SAFE
The paperwork half of the job — incident narratives, warrant affidavits, records checks, phone-dump and financial-record triage, link analysis — is exactly what AI does well, and agencies are already buying tools for it. But the core of the job is sworn, physical, and adversarial: knocking on doors, interviewing hostile witnesses, working informants, processing scenes, executing arrests, and testifying under cross-examination where a human's credibility and personal accountability are the evidence. Judgment about probable cause and charging recommendations sits with a badge, not a model.
Headcount grew steadily across the period.
Median pay $83,170 → $93,790 -9.8% 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
-0.7% 117,900 → 117,100 on the projections basis
Hard to automate, but shrinking anyway
The work resists current AI, yet the BLS projects -0.7% by 2034. Whatever is shrinking this occupation, the evidence does not point to automation — demand, demographics, offshoring and industry decline all shrink jobs that no machine could do.
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.
~7,800 openings a year on average, including replacing people who leave.
SpyAnalystDetectiveAir MarshalCriminalistInvestigatorCrime AnalystSpecial AgentState TrooperCity DetectivePolice OfficerThreat AnalystUndercover CopVice DetectiveVictim LiaisonDefense AnalystNarcotics AgentCriminal AnalystPolice DetectivePolice InspectorSecurity AnalystAccident ExaminerCriminal ProfilerEnforcement Agent
Holding it up: judgment & accountability . Weakest point: trust premium .
Mixed — a routine tier and a judgment tier Cellebrite extraction reports, NIBRS-coded incident narratives, subpoenaed bank-record spreadsheets and CDR link charts are already machine-triaged, but nothing yet gets a reluctant witness to change their story on a third doorstep visit or reads a suspect's tell across a table during a Reid-style interview — that split, roughly half desk and half street, is what puts it at 12 rather than 16.
Hands-on in uncontrolled environments Detectives clear rooms on warrant service, kneel in the blood spatter to photograph and swab before the scene is released, drive surveillance in traffic, and physically take custody of resisting arrestees in whatever weather and neighborhood the call happens in; it stops short of 18 only because a meaningful share of the shift is spent at a desk building the case file.
Licensed human required and personally liable Peace-officer certification is a statutory prerequisite (POST or state equivalent), your sworn name goes on the warrant affidavit under penalty of perjury, and Section 1983 exposure follows you personally when qualified immunity fails — but the badge is a license to act rather than a professional monopoly on the analytical work, which caps this below the 17-plus range of physicians and attorneys.
Exists to be accountable for ambiguous calls You decide on incomplete facts whether probable cause exists, which charge to recommend to the DA, whether to Mirandize now or keep it consensual, and whether to bring in an unstable suspect tonight or hold surveillance — calls that get relitigated at a suppression hearing years later with your name on them and the case dying if you got it wrong.
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 (12/20 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 (13/20) is whether the law requires a licensed human to sign. Trust premium (12/20) is whether buyers specifically pay for a person. Judgment and accountability (17/20) 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 42 of this occupation's 69 points (61%).
Embodiment (15/20) 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 81/100, still SAFE.
Courts and legislatures formalizing that a sworn officer must personally attest, under penalty of perjury, to the factual basis of any AI-drafted warrant affidavit and disclose the tool's involvement — already in motion via King County (WA) prosecutors' 2024 ban on AI-written police reports, Utah SB 180 (2025) requiring disclosure of generative AI in police reports, and California SB 524 (signed 2025) mandating AI-drafted report disclosure plus officer certification of accuracy. If a Brady/Giglio-style disclosure duty attaches to tool use, the sworn signature becomes non-delegable and audit-tracked.
Suppression rulings or consent decrees that treat 'the model flagged it' as failing the independent-judgment requirement for probable cause (the Franks v. Delaware / Aguilar-Spinelli lineage applied to algorithmic tips), forcing a named detective to build and own independent corroboration before charging — as already litigated around facial-recognition-led arrests in Detroit's 2024 ACLU settlement, which requires independent evidence beyond an algorithmic match.
Task-mix shift: this job genuinely has two tiers. If report narratives, phone-dump triage, records checks and link analysis are absorbed by Axon Draft One-type tools and Cellebrite/Magnet analytics, the residual day is interrogation, informant handling, hostile-witness work, scene judgment and cross-examination prep — the tier no current model performs at usable quality. Watch caseload-per-detective ratios rising without narrative-writing hours.
Shift of unit composition toward warrant service, undercover and field-contact work as digital-forensics analysis is centralized into civilian analyst units — meaning the remaining sworn investigator role concentrates the physically unpredictable share. Watch agency reorganizations that move digital forensics to non-sworn analysts.
Limited route: private-sector demand for licensed human investigators where a testifying witness is required — insurance SIU and corporate internal investigations contracting ex-detectives specifically because a deposable human is needed. This is a real but small market and does not move public-agency staffing.
The limit. Headroom is mostly in liability_shield; trust_premium is near its realistic ceiling since the buyer is a government budget, not a client choosing a human, and budget pressure cuts the other way — agencies adopt these tools to close vacancy gaps, so higher per-detective resistance can coexist with fewer detective positions.
| New York-Newark-Jersey City, NY-NJ | 7,580 | $106,440 +13% |
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 4,830 | $156,750 +67% |
| Los Angeles-Long Beach-Anaheim, CA | 3,240 | $127,890 +36% |
| Atlanta-Sandy Springs-Roswell, GA | 3,070 | $69,900 -25% |
| San Diego-Chula Vista-Carlsbad, CA | 2,990 | $124,840 +33% |
| Chicago-Naperville-Elgin, IL-IN | 2,290 | $102,260 +9% |
| Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 2,160 | $97,580 +4% |
| Miami-Fort Lauderdale-West Palm Beach, FL | 2,110 | $99,080 +6% |
| San Jose-Sunnyvale-Santa Clara, CA | 280 | $158,640 +69% |
| Brunswick-St. Simons, GA | 240 | $158,060 +69% |
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 4,830 | $156,750 +67% |
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 69. 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.