COOKED
The job is compiling data from surveys and records, cleaning and coding it, running prescribed statistical computations, and building tables and charts for someone else to interpret — which is precisely the work current AI plus modern scripting already does at usable quality. There is no license, no signature, and no client relationship; the statistician or researcher above you owns the interpretation and the accountability. The surviving fragment is data acquisition and quality judgment in messy institutional settings — knowing which source is trustworthy, why a field is broken, how a survey instrument distorts responses.
This fall is concentrated in 2020 and has not recovered since.
Median pay $49,870 → $50,330 -19.3% 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
-2.5%
Percentage only. The projection counts a different population from the 4,710 above — it includes self-employed workers, which for this occupation is most of them, so the two headcounts are not comparable.
Exposed, and shrinking
Both signals point the same way: the tasks are largely automatable and the BLS projects -2.5% 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.
~800 openings a year on average, including replacing people who leave.
BookmanCompilerChart ClerkFiscal ClerkReport ClerkActuary ClerkChart ChangerData CompilerMileage ClerkProrate ClerkWheelage ClerkData TechnicianProration ClerkTariff CompilerAnalytical ClerkChart CalculatorTabulating ClerkActuarial AnalystCompilation ClerkStatistical ClerkTechnical AnalystData Editing ClerkResearch AssistantResearch Associate
Holding it up: task resistance . Weakest point: liability shield .
Core tasks are already automatable Recoding open-ended survey responses into category codes, deduplicating respondent records, running the prescribed regression or weighting procedure in SAS/R/SPSS, and formatting the output into tables are each tasks an LLM plus a scripting pass now completes end-to-end; the 4 rather than 0 reflects the residual chasing of missing questionnaires, reconciling a field office's spreadsheet against the master file, and noticing that a variable's coding changed between collection waves.
Fully desk- and screen-based The only physical elements are retrieving paper questionnaires or archived record files and occasionally operating scanning or data-entry equipment; everything else happens at one workstation, which is why this sits at 2 rather than 0.
No licence, no signature requirement No state licence, no certification exam, no signature block — an ASA accreditation exists for statisticians and is neither required nor typically held at assistant level, and any error you introduce into a dataset is caught and owned by the statistician who signs the report.
Executes defined procedures on defined inputs Codebooks, survey specifications, and imputation rules set in advance dictate almost every call you make; the discretion at 4 is deciding whether a suspicious outlier is a real value or a keying error, and whether a response rate is low enough to flag upward — decisions someone else then ratifies.
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 (3/20) 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 8 of this occupation's 14 points (57%).
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.
Financial Specialists, All Other EXPOSED
Data Scientists EXPOSED
Economists EXPOSED
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 30/100, still COOKED.
Task-mix shift: if the scripted tier (recode, merge, tabulate, chart) is fully absorbed, what remains is data provenance and defect diagnosis — reconciling conflicting administrative sources, spotting instrument-induced response distortion, deciding whether a broken field is imputable. This tier genuinely exists in survey shops (Census, NORC, RTI) and in clinical data management, and it is what audit trails actually ask about. Watch for job postings retitled 'data quality analyst' or 'data manager' with the same pay band.
Extension of FDA 21 CFR Part 11 electronic-records practice — where a named individual's e-signature attaches to each data query resolution and database lock — into AI-produced derivations. FDA's January 2025 draft guidance on AI in regulatory decision-making asks sponsors to document human credibility assessment of model outputs; if that hardens into a requirement that a named data manager attests to AI-generated dataset transformations, the clinical-trial slice of this SOC acquires a real signature.
Model-risk validation regimes (Fed/OCC SR 11-7, and its spread to insurance via NAIC's 2023 AI model bulletin adopted in 20+ states) requiring a named human to document input-data lineage and quality for any model used in pricing or underwriting. That documentation work is exactly this occupation's residual, and SR 11-7 already demands independent, attributable review.
Formal ownership of the disclosure-avoidance and data-suppression call. Census's differential-privacy rollout and state health-department small-cell suppression rules put a consequential, ambiguous decision (release or suppress) on whoever prepares the table. If agencies name that person in the release documentation rather than treating it as clerical, the role owns a real call.
Sworn-agent status under Title 13 (Census) and similar confidentiality oaths carry criminal penalty for improper disclosure and already attach personally. Broader use of sworn special-agent designations for contractor data staff handling linked administrative records would make the disclosure judgment personally consequential.
The limit. Even with every lever, this caps in the high 20s–low 30s. There is no license, no client who chose you, and no physical component — and the liability levers attach to specific sectors (clinical trials, regulated finance, federal statistics) that hold well under half of these 4,710 workers. For the rest, compiling and charting inside a general office, no plausible route exists; trust_premium in particular has none, because the buyer of a cleaned dataset has never known or cared who cleaned it.
| Louisville/Jefferson County, KY-IN | 320 | $45,660 -9% |
| San Juan-Bayamon-Caguas, PR | 230 | $27,490 -45% |
| Seattle-Tacoma-Bellevue, WA | 170 | $63,660 +26% |
| Detroit-Warren-Dearborn, MI | 150 | $49,660 -1% |
| Boston-Cambridge-Newton, MA-NH | 140 | $70,570 +40% |
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 90 | $59,690 +19% |
| Omaha, NE-IA | 80 | $60,380 +20% |
| Bridgeport-Stamford-Danbury, CT | 70 | $80,360 +60% |
| Bridgeport-Stamford-Danbury, CT | 70 | $80,360 +60% |
| Hartford-West Hartford-East Hartford, CT | 40 | $73,660 +46% |
| Boston-Cambridge-Newton, MA-NH | 140 | $70,570 +40% |
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 14. 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.