Boston Public Schools
Verified evidenceCoordination compressedCreator to judgeRoutes can be optimized as one district system instead of built school by school and then combined.
Enrollment · Boston, Massachusetts, United States
Collections: Human still decides · Queue eliminated
Executive brief
The breakthrough was changing the unit of optimization from each school to the whole district, then positioning human routers as exception designers rather than route constructors.
AI value · Buses and annual operating cost
Verified50 buses removed in the first implementation year and about $5 million saved, with average ride time remaining about 23 minutes
Implemented operational result reported by the peer-reviewed design team; not a randomized trial.
Before
Build and maintain routes manually for each school over multiple weeks. → Connect school-level routes and assign buses.
After
Generates district-wide student stops, route sequences, and multi-school bus reuse in about 30 minutes. → Review and adjust the optimized base plan for operational exceptions.
Human boundary
The optimizer proposes a feasible system plan; district staff approve and modify routes.
Why it matters
Routes can be optimized as one district system instead of built school by school and then combined.
Before
How the work ran before the change.
Human transportation routers
Build and maintain routes manually for each school over multiple weeks.
ControlLocal school and vehicle constraints
District planners
Connect school-level routes and assign buses.
ControlManual system reconciliation
What changed
Routes can be optimized as one district system instead of built school by school and then combined.
Decision rightHuman authority remains at the consequential boundary
After
How the same work runs now.
BiRD optimization algorithm
Generates district-wide student stops, route sequences, and multi-school bus reuse in about 30 minutes.
ControlRide-time, accessibility, vehicle, and policy constraints
Human routers
Review and adjust the optimized base plan for operational exceptions.
ControlDistrict retains final plan authority
Exception path
Human routers adjust for wheelchair vehicles, door-to-door service, monitor needs, student conflicts, and operational knowledge.
Work removed
Decision authority
The optimizer proposes a feasible system plan; district staff approve and modify routes.
Before
Human transportation routers
Build and maintain routes manually for each school over multiple weeks.
Control: Local school and vehicle constraints
District planners
Connect school-level routes and assign buses.
Control: Manual system reconciliation
After
BiRD optimization algorithm
Generates district-wide student stops, route sequences, and multi-school bus reuse in about 30 minutes.
Control: Ride-time, accessibility, vehicle, and policy constraints
Human routers
Review and adjust the optimized base plan for operational exceptions.
Control: District retains final plan authority
Work that left the path
Human role before
Routers spent weeks constructing and reconciling school-level routes.
Human role after
Routers validate and tweak a system-level optimized base plan and manage policy and student-specific exceptions.
AI roleFlexible integer-programming optimizer that assigns stops, sequences routes, and reuses buses across schools.
Approximately 650 buses in the manual solution→50 buses removed in the first implementation year and about $5 million saved, with average ride time remaining about 23 minutes
Fall 2017 implementation · Boston Public Schools district routing across public, private, and charter schools
Implemented operational result reported by the peer-reviewed design team; not a randomized trial.
Anti-pattern
Optimizing only cost while ignoring ride time, accessibility, and family-level constraints.
Questions
Portability conditions
Reputation risk
low
Evidence and authority
Current · updated
1 independent, 1 peer reviewed; publication outcomes are verified.
Bundle 1.0.0 · reviewed 2026-08-23 · stable ID 23a3326b0ffa1d4b
Montgomery County, Maryland · Coordination compressed
A multilingual assistant can answer first-line 311 questions while urgent or unresolved needs route to people.
Greater Houston primary stroke centers · Coordination compressed
Large-vessel occlusion at a non-thrombectomy primary stroke center has to be recognized and the receiving team activated through sequential imaging review and conventional communication before transfer can start.
City of Pittsburgh and Carnegie Mellon University · Coordination compressed
That urban intersections should follow fixed timing plans optimized offline for predictable rush periods.
UPS · Coordination compressed
That each driver or dispatcher must determine the daily stop sequence from experience and static route plans.