TL;DR
- Automated scheduling raised trips per vehicle hour by 8 to 31 percent at six paratransit agencies that shared before-and-after data.
- States and brokers are adding GPS trip checks and data tools to NEMT contracts to catch billing problems before claims are paid.
- AI can give confident answers that are wrong, and speech recognition tools have made more errors for some groups of callers than others.
- US laws still apply when AI is used, including federal phone rules, HIPAA, and the False Claims Act.
- The NEMT provider stays responsible for every trip record and claim, whether a person or a tool created it.
Non-emergency medical transportation (NEMT) runs on phone calls, trip schedules, and trip records. Many software companies now sell AI tools for each of these tasks. This post looks at what public data shows about where these tools help, where they fall short, and which US laws apply when an NEMT provider uses them.
Where AI Helps in NEMT
AI and automated software help most with tasks that follow clear rules and involve large amounts of data, such as building schedules and checking trip records for errors.
Scheduling and Routing
The best public data on automated scheduling comes from ADA paratransit, which is run by public transit agencies. Like NEMT, paratransit is a demand-response service, where riders book trips and the provider assigns them to vehicles.
A 2023 study for the Transit Cooperative Research Program looked at continuous dynamic optimization, which is software that keeps updating trip schedules throughout the day. According to the US Department of Transportation’s summary, six agencies shared data from before and after they started using it. Their results were:
- Trips per vehicle hour rose between 8 and 31 percent.
- On-time performance rose 17 percent on average.
- Operating costs changed from a 30 percent decrease to a 4 percent increase, with an average reduction of 13 percent.
The same study found that customer complaints fell at most agencies that shared data, but rose 50 percent at one. The technology was first built to schedule trips that were still unassigned after dispatchers had gone home for the day.
These results come from transit agencies. Results at a private NEMT company can differ based on its fleet size, service area, and trip mix.
Trip Verification and Billing Checks
Billing errors and fraud are a known problem in Medicaid NEMT. A 2026 Health Management Associates review found that states and health plans are adding technology to their NEMT contracts to prevent these problems. The strategies it lists include:
- GPS trip verification, electronic trip logs, and real-time data analysis to flag unusual billing patterns before claims are paid.
- Extra verification for standing orders, which make up the largest share of trips.
- Mobile apps that let riders submit mileage reimbursement while brokers check the miles with GPS.
The same review notes that Wisconsin’s 2025 NEMT contract included penalties for fraud that the broker fails to detect. Some of these tools use AI, and others use simpler rules. The review notes that they replace paper logs and manual checks that were prone to error.
Reminders and Trip Updates
Some NEMT providers use AI voice or text tools to send trip reminders and status updates. These calls and texts fall under federal phone rules, which are covered in the section on US laws below.
Where AI Falls Short
AI tools can make mistakes that are hard to spot, and they can work better for some riders than others.
Confident Answers That Are Wrong
The National Institute of Standards and Technology (NIST) lists “confabulation” as one of the main risks of generative AI in its Generative AI Profile. NIST describes this as AI producing false information stated with confidence. In NEMT, this could mean a wrong pickup time, a wrong address, or the wrong level of service entered into a trip record.
Understanding Every Caller
A 2020 Stanford study tested speech recognition systems from Amazon, Apple, Google, IBM, and Microsoft. All five made nearly twice as many errors for Black speakers as for white speakers. On average, the systems got 35 of every 100 words wrong for Black speakers and 19 of every 100 words wrong for white speakers.
The study did not test older adults or people with speech disabilities. It does show that the accuracy of voice tools can change from one caller to another. This matters in NEMT, because Medicaid members who qualify based on age or disability use NEMT at higher rates than average, according to a CMS report to Congress.
Results That Vary by Operation
The paratransit study above found a wide range of results. Productivity gains ranged from 8 to 31 percent, and one agency saw complaints go up. A tool that works well for one operation may not produce the same results for another.
US Laws That Apply When NEMT Providers Use AI
No single federal law covers AI in NEMT. Existing laws still apply to the tasks AI performs, such as making calls, handling health information, and creating records used for billing.
| Law | What It Covers | How It Applies to AI |
|---|---|---|
| Telephone Consumer Protection Act (TCPA) | Calls that use an artificial or prerecorded voice. | The FCC ruled in 2024 that AI-generated voices count as artificial, so these calls require the prior express consent of the person being called. |
| HIPAA | Vendors that create, receive, maintain, or send protected health information for a covered entity. | An AI vendor that handles this information is a business associate and needs a Business Associate Agreement, even if the data is encrypted. |
| False Claims Act | False claims submitted to the federal government. | “Knowingly” includes reckless disregard of whether information is true, and no intent to defraud is required. |
| California AB 3030 | AI messages about clinical information from clinics and doctors’ offices. | Requires a notice that AI wrote the message. It does not cover scheduling or billing messages. |
| ACA Section 1557 rule | Tools used in clinical care decisions. | Requires steps to prevent discrimination. It does not cover scheduling or billing tools. |
Sources: FCC Declaratory Ruling FCC 24-17; HHS Guidance on HIPAA and Cloud Computing; 31 U.S.C. 3729; California AB 3030; 45 CFR 92.210.
Federal Phone Rules
In February 2024, the Federal Communications Commission confirmed that the TCPA’s rules on artificial or prerecorded voices include AI technology that creates human-sounding voices. As a result, calls made with these voices require the prior express consent of the person being called. NEMT providers that use AI voice tools for reminders or updates need records showing that riders gave consent.
Health Information Privacy
Trip requests often include protected health information, such as a rider’s Medicaid ID, medical appointment, or mobility needs. HHS guidance on cloud services states that a vendor that creates, receives, maintains, or sends this information for a covered entity is a business associate. This applies even when the vendor only stores encrypted data. An NEMT provider that is a HIPAA covered entity needs a Business Associate Agreement with any AI vendor that handles this information.
Billing Accuracy
The False Claims Act makes it illegal to knowingly submit false claims to the federal government, including Medicaid. The law defines “knowingly” to include acting in reckless disregard of whether information is true, and it does not require proof of intent to defraud. Trip records and claims built from AI output need the same review as any other record before they are billed.
State and Clinical AI Rules
Some rules written for AI in health care do not reach NEMT scheduling. California’s AB 3030, in effect since January 1, 2025, requires clinics and doctors’ offices to tell patients when AI wrote a message about their clinical information. The law does not cover appointment scheduling or billing messages. The HHS rule under Section 1557 of the Affordable Care Act covers AI tools used in clinical care decisions, and it does not apply to scheduling or billing tools.
Where Trained Agents and Dispatchers Fit
AI tools can handle parts of the work, and people remain responsible for the results. In NEMT, phone agents and dispatchers do different jobs.
Phone agents answer calls, book trips, and collect trip details. The level of service, pickup and drop-off addresses, and appointment details they collect need to be correct, because they become the record used to bill for the trip.
Dispatchers work with drivers throughout the day. When a rider cancels or reschedules, the dispatcher assigns the driver a new ride so the vehicle does not travel empty. Scheduling software can suggest these changes, and dispatchers confirm them with drivers and riders. When a provider uses AI tools, trained staff can review what the tools produce. That includes checking trip details before they are billed and taking over calls the tools do not handle well.
Trained NEMT Agents and Dispatchers for Your Operation
Pac Biz answers your phones with trained NEMT agents, so no call goes unanswered and no trip gets lost. We also provide dispatchers who coordinate with your drivers throughout the day.
Using AI and People Together in NEMT
Public data shows that automated scheduling and data tools can improve productivity and help catch billing problems. The same data shows that results vary, and research shows AI tools can make errors that are hard to spot. US laws on phone calls, health privacy, and billing apply whether a person or a tool does the work. NEMT providers that use AI still need trained staff to check records, handle calls the tools cannot, and keep every trip record accurate.
Sources
- US Department of Transportation, ITS Deployment Evaluation. Continuous Dynamic Optimization (CDO) Technology for Automated Paratransit Scheduling Operations Improved Transit Service Productivity Up to 31 Percent. Summarizing TCRP Synthesis 168 (2023). https://www.itskrs.its.dot.gov/2023-b01747
- Health Management Associates. Strategies to Address Fraud, Waste, and Abuse in Non-Emergency Medical Transportation. March 9, 2026. https://www.healthmanagement.com/blog/strategies-to-address-fraud-waste-and-abuse-in-non-emergency-medical-transportation/
- National Institute of Standards and Technology. NIST AI 600-1, Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile. July 2024. https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf
- Koenecke, A., et al. Racial Disparities in Automated Speech Recognition. Proceedings of the National Academy of Sciences, 2020. https://pmc.ncbi.nlm.nih.gov/articles/PMC7149386
- Centers for Medicare & Medicaid Services. Expanded Report to Congress: Non-Emergency Medical Transportation in Medicaid, 2018–2021. https://www.medicaid.gov/medicaid/benefits/downloads/nemt-rtc-2018-2021.pdf
- Federal Communications Commission. FCC Confirms that TCPA Applies to AI Technologies that Generate Human Voices. Declaratory Ruling FCC 24-17, released February 8, 2024. https://www.fcc.gov/document/fcc-confirms-tcpa-applies-ai-technologies-generate-human-voices
- US Department of Health and Human Services. Guidance on HIPAA and Cloud Computing. https://www.hhs.gov/hipaa/for-professionals/special-topics/health-information-technology/cloud-computing/index.html
- 31 U.S.C. 3729, False Claims. https://www.law.cornell.edu/uscode/31/usc_sec_31_00003729—-000-.html
- California Legislature. AB 3030, Health Care Services: Artificial Intelligence. https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240AB3030
- Crowell & Moring. HHS Applies Discrimination Prohibitions to Use of Automated and Non-Automated Patient Care Decision Support Tools. May 8, 2024. https://www.crowell.com/en/insights/client-alerts/hhs-applies-discrimination-prohibitions-to-use-of-automated-and-non-automated-patient-care-decision-support-tools