COOKED
The core job — keying values from forms, scans, invoices, and handwritten records into databases and spreadsheets — is exactly what OCR, intelligent document processing, and LLM extraction already do at production quality and a fraction of the cost. What remains is exception handling: the 3-8% of documents the pipeline flags as low-confidence, plus reconciling mismatches against source systems, which requires knowing the business rules rather than typing fast. Employment has been falling for two decades and the automation curve has steepened, not flattened.
Part 2020 shock, part continued decline in the years since.
Median pay $33,490 → $41,340 -1.2% 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
-25.9% 141,600 → 104,900 on the projections basis
Exposed, and shrinking
Both signals point the same way: the tasks are largely automatable and the BLS projects -25.9% 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.
~9,500 openings a year on average, including replacing people who leave.
EncoderData KeyerKeypuncherTeletypistVaritypistData TypistTeleprinterCard PuncherRecord ClerkComputer AideBraille TypistComputer ClerkData CollectorData ProcessorEncoding ClerkTeletype ClerkCheck ProcessorBraille OperatorData Entry ClerkData Input ClerkData TranscriberSimplex OperatorKeypunch OperatorPerforator Typist
Holding it up: judgment & accountability . Weakest point: liability shield .
Core tasks are already automatable Keystroke-level transcription from scanned invoices, remittance advices, and survey forms is measured in keys-per-hour and verified by double-entry — a task with a machine-readable input, a machine-checkable output, and no step that requires leaving the screen, which is why it sits at 2 rather than mid-band alongside clerks who also chase people by phone.
Fully desk- and screen-based The only physical elements are feeding a flatbed or high-speed scanner and handling paper batches at a desk, and even that disappears at any employer that has moved to emailed PDFs or portal uploads, so this is a 2 rather than a 0 only because some shops still put stacks of paper in your hands.
No licence, no signature requirement There is no credential of any kind — no state licence, no board, no registry — and an error in a keyed field is corrected by a supervisor or caught in QA sampling, so nothing about the job legally requires that a specific named human did the keying.
Executes defined procedures on defined inputs The 6 is earned on the exception queue — deciding whether an illegible date is 2019 or 2014, which of two vendor spellings matches the master record, whether a mismatched total means transposition or a genuinely wrong source document — but these are calls made against documented business rules and escalated when unclear, not owned end to end, which keeps it at the ceiling of the procedural band rather than into real discretion.
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 (2/20 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 (0/20) is whether the law requires a licensed human to sign. Trust premium (2/20) is whether buyers specifically pay for a person. Judgment and accountability (6/20) 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 8 of this occupation's 12 points (67%).
Embodiment (2/20) 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 26/100, still COOKED.
Task-mix shift is genuinely bilevel here: if bulk keying is fully absorbed by IDP pipelines, the surviving headcount is exception-queue work — adjudicating low-confidence extractions, resolving source-system mismatches, and correcting handwritten/damaged records that OCR flags. That residual work is harder per-unit and requires business-rule knowledge. This raises the score of the remaining job while shrinking the job count; it is not a jobs reprieve.
Sector-specific rules that require a named human to verify machine-extracted values before they enter a system of record: e.g. FDA 21 CFR Part 11 electronic-records requirements in clinical trial data entry (double-data-entry and human verification of source documents), and CMS/state Medicaid rules on encounter-data attestation. If such attestation requirements are extended explicitly to AI-extracted fields rather than to keyed fields generally, a human verifier becomes structurally required in regulated domains (clinical research coordination, court e-filing dockets, land/deed recording).
Exception adjudicators becoming the documented human-in-the-loop for automated decision systems — e.g. under Colorado SB 24-205 or the EU AI Act's Article 14 human-oversight duty, where a person who overrides or confirms a low-confidence extraction that feeds an eligibility or benefits determination is the recorded decision-maker. This turns keying into an auditable override role.
Physical handling of fragile, classified, or non-digitizable source material — county deed vaults, sealed court records, archival paper that cannot leave a room or be scanned by vendor equipment. This is a small niche and applies to a minority of positions.
The limit. Even with every lever, this occupation stays low. The liability and oversight routes preserve a thin layer of verifier roles inside regulated verticals, not the 127,000-person occupation; historical employment decline continues regardless. There is no realistic trust-premium route — no buyer pays extra for human-typed data, and the value proposition was always speed and cost.
| New York-Newark-Jersey City, NY-NJ | 10,110 | $45,350 +10% |
| Los Angeles-Long Beach-Anaheim, CA | 5,900 | $47,040 +14% |
| Dallas-Fort Worth-Arlington, TX | 5,610 | $43,110 +4% |
| Houston-Pasadena-The Woodlands, TX | 4,430 | $39,520 -4% |
| Chicago-Naperville-Elgin, IL-IN | 4,270 | $44,650 +8% |
| Phoenix-Mesa-Chandler, AZ | 2,650 | $44,160 +7% |
| Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 2,340 | $41,280 +0% |
| Miami-Fort Lauderdale-West Palm Beach, FL | 2,290 | $40,160 -3% |
| San Jose-Sunnyvale-Santa Clara, CA | 720 | $54,510 +32% |
| Albuquerque, NM | 220 | $53,320 +29% |
| Bismarck, ND | 30 | $53,140 +29% |
We have no reported case of a named organisation automating this occupation. Not one deployment, not one announcement.
That is worth saying out loud next to a score of 12. The verdict above is about what the work exposes — what current AI could do to these tasks. It is not a claim that anyone has done it. For this occupation those two things have come apart completely: the capability argument is on this page, and the evidence column is empty.
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.