Leadership insights shaping the future of inclusive mobility.

Most agency conversations stop at copilot tools and ChatGPT. Deeper integration, where agentic AI sits inside workflows like accounting, payroll, HR, and eligibility processing, is far less common. The barrier is the data foundation.

Contract interpretations drift over years. On-time performance can be calculated five ways inside the same operation. Layer AI onto inconsistent definitions and you get confident answers built on shifting ground. The work happens upstream of any AI deployment. Standardize the definitions, name the owners, build the audit habit.

The accessibility conversation centered on universal mobility. The principle is simple. Design accessibility into the system from the start, not as a compliance review at the end.

TriMet in Oregon came up as a working example. The agency builds disability awareness into hiring, training, and operations, so accessibility is structural rather than bolted on.

The concern about AI is specific. As call centers automate, agencies are deploying these systems without consistent input from the disability community. The research on how riders with verbal, auditory, cognitive, and intellectual disabilities interact with AI agents is still thin. The fix is to invite disability advocates into product development from day one, not at compliance review.

COTA faced real risk in replacing their 25-year-old paratransit software system. They handled it with a sandbox procurement model. Rather than trust vendor claims, they tested the software against 105 specific business needs over six months, bringing in dispatchers, reservationists, and drivers before committing.

The operations team uses the software every day, so COTA let them lead the procurement and evaluation. That produced higher confidence and better alignment with daily needs than letting IT dictate the choice.

COTA successfully moved its core paratransit service over in phases. They tackled the core service first, added a digital wallet to cut reliance on cash, and are now digitizing eligibility step by step.

Turning the feature on is roughly 10% of the work. COTA's rider training shows what the other 90% looks like. Monthly orientations now center on the rider app and digital wallet. Travel trainers are out in the community teaching customers how to use the platform. iPads sit at the front desk so walk-in customers can set up accounts on the spot, and outreach to nonprofits and medical providers is planned ahead of the eligibility rollout. Any agency planning a similar transition should budget for this from the start.

AI in transit is shifting from answering questions to investigating on its own. Couple it tightly with a solid analytics engine and your whole staff can understand what's happening in the operation without waiting on a data science team.

AI isn't just for executives. Tools like Technician Copilot let mechanics speak into a system to log work orders and diagnose issues, capturing asset data that used to disappear on paper clipboards. That feeds straight into vehicle availability, and from there into on-time performance.

AI is being deployed faster than the research on how riders with verbal, auditory, cognitive, or intellectual disabilities interact with it. The disability community isn't consistently at the table when those decisions get made.

The same deployment can produce accessibility gains and accessibility risks at once. For riders with speech impairments, language barriers, or conditions that make phone calls hard, AI Voice is turning out to beat a human call center. It's patient, it asks for clarification without frustration, and it works in more than 60 languages by default. Barriers that used to route these riders away from phone service drop significantly. Closing the gap the rest of the way takes more involvement from the community.

Before adopting any AI tool, agencies need a way to stress-test what they're actually buying. Four questions are worth asking: 

Autonomous vehicles aren't a distant future anymore. They're taking market share in cities like San Francisco, where Waymo has passed Lyft. The economics could reshape how service gets delivered. Today's ride-hailing runs $2.50 per mile, an NYC bus runs $0.50 per mile, and projected autonomous shared service runs $0.10 per mile.

The useful question isn't whether AVs arrive. It's what operations looks like with 10,000 of them in a city, not just two.

The assumption that AVs replace operators one-for-one is probably the wrong model. Vancouver's SkyTrain has run fully autonomous for 40 years, and the work didn't disappear. It shifted. Staff help riders find platforms, assist with the system, and handle customer support. The reframe is worth sitting with. The same budget can deliver far more service when the roles change rather than vanish.

The agencies that end up well positioned aren't running a separate AV strategy. They're doing the foundational work that already pays off elsewhere. Clean, trusted data across modes. An architecture that integrates multiple providers. Policies that anticipate edge cases. The harder questions, labor agreements, curb management for riders with disabilities, paratransit boarding assistance, are ones the industry is working through together, not problems any single agency has to solve alone.