EXPOSED
Laying out, aligning, clamping, and tack-welding heavy plate and beams is physical work in a shop where parts warp, fit-up gaps vary, and nothing lands exactly where the drawing says — language AI does nothing about that. The real pressure is not chatbots but capital: CNC plasma tables, automated beam lines, and robotic weld cells already absorb the repeatable cutting, drilling, and long-run welding, which is why headcount concentrates in custom and low-volume work. Blueprint interpretation and cut-list/nesting math are the parts AI and CAM software eat first; there is no license protecting the trade, only AWS certifications and employer qualification tests.
Part 2020 shock, part continued decline in the years since.
Median pay $40,390 → $51,330 +1.7% 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
-16.3% 53,800 → 45,000 on the projections basis
Shrinking, but not obviously because of AI
The BLS projects -16.3% by 2034, but at 47/100 this work is only moderately exposed — not the profile of a job current AI can simply do. Occupations shrink for many reasons, and the score does not point at automation as this one's cause.
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.
~4,100 openings a year on average, including replacing people who leave.
FitterFabricatorLayout ManHand FormerShip FitterSteelworkerWeld FitterMetal FramerSteel FitterMetal Box MakerMetal FabricatorMill Beam FitterMotorcycle MakerSteel FabricatorStructural FitterMachine Cage MakerStructural PlannerMachine Shop FitterStructures MechanicComponent FabricatorMotorcycle AssemblerSheet Metal MechanicIndustrial FabricatorMotorcycle Fabricator
Holding it up: embodiment . Weakest point: trust premium .
Tasks largely resist digitisation Fitting a 40-foot girder means shimming, wedging, and re-clamping until the flange sits within tolerance on a plate that has already pulled from heat — that hands-on reconciliation of drawing to warped steel keeps this at 14, while nesting, cut lists, and beam-line drilling programs are genuinely already machine work, which is what stops it from scoring higher.
Hands-on in uncontrolled environments Overhead cranes, chain falls, grinders, and hot tack passes on stock that shifts as it cools puts this at 17: the shop floor is enclosed and lit, unlike ironworkers erecting steel in weather, but nothing about positioning a 2-ton assembly on horses happens from a console.
No licence, no signature requirement AWS D1.1 procedure qualification and employer weld tests gate who touches the joint, but the certification travels with the weld procedure and the PE stamp on the structure belongs to the engineer of record — a fitter faces no personal licensure exposure, hence 4 rather than 10.
Meaningful discretion Deciding tack sequence and weld order to control distortion, or calling a fit-up bad and sending a member back to the cutting table, is real discretion that costs money if wrong — but tolerances, WPS parameters, and inspection hold points are written down and a QC inspector signs off, which caps this at 9.
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 (14/20 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 (4/20) 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 (9/20) 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 16 of this occupation's 47 points (34%).
Embodiment (17/20) 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.
Glaziers 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 58/100, still EXPOSED.
Two-tier structure is genuine: nesting, cut lists, beam-line drilling and long straight welds are already going to CAM and robotic cells, leaving warp correction, heat-straightening, in-situ fit-up of out-of-tolerance parts, and repair/retrofit work. As shops finish automating the repeatable tier, remaining measured task mix shifts toward the judgment/hand-fit tier without any law changing.
AISC certification programs and state/municipal building codes already require special inspection of fabricated structural connections; if AWS D1.1/AISC 360 revisions or local amendments (as some jurisdictions do for seismic zones) required a named AWS-certified welder/fitter of record with personally traceable stamp on fracture-critical or seismic moment connections — akin to the AWS CWI signature already required on inspection reports — the shield rises. FHWA fracture-critical bridge member fabrication rules already come close to this.
If prefab/modular construction growth pushes more field tolerance decisions upstream into the shop — fabricator deciding whether an out-of-tolerance member is shimmed, reworked, or rejected before it ships, with rework cost and schedule liability landing on the fab shop under design-assist or delegated-connection-design contracts (already common in AISC delegated connection design practice) — the fitter's calls become consequential and documented.
The limit. Trust premium has no plausible route: structural steel buyers are GCs and EPCs buying to spec and price, and no end customer pays for hand-fit steel as such. The binding threat here is capital equipment, not language models, so institutional levers move the score more than any AI plateau does.
| Chicago-Naperville-Elgin, IL-IN | 2,290 | $68,280 +33% |
| New York-Newark-Jersey City, NY-NJ | 2,240 | $56,870 +11% |
| Houston-Pasadena-The Woodlands, TX | 1,580 | $50,520 -2% |
| Dallas-Fort Worth-Arlington, TX | 1,170 | $49,280 -4% |
| Los Angeles-Long Beach-Anaheim, CA | 920 | $56,990 +11% |
| Phoenix-Mesa-Chandler, AZ | 870 | $48,150 -6% |
| Mobile, AL | 780 | $66,820 +30% |
| Baton Rouge, LA | 770 | $58,410 +14% |
| Bremerton-Silverdale-Port Orchard, WA | 30 | $75,570 +47% |
| San Diego-Chula Vista-Carlsbad, CA | 690 | $68,350 +33% |
| Chicago-Naperville-Elgin, IL-IN | 2,290 | $68,280 +33% |
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 47. 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.