Automation removed the shift-change handoff, not network control.
AI value · Driver-transfer road exposure
Company-reported
Road travel eliminated
Avoided exposure, not injuries prevented.
Before
Drives the heavy-haul train between mine and port until the 12-hour shift limit. → Travels to the train for mid-journey driver changes, requiring about 1.5 million road km annually.
After
Controls speed, braking, and responses to restrictions and alarms across the private mainline. → Monitor missions and manage stops, alarms, and network exceptions.
Human boundary
Automation drives routine missions; remote control owns network decisions.
Why it matters
Mainline driving can be automated while exceptions move to remote operations.
This case is company-reported. Use it for the operating-model shift; do not treat the numbers as independently measured.
How the work changed
Before
How the work ran before the change.
Step 1 of 2
Train driver
Drives the heavy-haul train between mine and port until the 12-hour shift limit.
ControlRail rules and onboard safety systems.
Step 2 of 2
Relief crew and road vehicle
Travels to the train for mid-journey driver changes, requiring about 1.5 million road km annually.
ControlPublished sources quantify road exposure but do not provide baseline injury count.
What changed
Mainline driving can be automated while exceptions move to remote operations.
Decision rightAI handles the default; humans own exceptions
After
How the same work runs now.
Step 1 of 2
AutoHaul onboard control
Controls speed, braking, and responses to restrictions and alarms across the private mainline.
ControlAutomatic train protection, onboard cameras, and upgraded crossings.
Step 2 of 2
Remote rail operators
Monitor missions and manage stops, alarms, and network exceptions.
ControlHuman network authority; exception latency and intervention thresholds are not published.
Process model built from the published workflow evidence for Rio Tinto Iron Ore. Every step, actor, and control appears in full below.Every step, actor, and control
Exception path
Stopped missions and alarms follow rail-safety control procedures.
Work removed
Mid-journey driver changes
1.5 million km annual relief-driver road travel
Decision authority
Automation drives routine missions; remote control owns network decisions.
Before
01
Train driver
Drives the heavy-haul train between mine and port until the 12-hour shift limit.
Control: Rail rules and onboard safety systems.
02
Relief crew and road vehicle
Travels to the train for mid-journey driver changes, requiring about 1.5 million road km annually.
Control: Published sources quantify road exposure but do not provide baseline injury count.
After
01
AutoHaul onboard control
Controls speed, braking, and responses to restrictions and alarms across the private mainline.
Control: Automatic train protection, onboard cameras, and upgraded crossings.
02
Remote rail operators
Monitor missions and manage stops, alarms, and network exceptions.
Control: Human network authority; exception latency and intervention thresholds are not published.
Work that left the path
Mid-journey driver changes
1.5 million km annual relief-driver road travel
Human role before
Train drivers operated mine-to-port services until shift limits, while relief crews travelled by road for mid-journey handoffs under rail-safety controls.
Human role after
Remote operators monitor missions and manage exceptions.
AI roleControls train speed, braking and alarm responses on the private network.
Outcomes
Driver-transfer road exposure
Company-reported
About 1.5 million km/year→Road travel eliminated
Operational disclosure · 200 locomotives; 1,700+ km track
Avoided exposure, not injuries prevented.
What leaders can reuse
Anti-pattern
Do not infer injuries prevented.
Questions
01Where is the operating threshold set and who can override it?
02What measured result would trigger rollback or retraining?
03Which residual decisions must remain human-owned?
Portability conditions
Private network
Remote intervention
Rail assurance
Reputation risk
low
Evidence and authority
What the public record supports.
Current · updated
2 primary; publication outcomes are reported.
Bundle 1.0.0 · reviewed 2026-08-23 · stable ID 977fe9cae6f167e2
