New York City Fire Department
Mixed evidenceExceptions onlyQueue eliminationContinuous decisioningFire companies can prioritize inspections by predicted risk instead of fixed cycles and local knowledge.
Engineering design · New York City, United States
Collections: Queue eliminated · Embodied work
Executive brief
The measurable first-order result is better inspection yield, not proven fire prevention; leaders should separate queue quality from ultimate safety outcomes.
AI value · Building violations found after FireCast 2.0 deployment
Company-reportedAverage violations increased 19% in the first 30 days and remained 10% higher after 60 days
Contemporary independent report quoting FDNY results; more violations found is a targeting metric, not proof that fires were prevented.
Before
Select and schedule inspectable buildings from policy requirements and local information. → Inspect assigned premises and record hazards.
After
Scores and ranks buildings from fire history, characteristics, violations, complaints, and other risk factors. → Receive prioritized lists, inspect buildings, and issue or follow up violations.
Human boundary
The model determines priority order; uniformed personnel decide findings and enforcement based on physical inspection.
Why it matters
Fire companies can prioritize inspections by predicted risk instead of fixed cycles and local knowledge.
Before
How the work ran before the change.
Fire officers
Select and schedule inspectable buildings from policy requirements and local information.
ControlMandatory cycles and professional judgment
Firefighters and inspectors
Inspect assigned premises and record hazards.
ControlFire code
What changed
Fire companies can prioritize inspections by predicted risk instead of fixed cycles and local knowledge.
Decision rightAI handles the default; humans own exceptions
After
How the same work runs now.
FireCast
Scores and ranks buildings from fire history, characteristics, violations, complaints, and other risk factors.
ControlPredictive risk ranking
Fire officers and field units
Receive prioritized lists, inspect buildings, and issue or follow up violations.
ControlHumans conduct inspections and enforcement
Exception path
Mandatory inspections, public complaints, and officer judgment can trigger work outside the risk-ranked list.
Work removed
Decision authority
The model determines priority order; uniformed personnel decide findings and enforcement based on physical inspection.
Before
Fire officers
Select and schedule inspectable buildings from policy requirements and local information.
Control: Mandatory cycles and professional judgment
Firefighters and inspectors
Inspect assigned premises and record hazards.
Control: Fire code
After
FireCast
Scores and ranks buildings from fire history, characteristics, violations, complaints, and other risk factors.
Control: Predictive risk ranking
Fire officers and field units
Receive prioritized lists, inspect buildings, and issue or follow up violations.
Control: Humans conduct inspections and enforcement
Work that left the path
Human role before
Fire officers assembled inspection priorities without a citywide predictive ranking.
Human role after
Officers allocate field capacity from a risk-ranked list and retain inspection and enforcement judgment.
AI rolePredictive analytics that assigns building fire-risk scores and prioritizes inspection lists.
Violation discovery rate before FireCast 2.0→Average violations increased 19% in the first 30 days and remained 10% higher after 60 days
First 60 days after FireCast 2.0 deployment · New York City risk-based building inspections
Contemporary independent report quoting FDNY results; more violations found is a targeting metric, not proof that fires were prevented.
Anti-pattern
Using risk rank as evidence of a code violation or claiming prevented fires from increased violation discovery.
Questions
Portability conditions
Reputation risk
medium
Evidence and authority
Current · updated
2 independent, 1 primary; publication outcomes are verified and reported.
Bundle 1.0.0 · reviewed 2026-08-23 · stable ID 4ab7ea0afa00d969
Pacific Gas and Electric Company · Exceptions only
That crews should be dispatched mainly after a fault, outage, or visible field report.
Waymo · Exceptions only
Within a defined domain, automated driving can control the vehicle while remote assistance handles exceptions.
Rio Tinto · Exceptions only
A central system can drive mine trucks on fixed routes while people supervise and manage exceptions remotely.
Georgia State University · Exceptions only
Enrollment support does not require counselors to initiate and answer every routine student interaction; a task-aware assistant can provide personalized outreach while routing exceptions to staff.