{
  "source": "Cooked Index — occupational AI risk register",
  "page": "https://cookedindex.com/jobs/telemarketers/",
  "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": "Telemarketers",
    "soc_code": "41-9041",
    "category": "Sales",
    "us_employment": 58430,
    "median_annual_wage": 35450
  },
  "verdict": "COOKED",
  "risk_resistance": 9,
  "contested": false,
  "near_boundary": false,
  "dimensions": {
    "task_resistance": 2,
    "embodiment": 1,
    "liability_shield": 1,
    "trust_premium": 3,
    "judgment_accountability": 2
  },
  "reasoning": {
    "task_resistance": "The dialer already picks the number, the screen already supplies the pitch and the rebuttal tree, and the only human contribution left is producing the audio — which is precisely what synthetic voice with sub-second latency now does, so this sits at 2 rather than 7 because there is no off-script residue like site inspection or credit assessment to fall back on.",
    "embodiment": "You sit in a headset at a workstation; nothing in the job requires touching, carrying, or going anywhere, and the 1 rather than 0 is only the fact that a body currently occupies a seat in a call center for shift-supervision purposes.",
    "liability_shield": "No state licences telemarketers; TCPA, the Do Not Call Registry, and state telemarketing registration bind the seller and the calling firm, not the individual on the phone, so a compliance violation costs the company statutory damages and costs you the shift, not a credential.",
    "trust_premium": "Nobody on the other end knows your name before you say it, calls are placed from purchased lists to strangers, and the 3 rather than 0 acknowledges only that outbound renewal and warm-lead lists occasionally reach someone who bought before — a list attribute, not a relationship you built.",
    "judgment_accountability": "Your discretion runs to picking rebuttal 3 instead of rebuttal 4 and deciding when to release the call; pricing, offer terms, credit approval, and the disclosure language are all locked upstream, which is why this is 2 and not the 8 a commissioned outside rep would get for negotiating terms."
  },
  "rationale": "The job is reading a script over the phone, handling a short list of predictable objections, and logging the outcome in a CRM — all three are now handled by conversational voice AI at lower cost and unlimited volume. Employment has already fallen sharply from its 2000s peak as predictive dialers and IVR absorbed the work, and current voice models close the remaining gap on tone and objection handling. There is no license, no physical presence, and buyers on the other end of the line are not paying for the relationship.",
  "outlook": "Expect continued steep decline over the next decade, with the residual jobs concentrated in complex B2B selling, campaign design, and oversight of automated calling systems.",
  "what_would_raise_it": {
    "levers": [
      {
        "dimension": "liability_shield",
        "change": "The FCC's Feb 2024 declaratory ruling already makes AI-generated voices 'artificial or prerecorded' under the TCPA, requiring prior express written consent; the pending AI-disclosure NPRM (CG Docket 23-362) would add mandatory 'this is an AI voice' disclosure at call start. If that finalizes AND states layer on sector rules — e.g. an insurance or securities regulator requiring that any solicitation resulting in an application be conducted by a registered/licensed human (FINRA Rule 2210-style supervision of 'retail communications'), the compliant channel for regulated products becomes a named human on a recorded line. Watch state insurance departments and the NAIC AI model bulletin adoption.",
        "plausibility": "already happening",
        "would_add": 5
      },
      {
        "dimension": "task_resistance",
        "change": "Two-tier split is real: the outbound-cold tier is gone, the remaining tier is complex B2B qualification, appointment-setting into six-figure deals, and Medicare Advantage / final-expense sales where the caller must navigate eligibility facts the caller misstates. If AI-disclosure requirements make disclosed-AI calls answer-rate-collapse (already visible in early carrier spam-labeling data), the survivor headcount is entirely judgment-tier and the measured score of the remaining job rises even as employment falls.",
        "plausibility": "plausible",
        "would_add": 4
      },
      {
        "dimension": "trust_premium",
        "change": "Carrier-level authentication (STIR/SHAKEN attestation plus branded caller ID) makes 'verified human agent at a named firm' a purchasable signal; some outbound health-plan and financial firms already market human-only contact as a compliance and brand-safety feature after AI-voice fraud incidents. This is a thin, buyer-side (not consumer-side) premium — the person answering still does not want the call.",
        "plausibility": "plausible",
        "would_add": 2
      },
      {
        "dimension": "judgment_accountability",
        "change": "If CMS's third-party marketing organization rules (already tightened for Medicare Advantage in the 2024/2025 rules, with recorded-call retention and scope-of-appointment requirements) extend to naming an accountable individual agent for each enrollment, the caller personally owns a consequential eligibility call under ambiguity rather than reading a script.",
        "plausibility": "plausible",
        "would_add": 3
      }
    ],
    "ceiling_note": "Even with every lever, this tops out around 25-30 and only for the regulated-product slice (Medicare, insurance, securities); those workers are increasingly classified as insurance agents (41-3021) rather than telemarketers, so the SOC code itself hollows out. Generic B2C outbound has no route at all — no license exists to create, and no consumer has ever preferred the call."
  },
  "adjudication": null,
  "employment_history": {
    "points": [
      {
        "y": 2017,
        "emp": 189670,
        "wage": 24460
      },
      {
        "y": 2018,
        "emp": 164160,
        "wage": 25250
      },
      {
        "y": 2019,
        "emp": 134800,
        "wage": 26290
      },
      {
        "y": 2020,
        "emp": 117610,
        "wage": 27920
      },
      {
        "y": 2021,
        "emp": 115130,
        "wage": 28910
      },
      {
        "y": 2022,
        "emp": 96520,
        "wage": 31030
      },
      {
        "y": 2023,
        "emp": 81580,
        "wage": 34480
      },
      {
        "y": 2024,
        "emp": 66430,
        "wage": 34410
      },
      {
        "y": 2025,
        "emp": 58430,
        "wage": 35450
      }
    ],
    "from": 2017,
    "to": 2025,
    "change_pct": -69.2,
    "comparable_from": 2019,
    "spans_soc_revision": true
  },
  "pivots": [
    {
      "slug": "retail-salespersons",
      "title": "Retail Salespersons",
      "verdict": "EXPOSED",
      "risk_resistance": 36,
      "median_wage": 35410,
      "overlap": 74,
      "skills_to_close": [
        "Mathematics",
        "Learning Strategies",
        "Active Learning",
        "Operation and Control"
      ]
    },
    {
      "slug": "concierges",
      "title": "Concierges",
      "verdict": "EXPOSED",
      "risk_resistance": 41,
      "median_wage": 38950,
      "overlap": 71,
      "skills_to_close": [
        "Coordination",
        "Active Learning",
        "Monitoring",
        "Social Perceptiveness"
      ]
    },
    {
      "slug": "waiters-and-waitresses",
      "title": "Waiters and Waitresses",
      "verdict": "EXPOSED",
      "risk_resistance": 49,
      "median_wage": 35230,
      "overlap": 70,
      "skills_to_close": [
        "Quality Control Analysis",
        "Monitoring",
        "Operations Monitoring",
        "Troubleshooting"
      ]
    }
  ],
  "license": "https://cookedindex.com/terms"
}