{
  "source": "Cooked Index — occupational AI risk register",
  "page": "https://cookedindex.com/jobs/bartenders/",
  "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": "Bartenders",
    "soc_code": "35-3011",
    "category": "Food",
    "us_employment": 756390,
    "median_annual_wage": 34340
  },
  "verdict": "SAFE",
  "risk_resistance": 68,
  "contested": false,
  "near_boundary": false,
  "dimensions": {
    "task_resistance": 16,
    "embodiment": 18,
    "liability_shield": 8,
    "trust_premium": 15,
    "judgment_accountability": 11
  },
  "reasoning": {
    "task_resistance": "A bartender's shift is building four different drinks at once from muddled mint, chilled glassware and hand-cut citrus while tracking six open tabs and the guy at the end who's had enough — the only pieces software touches are the par-level sheet and the pour-cost spreadsheet, which is why this sits at 16 and not 19.",
    "embodiment": "You are standing on rubber mats for eight hours, hauling kegs and ice bins, wiping spills, changing CO2 lines, and reaching across a wet crowded well that no two nights configure the same way — the environment is unpredictable enough that robot arms only work behind a rail with a fixed 12-drink menu.",
    "liability_shield": "You hold a TIPS or state alcohol-server card and dram-shop statutes can put liability on you for over-serving, but the licence that actually matters is the establishment's liquor licence held by the owner — your certification is a day-long course anyone can pass, so it blocks nothing structurally.",
    "trust_premium": "Regulars pick a bar because of who is behind it — you know their drink, you comp the third round, you remember the divorce — and that recognition is why they tip 20% instead of using the self-pour wall, though a large share of volume is still one-time tourists and wedding guests who'd take a drink from anyone.",
    "judgment_accountability": "Cutting someone off, spotting a fake ID, deciding when a table's argument becomes a door situation, and calling a cab are calls you make in seconds with real legal and safety consequences, but they run against house policy and posted ID rules rather than open-ended professional judgment."
  },
  "rationale": "Mixing drinks, pulling drafts, restocking coolers, cutting fruit, reading a crowded room and defusing a drunk customer are physical, social, real-time tasks that language models cannot touch and current robotics handles only in novelty kiosks with fixed menus. What AI does erode is the paperwork edge: inventory counts, pour-cost analysis, cocktail menu drafting, and scheduling. The bigger threats to this job are self-pour taps, tablet ordering, and tip-economics — not an AI that can hold a bar.",
  "outlook": "Still hundreds of thousands of these jobs in 2035, but concentrated in craft, hotel, and high-volume social venues while casual-dining and self-pour formats quietly shed bar headcount.",
  "what_would_raise_it": {
    "levers": [
      {
        "dimension": "liability_shield",
        "change": "State ABC boards moving from voluntary to mandatory certified server training with personal liability — e.g. states adopting mandatory RBS/TIPS-style certification (California's RBS program since 2022, Washington MAST) plus dram-shop statutes that name the individual pourer, not just the licensee. Extension of mandatory certification to the remaining ~20 states without it, and any rule that self-pour taps and tablet ordering require a certified human to authorize each additional pour, would make a licensed human the legally required checkpoint on intoxication calls.",
        "plausibility": "already happening",
        "would_add": 5
      },
      {
        "dimension": "judgment_accountability",
        "change": "Dram-shop and social-host liability litigation increasingly turning on the specific cut-off decision, plus insurer requirements (liquor liability carriers demanding documented refusal-of-service logs signed by the bartender on shift) that formalize the 'who decided to stop serving' call as an auditable, attributable judgment rather than informal discretion.",
        "plausibility": "plausible",
        "would_add": 4
      },
      {
        "dimension": "trust_premium",
        "change": "Continued growth of the craft cocktail and omakase-bar segment where the bartender is the product and menus are bespoke — plus any consumer backlash making 'staffed bar, no QR ordering' an advertised feature. This is a segment effect: it raises the premium for maybe the top tier of venues, not for the high-volume chain and stadium bars where most of the 756k work.",
        "plausibility": "plausible",
        "would_add": 2
      },
      {
        "dimension": "task_resistance",
        "change": "Task-mix shift as self-pour walls and tablet ordering absorb the routine beer-and-well-drink tier, leaving crowd-reading, intoxication assessment, conflict de-escalation and bespoke build as the residual job. Note this shrinks headcount while raising per-role resistance — a smaller job that is harder to automate.",
        "plausibility": "already happening",
        "would_add": 2
      }
    ],
    "ceiling_note": "The realistic ceiling is roughly high-70s. The binding constraint is not AI capability but labor economics: self-pour, tablet ordering and tip-credit changes reduce the number of bartender slots without any model needing to mix a drink. A higher liability shield protects the role's content while headcount still falls."
  },
  "adjudication": null,
  "employment_history": {
    "points": [
      {
        "y": 2017,
        "emp": 613350,
        "wage": 21690
      },
      {
        "y": 2018,
        "emp": 631480,
        "wage": 22550
      },
      {
        "y": 2019,
        "emp": 646850,
        "wage": 23680
      },
      {
        "y": 2020,
        "emp": 486720,
        "wage": 24960
      },
      {
        "y": 2021,
        "emp": 485330,
        "wage": 26350
      },
      {
        "y": 2022,
        "emp": 613070,
        "wage": 29380
      },
      {
        "y": 2023,
        "emp": 711140,
        "wage": 31510
      },
      {
        "y": 2024,
        "emp": 745610,
        "wage": 33530
      },
      {
        "y": 2025,
        "emp": 756390,
        "wage": 34340
      }
    ],
    "from": 2017,
    "to": 2025,
    "change_pct": 23.3,
    "comparable_from": 2019,
    "spans_soc_revision": true
  },
  "pivots": [],
  "license": "https://cookedindex.com/terms"
}