COOKED
The modal customer service rep answers inbound calls, chats, and emails against a scripted knowledge base, looks up order and account status, processes returns and credits, and logs the interaction in a CRM — the exact loop that LLM agents with tool access now handle at acceptable quality and a fraction of the cost. A minority of the title works in-person (bank branches, utility counters, retail service desks) or handles escalations where an angry customer needs a human who can bend policy and own the outcome; that tier is real but small relative to 2.6 million jobs. There is no license, no signature requirement, and no personal liability to slow deployment.
Part 2020 shock, part continued decline in the years since.
Median pay $34,710 → $44,770 +3.2% in real terms
This line is counted by the Bureau of Labor Statistics — the one figure on this page that isn't a judgement of ours. Headcount moves on demand, offshoring, demographics and the business cycle, and automation is one term among several, often not the loudest.
So a falling line is not evidence that AI did it, and a rising one is not evidence that it won't. Both happen in this register: some occupations resist automation and shrink anyway, others are highly automatable and keep growing. The marked year is 2020.
BLS projection, 2024–2034
-5.5% 2,814,000 → 2,660,300 on the projections basis
Exposed, and shrinking
Both signals point the same way: the tasks are largely automatable and the BLS projects -5.5% by 2034. This is the case where the score and the forecast agree, and it is the one worth taking seriously.
Different clocks. The score is what current AI could do to this work today. The projection is how many of these jobs will exist in 2034. Everything between the two — how fast employers actually adopt, whether demand grows in the meantime — is why they can point opposite ways without either being wrong.
~341,700 openings a year on average, including replacing people who leave.
Bill AdjusterHub AssociateReturns ClerkTrouble ClerkService WriterWarranty ClerkService AdvisorAccount AdjusterAdjustment ClerkComplaints ClerkCustomer AdvocateComplaints AdjusterGuest Service AgentClerical AdjudicatorLost and Found ClerkPhone RepresentativeVerifying SpecialistCompensation AdjusterAccount RepresentativeCustomer Service AgentCustomer Service ClerkMembership CoordinatorService RepresentativeWarranty Administrator
Customer Success ManagerTechnical Account Manager
Holding it up: trust premium . Weakest point: liability shield .
Core tasks are already automatable Intent classification, order lookup, refund issuance under a dollar threshold, and after-call CRM notes are the four things that fill the shift, and all four are already shipping in production deflection stacks — the 4 rather than a 10 reflects that even the harder variants (multi-system account reconciliation, warranty eligibility) are lookups against documented rules, not novel work.
Fully desk- and screen-based Headset, two monitors, and a seated queue is the job for most of the 2.6 million; the 4 rather than a 0 is the branch teller and utility counter minority who hand over paperwork, verify ID in person, and swap out equipment at a service desk.
No licence, no signature requirement Nothing you say on a call requires a credential — the exception is the insurance and securities CSR who needs a state producer license or Series 6/7 to quote or bind, and at roughly a slice of the title that pulls the score to 1 instead of 0.
Executes defined procedures on defined inputs Your discretion is bounded by a refund ceiling, a retention offer matrix, and an escalation trigger written by someone else; the 4 covers the real judgment in reading whether a caller is about to churn or file a CFPB complaint and deciding to escalate rather than close.
The verdict above describes this occupation as a whole. Almost nobody does the typical version of a job — tick what's actually in your week and see how your own mix sits.
Your task mix speaks to task resistance (4/20 here) — how much of the day's work current AI already does. That is the dimension the boxes above are about.
It cannot move the other three. Liability shield (1/20) is whether the law requires a licensed human to sign. Trust premium (7/20) is whether buyers specifically pay for a person. Judgment and accountability (4/20) is whether the role exists to own consequential calls. Those are facts about the occupation's standing, not about which tasks are in your week — a paralegal who does only trial exhibits still holds no licence. Together they are 12 of this occupation's 20 points (60%).
Embodiment (4/20) is also a property of the work rather than the worker, but we don't tag individual tasks as physical or not, so the picker can't tell you anything about it. That's a limit of this tool, not a claim.
Did we get the list right? Tell us what's missing — the tasks are written from the outside, and you're reading this from the inside.
The moves above are yours to make. This is the other half: what would have to change in the world for the occupation itself to score higher. None of it is in any one person's gift, but it is where the floor actually comes from. Scores here are not a one-way ratchet. Only two of the five dimensions — task resistance and embodiment — track what machines can do. The other three track law, what buyers will pay for, and who is answerable, and those move in both directions, often in response to the same pressure AI creates. If every lever below landed, this occupation would score around 39/100 — EXPOSED.
Tier collapse: once bots absorb order-status, returns, and password resets, the surviving headcount is escalation/retention work — de-escalating abuse, reconstructing what the bot got wrong, exception handling outside policy. This raises the average difficulty of the remaining job even as headcount falls sharply. Already visible at Klarna, which cut agents then rehired for a 'human always available' escalation tier.
Enforcement extension of existing sectoral rules to AI channels: e.g. CFPB/state UDAP actions or FCC TCPA rulings holding that debt-collection, insurance-claim, or utility-shutoff conversations require an identified human agent of record; several state insurance departments already require licensed producers for anything touching coverage advice. Would force a named human on regulated-vertical queues.
Formal authority expansion: giving retained reps binding goodwill-credit and policy-exception limits (a real dollar authority documented in the CRM) rather than script adherence, because the bot cannot be given spend authority insurers or auditors will accept. Makes the human the accountable decision point on refunds, fraud holds, and retention offers.
AI-disclosure and human-escalation statutes: Utah's AI Policy Act (2024) already requires disclosure when a consumer asks if they're talking to AI; California SB 1018-style bills and the EU AI Act Art. 50 push the same. A mandatory 'right to a human on request' clause — proposed in several state bills and in Colorado's SB 24-205 framework — would statutorily preserve a human queue.
Union contract language: CWA has bargained for staffing floors and AI-consultation clauses at AT&T and in the 2023-25 cycle; a ratified minimum-human-agent headcount or no-displacement-by-automation clause at a major telco/airline would convert trust into a contractual floor. Also marketable 'talk to a real person' positioning as a differentiator in banking and airlines.
The limit. Even with every lever, this stays a low-score occupation and the headcount collapse is largely independent of the score: the levers protect the character of a much smaller surviving tier, not the 2.6 million. No licensure exists to build a shield on, and there is no plausible route to broad consumer willingness to pay for a human on routine transactions.
| New York-Newark-Jersey City, NY-NJ | 125,570 | $49,590 +11% |
| Dallas-Fort Worth-Arlington, TX | 96,930 | $44,990 +0% |
| Chicago-Naperville-Elgin, IL-IN | 75,240 | $47,100 +5% |
| Phoenix-Mesa-Chandler, AZ | 68,930 | $46,410 +4% |
| Los Angeles-Long Beach-Anaheim, CA | 66,590 | $48,840 +9% |
| Houston-Pasadena-The Woodlands, TX | 65,510 | $40,380 -10% |
| Miami-Fort Lauderdale-West Palm Beach, FL | 63,440 | $39,560 -12% |
| Atlanta-Sandy Springs-Roswell, GA | 59,990 | $42,710 -5% |
| San Jose-Sunnyvale-Santa Clara, CA | 8,860 | $60,480 +35% |
| San Francisco-Oakland-Fremont, CA | 19,930 | $57,830 +29% |
| Seattle-Tacoma-Bellevue, WA | 25,840 | $53,550 +20% |
Oracle · Uber · Walmart · McDonald's · Block · IKEA · Home Depot · Monday.com · DeepL · Salesforce · Commonwealth Bank of Australia · Commonwealth Bank of Australia; Microsoft; Uber · Centrica · Mews · Yorkshire Building Society · KPN · DocMorris · CVS Health · Tower Insurance · Ibex 35 banks · The Home Depot · Internal Revenue Service · Airbnb · Travelers · Etiqa Insurance · Klarna · Chubb · Allianz · IAG · HSBC · 1&1 · Coinbase · Verizon · Heathrow Airport · Commerzbank · Cisco; Block · Expedia · Ergo · Suncorp
German telecom provider 1&1 reportedly plans to cut 60 positions at its Zweibrücken site while shifting to AI automation.
Quartz reports Block posted a Q2 2026 earnings beat following layoffs attributed to AI adoption.
BeInCrypto reports that Block's earnings follow layoffs the company attributed to AI adoption.
Yahoo Finance reports that Block's earnings follow layoffs the company attributed to AI adoption.
CX Today reports IKEA has deployed its AI assistant Billie to handle routine customer queries and retrained 8,500 call centre staff into other roles.
HC Amag reports Commonwealth Bank of Australia cut contractor call centre roles following an AI rollout in its customer service operations.
Has AI actually changed your work? One tap, anonymous, and the running tally is public. Nothing else is asked of you.
Rather than check back: get the digest and we'll tell you what changed — or watch a single occupation from its own page.