BT scheduled field engineers dynamically and saved $150M a year

BT already had Work Manager to automate field communications, but repair and installation work kept changing as new faults, appointments, skills, and absences appeared. A simple real-time allocator could miss longer-lookahead tradeoffs.

In 1996 BT Labs added Dynamic Scheduler, mixing heuristic search and constraint-based reasoning. The system let controllers see provisional schedules, run what-if views, and manage jeopardy before appointments failed.

BT rolled it out nationally from 1997. By 1998 it covered 20,000 engineers and was reported to save $150 million per year, with an estimated $250 million per year when extended to a 40,000-person workforce.

Reschedule continuously, but keep plans stable

The system optimized a live service network where tomorrow's perfect plan is useless if today's customers already moved.

It balanced efficiency with schedule stability, which made controllers willing to trust it.

The interface turned optimization into operational control, not a black-box dispatch list.

In field service, the best schedule is not just cheapest; it must react to new work without destroying trust in the plan.

INFORMS highlighted the work as an Edelman finalist, and the Interfaces paper became a durable reference for large-scale workforce scheduling in telecommunications. The idea traveled well: dynamic scheduling must be both predictive and reactive.