Safety NetDispatch
Safety Net Dispatch tool · Construction Site of the Future

AI Is More Likely to Rewrite Construction Roles Than Replace Them

The AI Exposure Matrix is a judgment-based model of relative exposure, not a job-loss forecast. It assesses how much each role's current task mix is reachable by present and emerging AI capabilities, and uses broad directional horizons rather than precise timelines.

Role type
Construction roles
Exposure band Higher Moderate Lower
MORE / SOONER CHANGE LESS / LATER CHANGE Cashier </> Software Eng. Lawyer Radiologist BIM Coord. Project Mgr. Estimator Equip. Operator Safety Director Superintendent Foreman Field Laborer
Now Nearer term Medium term Longer term
Directional change horizon →
Compare task mixes, not job-loss odds. Higher, Moderate, and Lower are judgment-based exposure bands. Tap a role to see the work AI can assist with and the decisions people still own.
Construction roles
Other industries (comparison)
Exposure = a relative assessment of how much a role’s current task mix is reachable by present and emerging AI capabilities. Horizons are directional scenarios, not forecasts. Tap any role for details.

,

,
Exposure band ,
Horizon (across) ,
Impact (up) ,

What's changing

,

Where human judgment remains central

,

Possible direction

,

Bottom line ,

Safety Net Dispatch

Why construction is exposed differently

The full write-up behind this matrix is coming to Safety Net Dispatch: practitioner notes on AI, field data, and what actually changes on a jobsite. Subscribe to get it when it ships.

Subscribe to Safety Net Dispatch

Don’t see your role, or disagree with a placement?

Suggest a role →
Methodology: This matrix is a practitioner-built thinking tool. Role placements are synthesized judgments informed by occupational task data, observed AI-use research, labor-market analysis, and construction-technology research. The bands and horizons are not job-loss predictions or independently validated forecasts. The cited sources measure different constructs and were used as inputs, not averaged together as one compatible dataset. Earlier versions showed author-assigned percentages. Those calibration values are retired from the reader-facing model because they implied more precision than the evidence supports. The current bands do not correspond to measured percentages or probabilities.

Band definitions: Higher = a substantial portion of the role already occurs in digital, structured, or repetitive workflows. Moderate = AI can reach meaningful administrative or analytical tasks, while important work remains physical, situational, or judgment-dependent. Lower = the core work remains highly physical, variable, interpersonal, or dependent on real-time field judgment.

How to read the chart: Impact describes the possible scale of task-mix change; it is not the exposure band. Horizons (already happening, nearer term, medium term, longer term) express relative sequencing, not dates. Coordinates and spacing are illustrative. The comparison roles are selected examples, not a representative sample of other industries. This model includes AI, robotics, and conventional automation; not every example is generative AI. Role-panel directions are scenarios, not employment forecasts.

Sources: Anthropic Economic Index, February 2025 (observed Claude use, not economy-wide adoption); Goldman Sachs labor-exposure research; U.S. Department of Labor O*NET; World Economic Forum Future of Jobs Report 2025; McKinsey construction AI and productivity research; ILO, Generative AI and Jobs: A Refined Global Index of Occupational Exposure (2025); OSHA construction responsibilities and authorized and competent-person definitions.