Artificial Intelligence, Public Transit, and the Geometry of Cities

Philip Pugliese (Co-Founder)

Overview

Public transit leadership today operates in an environment defined by rapid technological change and stubborn physical constraints. New mobility modes—micromobility, ride-hailing, on-demand shuttles, and increasingly autonomous vehicles—promise flexibility and personalization. At the same time, transit agencies remain accountable for moving large numbers of people safely, affordably, and equitably through dense urban environments that are governed by immutable rules of geometry, time, and space.

Artificial Intelligence (AI) is often presented as a solution to this tension. For transit executives, the relevant question is how it can support management decisions under real-world constraints.

The Changing Operating Environment

Urban mobility is no longer defined by a single mode. Riders increasingly combine buses, rail, bikes, scooters, walking, and ride-hailing within a single trip. Autonomous vehicle technology promises future efficiency gains but remains constrained by regulatory, safety, and cost realities. Electrification introduces new operational dependencies on energy systems and charging infrastructure. Demand is more volatile, labor markets are tighter, and public expectations for reliability and transparency are higher.

Despite this complexity, the fundamental problem facing public transit has not changed: how to move people efficiently through limited space. Roadway width, signal cycles, dwell times, vehicle capacities, and stop spacing impose hard limits on throughput. No amount of software can eliminate congestion on a constrained corridor or increase capacity without trade-offs.

This is where AI must be understood not as a replacement for planning fundamentals, but as a management tool for navigating constraints.

From Static Planning to Adaptive Management

Traditional transit planning relies on averages, historical patterns, and periodic updates. Schedules, routes, and resource allocations are often fixed for months or years at a time. In a stable environment, this approach is serviceable. In today’s dynamic conditions, it creates risk.

AI enables a shift from static planning to adaptive management. Predictive models can forecast ridership, congestion, vehicle availability, and labor shortfalls with significantly greater accuracy than rule-of-thumb approaches. For transit leadership, this translates into earlier visibility of emerging problems and more time to respond deliberately rather than reactively.

The value is not automation for its own sake. It is reducing uncertainty in executive decision-making.

Managing Trade-offs, Not Eliminating Them

Public transit leaders rarely face binary choices. Nearly every operational decision involves trade-offs: coverage versus frequency, reliability versus cost, peak capacity versus all-day service, operator equity versus schedule efficiency. These trade-offs are further constrained by labor agreements, funding requirements, accessibility mandates, and political oversight.

Planning and Optimization AI supports management by explicitly modeling these competing objectives. Rather than producing a single “optimal” answer, modern AI systems can generate a range of feasible options and quantify their implications. Leadership can then choose strategies that align with current priorities—whether that is cost containment, service equity, or resilience—while understanding what is being sacrificed elsewhere.

This capability is especially important as agencies integrate micromobility and on-demand services into their networks. These modes do not eliminate geometric constraints; they redistribute demand across time and space. AI helps evaluate when such redistribution improves system performance and when it merely shifts congestion or cost.

AI in a World of Physical Constraints

It is tempting to view autonomous vehicles and AI-driven dispatch as a path around the limitations of fixed-route transit. In reality, autonomy does not repeal geometry. Ten autonomous vehicles still occupy the same road space as ten human-driven vehicles. Increasing the number of AVs will improve flexibility yet will not improve network capacity.

For public transit leadership, the implication is clear: high-capacity modes remain essential in dense corridors, while flexible services are most effective as feeders, complements, or coverage solutions. AI adds value by coordinating these layers—aligning fixed-route capacity with variable-demand services, optimizing transfers, and managing curb space and dwell times more intelligently.

In this sense, AI becomes a system-level integrator rather than a mode-specific solution.

Human-AI Teaming in Transit Management

Another misconception is that AI replaces professional judgment. In practice, its greatest value lies in supporting human decision-makers. Dispatchers, planners, and executives bring contextual knowledge, institutional memory, and ethical judgment that algorithms do not possess. AI excels at processing scale and complexity beyond human capability.

Agentic AI systems are beginning to automate repetitive operational tasks—monitoring vehicle locations, adjusting assignments, coordinating charging, or flagging exceptions—while escalating only high-risk or ambiguous situations to human staff. This allows leadership to redeploy scarce human expertise toward oversight, strategy, and stakeholder engagement rather than constant firefighting.

For agencies facing workforce shortages, this form of augmentation is often more impactful than headcount expansion.

Accountability, Explainability, and Trust

Public transit operates under intense public and regulatory scrutiny. Decisions must be explainable, auditable, and defensible. Black-box systems that cannot justify their outputs are ill-suited to this environment.

Purpose-built AI systems for transit increasingly combine machine learning with rule-based and optimization methods, ensuring that recommendations respect hard constraints and policy requirements. For leadership, this means AI can be used not only to improve performance but also to strengthen governance, providing clear rationales for decisions and measurable performance baselines.

Conclusion

Artificial Intelligence will not change the geometry of cities, eliminate congestion, or remove the need for hard choices in public transit. What it can do is help leadership see those constraints more clearly, evaluate options more rigorously, and act with greater confidence in a volatile environment.

As micromobility, autonomy, and new service models continue to emerge, the agencies that succeed will be those that treat AI not as a technology experiment, but as a core management capability—one that respects physical realities while enabling smarter, faster, and more transparent decisions.

In the end, the role of AI in public transit is not to replace fundamentals, but to help leaders manage them better.