Handing Off the Medicaid Enrollment Maze to an AI Platform That Does It for You

AI-powered support for Medicaid self-direction enrollment can help family caregivers find programs that pay relatives for home care and work through the paperwork to enroll. With new requirements arriving in 2027, its most important role is preventing eligible families from losing benefits through administrative failures. Families need end-to-end assistance with human oversight.

Why Medicaid enrollment is about to get significantly harder for families and caregivers

The One Big Beautiful Bill Act (OBBBA), signed July 4, 2025, sets dates for two changes to the rules. Work requirements begin January 1, 2027, and Medicaid expansion adults must renew twice a year starting December 31, 2026. Expansion adults, who are generally ages 19–64 and enrolled through the ACA expansion, used to renew once a year. These changes compound each other. Families will face more frequent reviews while also needing to establish work compliance or an exemption.

University of North Carolina health policy and social medicine professor Jonathan Oberlander warns that people who work and remain eligible can nonetheless lose coverage because of administrative complexity. CMS projects that a substantial share of projected coverage losses will result from failures to submit correct documentation. Eligibility alone will not protect an enrollee.

That risk becomes more serious because losing Medicaid does not necessarily open another affordable route to insurance. People who lose coverage for failing to meet work requirements are generally blocked from receiving marketplace premium subsidies. A documentation failure can therefore lead into a coverage gap with no easy exit. For families already managing care, the new schedule makes administrative follow-through part of protecting access to treatment and support.

Why most projected coverage losses will be paperwork failures, not eligibility failures

Administrative attrition describes the specific failure at issue. A person can qualify for coverage yet disappear from the rolls because a notice never arrived, a form was confusing, or a response missed its window. The Commonwealth Fund’s 2025 analysis found that most Medicaid disenrollments are procedural, including these failures among members who remain eligible.

Weill Cornell’s JAMA Health Forum Special Communication cites earlier research that found some eligible Arkansas enrollees lost coverage under a similar work requirement because they had trouble with the paperwork. The federal rule brings those same conditions to the whole country. The renewal cycle following the end of pandemic-era continuous enrollment protections also showed how unevenly administrative barriers fall. Children’s churn rose across racial and ethnic groups, with Hispanic children seeing the largest increase because they faced more barriers to keeping coverage, even though rates of ineligibility were no higher.

Caregiver exemptions do not eliminate this failure mode. Family caregivers responsible for children ages 13 and under, or disabled individuals of any age, are exempt from the work requirement, but they must document and claim that exemption. The protection itself requires another administrative step. Dr. Beth McGinty of Weill Cornell describes the difficulty plainly. “It's already very complicated to figure out how to apply for Medicaid and stay enrolled. It differs by state. Forms are confusing, and this new requirement creates additional administrative complexity.”

An exemption that exists in law cannot prevent disenrollment if the evidence needed to establish it never reaches the agency. Assistance must address that gap between entitlement and documentation.

What family caregivers stand to lose if their Medicaid coverage lapses

Ellie left her job when she could no longer balance employment with full-time care for her mother Jane, who has Alzheimer’s, and her father. Medicaid’s HCBS program paid her to become her mother’s full-time caregiver. If the coverage that pays for that care ended, so would her income. Her story shows why enrollment is not solely a question of access to medical appointments.

Corey Briskey in North Carolina faced a different consequence of unsupported caregiving. After years without support, he developed food sensitivities, eczema, and debilitating brain fog. Medicaid coverage let him hire respite care. Briskey described the change in concrete terms, saying “the day the respite worker arrived, I felt something I hadn't felt in years: relief”. A lapse can threaten that relief as well as the financial support that makes it possible.

All reporting states except Alaska allow Medicaid enrollees to self-direct home care in at least some circumstances, and all allow enrollees to select, train, and dismiss their own caregivers. All responding states pay family caregivers under some circumstances and offer other support, including respite care. The program’s financial management agency pays caregivers, usually as W-2 income, though some programs issue a 1099 instead.

CMS lists several ways to get self-directed services, including the Home and Community Based Services (HCBS) State Plan Option, the Community First Choice Option (CFCO), and the Self-Directed Personal Assistance Services (PAS) State Plan Option. Payment through these program pathways depends on the relative’s eligibility. Families must keep their coverage and continue to qualify for the program. Nearly half of U.S. family caregivers already say caregiving has hurt their finances, and many have cut their work hours or left jobs. For those households, pay for care and help with respite costs may keep an already tight budget afloat.

How the enrollment maze defeats people before AI enters the picture

Careforce CEO Huzaifa Sial identifies the underlying problem. “Most people don't know what they need, and if they do, they have a hard time getting there. That's the hidden execution problem that nobody sees.” His distinction separates discovery from completion. A family must first recognize an available benefit, then navigate the process that turns it into actual support.

State variation makes both tasks harder. Programs include CDASS in Colorado, IRIS in Wisconsin, CDPAP in New York, and CFSS in Minnesota, with different names, forms, and processes. Knowing the term “self-direction” does not identify the right application. Timelines typically run several months from the start of an application to the first paycheck, so even successful navigation demands sustained attention.

KFF’s focus groups found that family caregivers often struggled to make ends meet and reduced work hours because of caregiving demands. Those demands also leave less time to apply for benefits. Paid caregiver wages, respite funding, and medical equipment coverage can go unclaimed because establishing eligibility becomes a second job. The process serves agency administration more readily than an applicant’s limited time.

A large majority of caregivers report burnout, while a substantial share have no support from friends or family. This is not simply a need for clearer explanations. The structural gap is between knowing what must happen and having enough time, support, and continuity to make it happen.

What AI-assisted enrollment actually does in practice, and where it already works

Kern Family Health Care, the largest Medi-Cal provider in Kern County, deployed “Angelica,” a conversational AI program from Oakland startup Careforce. Angelica schedules appointments, reminds members to finish paperwork, reschedules around conflicts, and communicates in a member’s preferred language. Human staff fill out and check applications at the appointments it helps set up.

Vanessa Barahona, 41, completed her renewal with Angelica’s assistance. When an appointment conflicted with her cleaning job, Angelica rescheduled it and communicated with her in Spanish. Barahona said “It felt like I was talking to a real person”.

Kern Family spent approximately $370,000 on the software. Making hundreds of thousands of calls to members would otherwise have required substantially more spending on staff. Careforce is also working with Central California Alliance for Health in the same capacity. These deployments demonstrate an operating outreach-and-renewal model, though they do not establish that every step of Medicaid self-direction enrollment can already be automated.

Weill Cornell’s communication describes a complementary agency-level role. AI could link enrollment records with payroll, tax, or other public-program data to show that members meet the work requirement or qualify for an exemption without asking them for more documents. It could also analyze call center transcripts, help desk messages, and website activity to identify where applicants struggle. That shifts some work away from the applicant.

The same communication describes digital assistants that explain requirements and needed documentation, as well as analysis of anonymized call transcripts to help agencies adapt outreach week by week. About a quarter of state Medicaid programs already use AI chatbots for consumer assistance. These capabilities serve different purposes. Neither has been shown to handle every step of enrolling caregivers for compensation without human help.

Where AI-assisted enrollment has real limits and why human oversight remains essential

Identifying someone at risk of losing coverage does not explain every reason they cannot finish a renewal. Transportation gaps, housing instability, language barriers, and distrust can defeat an otherwise effective reminder. Automation cannot resolve those conditions by itself. A platform may communicate in a preferred language, as Angelica did, without being able to address the full circumstances behind a missed appointment.

Bias creates a separate risk. McGinty warns that biases in existing data can be reproduced through AI and calls for human involvement, careful monitoring, and oversight. That objection reaches beyond whether an assistant gives a correct answer. Systems used to identify people for help or route their cases can reproduce disparities already present in the information they use.

Stanford economics professor Mark Duggan also worries that insurers could use such software to cherry-pick patients. Duggan puts the oversight requirement directly, saying “When you have a new technology like this, you need to police it”. The purpose and use of the software deserve scrutiny as much as its technical performance.

Access will also vary by state. Weill Cornell researchers argue that federal assistance will be vital for states with less developed AI capabilities and IT infrastructure. The benefit cannot be assumed to be available everywhere. Kern Family’s arrangement illustrates the more credible design, with AI handling navigation and scheduling while human staff complete and verify applications. At the scale the 2027 requirements demand, a well-designed human-AI hybrid that preserves human judgment is better suited than unaided human navigation.

How can family caregivers use AI enrollment platforms to protect coverage and get paid through Medicaid?

For families preparing now, the practical step is to move as much enrollment and exemption work as possible onto an assistance platform built to support completion. The starting questions are which coverage needs protection, which state program applies, and whether the caregiver qualifies for an exemption. Caregivers responsible for children 13 and under or disabled individuals of any age should document and assert that exemption proactively. An exemption left undocumented is an unfinished task.

Caregivers then need to find the self-directed care program that applies where they live and check whether a program they found online is offered locally. State examples include California IHSS, New York CDPAP, Ohio PASSPORT, and Texas STAR+PLUS, with differing names, structures, eligibility rules, and pay rates. Applications should begin well ahead of the late-2026 transition because reaching a first paycheck typically takes several months. December 31, 2026, is the effective date for the renewal-frequency change, with renewal deadlines varying by family.

For caregivers and home care agencies comparing software, the clearest criteria are benefit discovery, documentation and submission support, accessible multilingual outreach, and human escalation. These are distinct capabilities. A tool that surfaces an unfamiliar caregiver compensation program solves discovery, while one that guides documentation and submission tackles execution. A chatbot that answers questions but stops there does not provide end-to-end enrollment support.

The named examples offer a narrower comparison than a vendor ranking would imply. Careforce has a concrete outreach-and-renewal deployment with human application support. State Medicaid chatbots provide consumer assistance, while Weill Cornell’s agency-level proposals describe ways to reduce paperwork through data linkage and better navigation. The evidence does not establish a best vendor for complete Medicaid self-direction enrollment.

The appropriate platform is therefore one that can reveal relevant Medicare, Medicaid, and caregiver compensation programs, then guide the family through enrollment. Families should look for help that stays with the application and recognizes when a person must take over. Handing off paperwork is useful only when responsibility for completing it remains clear.