By Kalin Stanojev, Vice President, AI Platform at SafeRide Health
For most health plan leaders, the loudest stories about AI over the past two years have been cautionary ones: cases of prior authorization tools making high-stakes denials at scale, with little oversight or explanation about how they arrived at those decisions. The backlash now colors how many health plans receive pitches that include AI.
The best way to cut through AI skepticism is to get specific about which jobs AI is ready to do well and then show health plans where it's working. Medical transportation is a good place to make this concrete, because the work splits cleanly into tasks AI handles well and tasks it does not.
The most effective non-emergency medical transportation (NEMT) brokers today are focused on improving the efficiency of their processes, tools, and operations. A lot of that is through automation and AI. A 2025 survey from the National Association of Insurance Commissioners found that 84% of responding health insurers were using AI for fraud detection, utilization management, and prior authorization—and transportation companies are no different.
Guardrails and clear roles matter here: AI should take the high-volume, quick-analysis work so people are free for the decisions that turn on judgment and empathy. AI can make human work better, too, by surfacing the right context and direction, so people are already informed before they handle the hard calls. Used that way, the same tools improve cost, member experience, and benefit integrity.
Using AI to Make Humans’ Jobs Easier and More Effective
AI at its best makes people more efficient and effective in their jobs, saving them time spent on “busy work” that could be used for important human interactions. For example, an Interactive Voice Response (IVR) system with AI agents can field phone calls and direct members to the best solution, saving call representatives’ bandwidth for critical needs while ensuring that members continue to receive fast, accurate, human-centered service.
In the medical transportation space, when a member calls into their transportation provider, agents (whether human or AI) should have secure access to the member’s past rides, health plan benefit structure, eligibility, and specific mobility needs. Many ride bookings can be handled quickly and seamlessly by the IVR, freeing care teams to focus on the calls that genuinely require human judgment, empathy, and problem-solving. In addition, all calls can and should be recorded for sentiment analysis and quality control.
Recorded calls and sentiment data do more than catch quality issues. Across many rides, they start to show what an individual member values, whether it's a particular vehicle type, a preferred pickup window, or extra help at the door, so their next booking and ride experience can be shaped around them. Listening at that scale is exactly what AI is suited to, and it turns routine quality control into a way to personalize the experience.
Turning Better Data into Better Decisions
AI that relies on fragmented, disconnected data produces fragmented answers. SafeRide anchors its operations in a proprietary technology platform that unifies the entire transportation operation in one place. Rather than relying on disconnected systems for scheduling, dispatch, routing, GPS tracking, and reporting, the platform brings all functions into one secure, cloud-based architecture, and it captures detailed data on every ride. This allows for consistent benefit administration across markets, transparent oversight, and a higher-quality, more predictable experience for health plans. It also gives the models far more to work with, from fulfillment trends to provider performance.
That data can be aggregated and used in real time to monitor and improve operations; once services are complete, health plans then receive detailed performance reports. We are now working to make that data work harder, organizing it so AI can move faster from what happened to why, and to what to do about it. A connected platform produces cleaner data to begin with, and AI turns it into insight faster than a team working by hand. That insight points to a specific action: reassigning a struggling route, flagging a provider trend, or spotting a pattern in grievances across a market rather than logging each one.
Getting Members to the Right Place
When technology is fully integrated into NEMT operations, routing and trip assignments are determined by a combination of algorithmic optimization and real-time operational judgment. For instance, SafeRide’s Ride Assignment Algorithm powers about 95% of all rides and takes many factors into account when dispatching a ride to a member, including supply and demand, required modality, and historical provider performance. Leading NEMT providers and rideshare options are scored based on all these factors, with the highest-scoring provider scheduled for the ride. These calculations all take place in seconds. If a transportation provider cannot fulfill a trip, the ride automatically reenters the queue and is reassigned to the next best option, so the member doesn’t miss a ride or their appointment.
Getting a member to the right place is deceptively hard. A member often knows roughly where they're going, but not the exact address, suite, or which of several providers at that location they're seeing. The provider side is just as tangled: Many providers share one address, some providers operate across many sites, and names don't always map cleanly to locations. There’s often also a rules layer, in which approved providers vary by plan and state, and the right destination is the nearest match that's also covered.
AI is well-suited to this ambiguity when it's applied with care, pulling together address, provider, and eligibility sources to resolve incomplete information into one correct, compliant destination. Sometimes the fix is as small as a missing street suffix; sometimes it's choosing among a dozen clinics at one address. The member never sees it, but AI can help determine whether the ride reaches the right place and whether it holds up as compliant, and then reconciles it cleanly afterward.
Setting Up Guardrails Across the Entire Ride Journey
Effective benefit integrity frameworks validate benefits and eligibility in real-time at booking; automatically enforce rules related to modality, distance, and authorizations; detect duplicate rides, inflated mileage, and utilization anomalies; and create full audit trails and role-based access controls. These are crucial steps to prevent fraud, waste, and abuse (FWA).
SafeRide is now working to run all that information through machine learning to assign a probability score to each ride, so instead of being reactive, we can be even more proactive in preventing FWA. For scheduled rides that have a high risk of being wasteful or fraudulent, we can have our specialized team review them in greater depth. For instance, if someone is traveling 200 miles to a dialysis center when there's one two miles down the road, our team can surface the ride to the health plan for review and, if appropriate, address it. This allows us to manage benefits more effectively and equitably.
We also recognize the importance of tightening FWA checks without impacting the member experience. Flagging so many rides that members face delays and denials is the failure mode that gave prior authorization its reputation. The test that matters is whether member experience holds up alongside enforcement.
In a six-month period with a Medicaid plan, for example, SafeRide implemented a connected set of interventions across the ride journey, bringing utilization down to a stable lower level while member experience metrics held steady throughout. That’s a sign the system was designed with intention, but it only stays that way if you monitor member sentiment as closely as you track recovered dollars.
Being Honest About What AI Has Not Earned Yet
Promoting AI's strengths matters, but being honest about what it can't yet do well matters just as much. The dialysis case above is a good example: A 200-mile trip might be a case of FWA, or it might be a member traveling that far for a legitimate reason, and the data alone cannot make that determination. That’s exactly why the judgment call belongs with the right team at the member’s health plan. Reading intent, spotting collusion among multiple parties, and detecting brand-new schemes that look nothing like past data are all harder for a model than flagging a simple billing anomaly. None of this is a permanent ceiling, though. These tools are improving quickly, and the line between what AI can handle on its own and what still needs a person will keep shifting.
This is where we focus our own work. We're pushing on how AI gets applied to non-emergency medical transportation, and we're just as deliberate about doing it in ways that are auditable, controllable, and secure enough for enterprise scale. The honesty about what AI can't do yet is what makes that worth doing, because it keeps the investment pointed at real value. As the models improve, we keep narrowing the work that reaches a person to the cases that truly need human judgment.
A Future Fueled by Technology and People
A smarter platform and human-guided technology produce better outcomes for everyone: Members who reach their appointments, health plans that control costs and reduce FWA exposure, and a system that works equitably at scale. The stakes are real on both sides: A study using Medicaid claims data found that reliable transportation to dialysis appointments saves Medicaid more than $41,000 per patient annually in avoided complications and hospitalizations. Success often comes down to the quality of the technology and human skill behind the benefit.
__
About Kalin Stanojev
Kalin Stanojev leads AI strategy and platform at SafeRide Health, where he is helping shape how AI transforms the company’s operations, products, and ability to serve the healthcare transportation market at scale. A former founder with deep product and technology leadership experience, he has spent the last decade building leading-edge applied AI systems that turn rapid innovation into scalable business value, with accountability and control built in by design.
About SafeRide Health
SafeRide Health is a healthcare technology and services company working to improve access to care through smarter, more reliable non-emergency medical transportation. SafeRide helps health plans and care organizations coordinate rides across a national network while improving visibility, compliance, member support, and operational performance. To learn more, visit saferidehealth.com.