When the Driver's Seat Is Empty

What autonomous vehicles mean for people who need more than just a ride

This is Part 3 of a three-part series on the evolution of healthcare mobility. [← Part 1: The Rideshare Revolution Comes to Healthcare] [← Part 2: The Hybrid Network Era]

Myrna Peterson is 76 years old. She has quadriplegia and uses a wheelchair. She lives in Grand Rapids, Minnesota — a town of about 11,000 people in the rural north-central part of the state, where transportation options are scarce and the nearest major city is over 75 miles away.

Myrna is also a regular rider of autonomous vehicles.

Through the goMARTI program — Minnesota's Advanced Rural Transit Initiatives — Myrna rides Toyota minivans outfitted with cameras, radar, GPS, and laser sensors that navigate Grand Rapids' streets without a human driver. The program has delivered over 25,000 rides and recently expanded into its second phase, combining autonomous vehicles with electric and hybrid options. Myrna's advocacy helped attract the government funding that made it possible. As she told reporters, the service gives her something she hadn't had in years: the ability to go where she wants, when she wants.

Her story captures both the extraordinary promise and the hard questions of autonomous vehicles in healthcare mobility. The technology is real, it's working, and it's serving people whose lives are most constrained by transportation barriers. But for the populations at the center of healthcare mobility — older adults with complex needs, people with intellectual and developmental disabilities, wheelchair users, individuals with behavioral or cognitive challenges — the arrival of AVs raises a set of questions that the industry needs to confront honestly.

This isn't theoretical anymore

The autonomous vehicle landscape has matured faster than many expected. Waymo provided more than 14 million trips in 2025, more than tripling its ride volume from the prior year. The company now operates fully driverless vehicles — no human safety operator — in Austin, San Francisco, Phoenix, Atlanta, and Los Angeles, with expansion planned to over 20 additional cities in 2026, including Detroit, Las Vegas, Nashville, San Diego, and Washington, D.C. Zoox, owned by Amazon, has roughly 50 robotaxis running in San Francisco and Las Vegas with paid service planned for late 2026. More than 35 states have passed AV-related laws, and 25 states introduced 67 new AV bills in early 2025 alone.

But the AV deployments most relevant to healthcare mobility look quite different from a Waymo robotaxi in San Francisco. They look like the programs specifically designed around older adults and people with disabilities.

In Detroit, May Mobility launched "Accessibili-D" in June 2024: a free autonomous vehicle service for residents age 62 and older and those with disabilities. The program deploys three AVs — two of them wheelchair-accessible — across 68 stops in 11 square miles of downtown Detroit, six days a week. By mid-September 2024, 82% of riders were repeat users, riding to medical appointments, grocery stores, and social activities. In Arlington, Texas, May Mobility's RAPID program achieved 99% on-time performance with vehicles operating fully autonomously 80% of the time and surpassing 100,000 rides.

These aren't tech demos. They're transportation services used by real people with real needs — and the early results suggest strong demand.

The economic case

The cost trajectory for autonomous vehicles is striking. Per-mile AV operating costs are projected to fall from roughly 84 cents in 2018 to 51 cents in 2025 and 33 cents by 2035. Insurance costs alone are expected to drop more than 50% by 2040. By 2030, per-mile robotaxi costs are expected to be competitive with private car ownership and significantly cheaper than current ride-hailing services.

For healthcare mobility, the implications extend beyond unit economics. Autonomous vehicles don't have shift limits. They don't call in sick — a fact that would have been especially useful during the pandemic-era driver shortages described in [Part 1 of this series]. They can be repositioned algorithmically across a service area based on predicted demand. They can operate in the early morning and late evening hours when human-driven service is expensive and hard to staff. And in rural areas — where driver recruitment is often the single biggest constraint on transportation supply — AVs could provide on-demand coverage that simply isn't feasible today. The goMARTI program in Grand Rapids, Minnesota is proof of concept.

At scale, the potential is transformative. An estimated 2 million individuals with disabilities could access employment if transportation barriers were eliminated, with associated healthcare savings estimated at $19 billion annually from prevented missed appointments alone.

The hard questions

But cost curves and pilot metrics only tell part of the story. For the populations at the center of healthcare mobility, the arrival of AVs raises a set of questions that don't have easy answers.

Physical accessibility is the most concrete challenge. Most wheelchair users currently require a human attendant to secure their wheelchair inside a vehicle before the trip can begin. Automated wheelchair securement systems are an active area of research but not yet widely available. Vehicle doorways need to accommodate 870mm width and 1508mm access depth to fit 95th-percentile wheelchair users — dimensions that not all AV platforms are designed for. Ramps or lifts must be side-entry and operable without assistance. And accessible pickup and dropoff infrastructure at healthcare facilities matters as much as accessible vehicles.

Cognitive and behavioral accessibility is the less visible but equally important concern. How does a person with a cognitive disability interact with a driverless vehicle interface? What about a non-verbal passenger who currently relies on their driver to understand their needs? Consider the PACE participant experiencing sundowning in the late afternoon, or the individual with I/DD who becomes anxious in unfamiliar situations. These are challenges that a trained, familiar human driver manages through experience, patience, and relationship — qualities that no vehicle user interface can replicate.

And then there's the gap that matters most for healthcare mobility: the driver as caregiver. In PACE programs, I/DD day programs, and other high-touch settings, the driver is far more than someone behind the wheel. The driver is an observer — noticing when a passenger seems more confused than usual, or is limping, or hasn't eaten. The driver is a safety monitor — managing elopement risk for passengers who may try to leave the vehicle, knowing the seizure protocol for passengers who need one. The driver is a communication bridge — relaying information to the care team about what happened during the ride. And the driver is a source of continuity and trust — the same face at the same time, day after day, which for many passengers with cognitive or behavioral needs is one of the most stabilizing elements of their routine.

When you remove the driver, you don't just lose a person behind the wheel. You lose a care function.

Here's a fact that makes the stakes concrete: as of 2025, neither the federal government nor any state requires autonomous vehicles to be accessible for people with disabilities. The U.S. Access Board has published guidance on inclusive AV design, and the National Council on Disability has recommended that AV products incorporate accessibility from the start. But these are voluntary frameworks, not mandates. If the industry follows the path of least resistance, autonomous vehicles will expand transportation options for able-bodied passengers long before they become meaningfully accessible to the people who need new options most.

Reimagining the human role

None of these accessibility challenges are reasons to reject autonomous vehicles. They're reasons to integrate them thoughtfully — and that starts with rethinking what the human role in healthcare transportation actually is.

An alternative model is already emerging in early pilot programs and academic research. Instead of replacing the human entirely, it redefines what the human does.

The concept is sometimes called the attendant model or the ride companion model. Instead of a driver, a caregiver or trained aide rides with the passenger in the autonomous vehicle. The aide's entire focus is on the passenger — not on traffic, not on navigation, not on parking. They can assist with boarding and wheelchair securement. They can monitor for health or behavioral changes. They can provide the reassurance and continuity that vulnerable passengers need.

Research on AV adoption among people with disabilities shows that on-board attendant presence significantly increases willingness to use autonomous vehicles. The effect is especially strong in shared-ride scenarios, where passengers express concern about traveling with strangers without a human operator present. Meanwhile, caregivers themselves benefit from not having to drive — they can focus entirely on the person they're supporting, and for passengers who currently ride with a family member, the AV could relieve that family member of driving duties while preserving the companionship.

For programs, this shift has real operational implications. The staffing model changes: fewer CDL-holding drivers, more trained aides and companions. Training requirements pivot from vehicle operation to behavioral support, health observation, and communication. And the system orchestrating it all must now track aide assignments, aide-passenger compatibility, and aide qualifications alongside every other variable.

This isn't a simple transition. But it may be the right one — a model where technology handles the driving and humans handle the caring.

The system that holds it all together

Across this three-part series, we've traced an arc: from the TNC revolution that brought on-demand supply and consumer-grade experience to healthcare transportation, through the hybrid network era that blended those new resources with dedicated fleets and traditional NEMT providers, to the approaching age of autonomous vehicles that could reshape the economics and geography of the entire system. At each stage, the supply has gotten richer — and the orchestration challenge has gotten harder.

Consider what a transportation program serving older adults or people with disabilities might need to manage in the next five to ten years: dedicated fleet vehicles, contracted NEMT providers, TNC partners, and autonomous vehicles — some with human aides aboard, some without. Each modality has different capabilities, different cost profiles, different strengths. The question of which one is right for a given trip depends entirely on who the passenger is and what they need.

That's why the most important piece of infrastructure in healthcare mobility isn't any particular vehicle or provider. It's the transportation system of record — the comprehensive data and decision layer that captures every passenger's accessibility profile, communication needs, behavioral considerations, and care requirements; that understands every provider's capabilities and constraints; and that matches the right resource to the right trip, every time, across the full network.

Without this layer, every new supply source creates more risk — more opportunities for a passenger to be matched with a vehicle or service level that doesn't meet their needs. With it, every new supply source genuinely expands access, independence, and quality of life for the populations who stand to benefit most.

The vehicles are going to keep getting smarter. The networks are going to keep getting more complex. The populations at the center of healthcare mobility — the Myrna Petersons, the Mrs. Alvarezes, the PACE participants and day program riders and dialysis patients — deserve a system that's smart enough to match.

Healthcare mobility has always been about more than getting from point A to point B. As the driver's seat empties and the network grows, the organizations that succeed will be the ones that never lose sight of that.

This is Part 3 of a three-part series. Read [Part 1: The Rideshare Revolution Comes to Healthcare →] and [Part 2: The Hybrid Network Era →].