EXPOSED
The daily work — pulling and cleaning data, coding optimization and simulation models in Python/R, running scenario analyses, and writing up findings in decks and memos — sits squarely in the zone current AI handles at usable quality. What persists is upstream: correctly framing a messy business problem as a solvable model, choosing which constraints and objectives actually reflect what the organization wants, and defending a recommendation that reroutes a supply chain or reshapes a staffing plan. There is no license and no signature requirement, so nothing regulatory slows the substitution; the modal analyst's output is an artifact, not a relationship.
Dipped in 2020, then grew past where it started.
Median pay $84,810 → $88,940 -16.1% 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
+21.5% 112,100 → 136,200 on the projections basis
Exposed, but growing
AI can already do a lot of these tasks, and the BLS still expects +21.5% more of these jobs by 2034. Demand for the output is growing faster than the work is being automated away — the mechanism BLS gives for software developers, and the combination people most often misread as an error.
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.
~9,600 openings a year on average, including replacing people who leave.
ResearcherRisk AnalystForms AnalystPolicy AdvisorPolicy OfficerLiaison PlannerMethods AnalystProcess AnalystSystems AnalystBusiness AnalystDecision AnalystMaterial LiaisonProcedure WriterResearch AnalystProcedure AnalystStandards AnalystTechnical AnalystMethods ConsultantMethods SpecialistResearch AssistantResearch AssociateResearch ScientistSystems ConsultantResearch Specialist
Holding it up: judgment & accountability . Weakest point: liability shield .
Mixed — a routine tier and a judgment tier Writing the LP formulation, tuning solver parameters in Gurobi/CPLEX, building discrete-event simulations, and generating sensitivity tables are all things an LLM now does at first-draft quality; the 9 rather than a 5 reflects that the problem-framing step — deciding that a nurse-scheduling complaint is actually a demand-forecasting failure, and that the constraint the client insists on is the one to relax — still requires sitting in rooms with operations staff who cannot state their own objective function.
Fully desk- and screen-based A 3 rather than 0 accounts for the occasional plant walkthrough, warehouse time-and-motion observation, or ride-along to see why the routing model's assumptions don't match how drivers actually load the truck — real, but a few days a quarter against a full-time seat in front of a solver and a BI tool.
No licence, no signature requirement There is no state license, no PE-style stamp, and no statutory sign-off on an optimization recommendation; the 2 rather than 0 is only for the narrow slice working under DoD/FFRDC clearance requirements or SOX-adjacent model-governance regimes, where the barrier is the clearance and the model-validation paperwork, not the analyst's credential.
Meaningful discretion A 12 sits above procedural work because you choose the objective — minimize cost, or minimize worst-case delay, or keep the union contract intact — and that choice is a value judgment nobody hands you in a spec; it stops short of 14+ because the executive who approves the network redesign owns the consequence, and your name appears on the analysis, not the decision.
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 (9/20 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 (2/20) is whether the law requires a licensed human to sign. Trust premium (8/20) is whether buyers specifically pay for a person. Judgment and accountability (12/20) 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 22 of this occupation's 34 points (65%).
Embodiment (3/20) 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.
Aerospace Engineers 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 47/100, still EXPOSED.
Task-mix shift is genuine here: the occupation has a routine tier (data pull, model coding, scenario runs, deck production) and a judgment tier (problem framing, constraint elicitation from stakeholders who disagree, validating that an optimum is not an artifact of a mis-specified objective). If headcount contracts to the framing/validation tier, the residual job's untractable share rises. Watch for job postings shifting from 'build models in Python' to 'translate business problems and audit model outputs'.
Narrow and only in adversarial settings: expert-witness and regulatory-testimony work (damages models, antitrust market simulations, capacity-adequacy filings at FERC/state PUCs) where a tribunal requires a human author who can be cross-examined on methodology. Federal Rule of Evidence 702 and Daubert already force a named human expert. This applies to a small slice of the occupation, not the modal analyst.
Model risk management regimes extending beyond credit/finance to operational optimization. Concretely: if the Fed/OCC SR 11-7 style model validation requirement, or the EU AI Act's high-risk obligations for workforce-management and critical-infrastructure systems, are read to cover staffing-allocation and network-routing optimizers, a named human must document intended use, challenge assumptions, and own the sign-off on model limitations. Analysts in banks and insurers already do this as 'model validation' roles.
No license exists and none is being proposed, so the only realistic route is contractual/internal rather than statutory: professional-body credentialing (INFORMS' Certified Analytics Professional) becoming a named requirement in federal contracting or in E&O insurance conditions for consultancies, or a named 'model owner' attestation required under an internal MRM policy. This is a weak shield — it does not create personal legal liability — so the gain is small.
The limit. Realistic ceiling is roughly the mid-40s. The core constraint is structural: there is no licensure pathway in motion, the output is a deliverable rather than a relationship, and the buyer is an internal executive who has no reason to pay a premium for human authorship. The bulk of any upside is task-mix contraction into the framing tier — which raises the score per surviving worker while reducing how many workers survive.
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 8,080 | $127,600 +43% |
| New York-Newark-Jersey City, NY-NJ | 6,890 | $103,110 +16% |
| Dallas-Fort Worth-Arlington, TX | 4,270 | $106,380 +20% |
| Chicago-Naperville-Elgin, IL-IN | 3,680 | $86,500 -3% |
| Los Angeles-Long Beach-Anaheim, CA | 3,480 | $98,270 +10% |
| Boston-Cambridge-Newton, MA-NH | 3,420 | $100,700 +13% |
| Atlanta-Sandy Springs-Roswell, GA | 3,250 | $77,510 -13% |
| Houston-Pasadena-The Woodlands, TX | 2,500 | $86,380 -3% |
| Colorado Springs, CO | 170 | $140,590 +58% |
| Lexington Park, MD | 390 | $135,300 +52% |
| Huntsville, AL | 350 | $132,000 +48% |
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 34. 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.