Construction worker — tasks, one by one
The unit of analysis is the task, not the job title. Each one below carries its direction, whether the judgement rests on evidence or on platform inference, the reasoning, and what it does not establish.
Every task on this page#
Building it where it goes
Still human-led✓ Evidence-backedForming, fixing, laying, finishing — in a space with no two corners alike and a tolerance that has to be absorbed as you go.
This is unstructured manipulation in an environment that changes hourly, which robotics has handled poorly for forty years — and the constraint is not dexterity but the absence of a fixed reference frame. Where construction robots work, the site has been changed to suit them rather than the other way round.
The real mechanism moving this work is not a robot arriving on site, it is the work leaving the site: prefabrication moves the same task into a factory where a fixed reference frame exists, and the job that results is a different job in a different place with different pay. Reading this task as safe because robots are bad on site misses where the change actually comes from.
Setting out and checking
Automating≈ Platform inferenceMarking where things go, checking level and line, and catching the error before it is buried.
Robotic total stations, laser layout and scan-to-model comparison are among the most successfully deployed technologies in construction, and the reason is the shape of the task: position against a model is measurable, and the answer is a number a machine produces better than a person with a tape.
Better measurement finds more errors and does not fix any of them, so the work it creates is rework, which is manual. Where scanning has been introduced the reported effect is usually fewer buried mistakes rather than fewer people, and the cost saved lands with the client rather than with the crew.
Working out what can happen today
Still human-led✓ Evidence-backedThe concrete is late, the scaffold is not struck, another trade is in the way — deciding what the crew does instead, right now.
Scheduling software plans; it does not know that this delivery driver will wait twenty minutes if asked or that the other foreman owes you a favour. The re-plan happens hourly against information that exists only as relationships and is never entered anywhere.
Being unautomatable is not being paid: this judgement is usually performed by a working foreman whose pay reflects the trade rather than the coordination. And the same task is being centralised in large contractors, which moves it to an office without automating it at all.
Not getting hurt
New task≈ Platform inferenceReading a site for what is about to go wrong: the load that is not secure, the edge nobody has guarded, the person working above you.
Site-safety monitoring is a genuine and growing deployment — cameras detecting missing protection, proximity systems on plant — and it targets a real problem, because construction's injury rate is the reason this task is core. What the systems catch is rule violations; what they cannot catch is the situation that is unsafe without breaking any rule.
Safety monitoring changes who is blamed as much as who is safe: a system that records every violation creates a record that can be used to attribute an accident to a worker rather than to a schedule. Where this has been introduced without changing the schedule, the reported effect on injury rates has been much smaller than the reported effect on documentation.