      },
      {
        "y": 2022,
        "emp": 21030,
        "wage": 38340
      },
      {
        "y": 2023,
        "emp": 20370,
        "wage": 41200
      },
      {
        "y": 2024,
        "emp": 20270,
        "wage": 42750
      },
      {
        "y": 2025,
        "emp": 15900,
        "wage": 45950
      }
    ],
    "from": 2017,
    "to": 2025,
    "change_pct": -29.2,
    "comparable_from": 2019,
    "spans_soc_revision": true
  },
  "pivots": [],
  "license": "https://cookedindex.com/terms"
}