AI Outbound Campaigns for State Benefit Recertification: Reducing Medicaid Churn

The Medicaid Unwinding in Numbers

The Families First Coronavirus Response Act kept Medicaid enrollment continuous from March 2020 through March 2023. When that continuous enrollment condition lifted on April 1, 2023, every state had to re-run eligibility redetermination on every enrollee for the first time in three years. The result was the largest single coverage event in Medicaid's history.

CMS and KFF's unwinding tracker have logged more than 25 million Medicaid and CHIP disenrollments through the reporting period that closed out the bulk of unwinding case action. Roughly 69 percent of those disenrollments were procedural - meaning the state could not confirm whether the person was still eligible because a form was not returned, a phone call was not answered, or the address on file was wrong and the renewal packet never reached the enrollee. State-by-state procedural rates have ranged from below 40 percent in the best-performing states to over 85 percent in the worst.

The churn cost is not one-time. KFF's longitudinal analysis shows that 40 to 60 percent of procedurally disenrolled beneficiaries re-enroll within a year, meaning states absorb the administrative cost of a full new eligibility determination for people who never should have lost coverage in the first place. MCOs lose continuity with members. Providers lose Medicaid revenue on already-scheduled care. And enrollees experience gaps that translate directly into missed prescriptions, deferred chronic disease care, and avoidable ER visits.

Unwinding is officially complete as a policy moment, but redetermination itself is now a permanent annual process. Every state is operating under CMS's 2024 eligibility and enrollment final rule, which tightened procedural disenrollment guardrails, required ex parte renewal attempts, and added operational reporting that keeps the spotlight on procedural termination rates. The problem did not end with unwinding. It became structural.

Procedural Disenrollment Is the Fixable Problem

Eligibility-based disenrollment is a policy outcome. If a person's income has risen above the Medicaid threshold, the state has to act on that finding. That is not a problem for AI to solve.

Procedural disenrollment is different. It is the failure of an administrative process. It is what happens when:

• The renewal packet goes to an address the enrollee left two years ago.
• The enrollee never opens the envelope because they assume it is junk mail.
• The enrollee opens it, intends to respond, and misses the deadline because work and caregiving got in the way.
• The form is returned incomplete and the follow-up call goes to a disconnected number.
• The state's eligibility call center has a 45-minute hold and the enrollee hangs up.
• The ex parte renewal fails and nobody tells the enrollee.

Every one of those failure modes is a touchpoint problem. The person exists, the person is (likely) still eligible, but the state cannot close the loop. State Medicaid directors have told CMS and their legislatures directly that the bottleneck is staffing and contact-rate, not policy. That is exactly the problem AI outbound voice campaigns are engineered to solve.

How an AI Recertification Campaign Actually Runs

A well-designed AI recertification campaign is not a robocall blast. It is a scheduled, personalized, compliance-aware outbound workflow that moves each enrollee through the redetermination funnel until they either complete renewal, confirm they are no longer eligible, or request a human caseworker.

1. Redetermination due-date trigger. The state's MMIS or integrated eligibility system publishes a list of enrollees whose renewal month is upcoming (usually 60, 45, 30, 14, and 7 days out). The AI platform ingests the list via secure SFTP or API and builds the call queue.
2. Ex parte attempt first. Before AI places an outbound call, the state runs its automated ex parte renewal using wage, SSA, and state data sources. AI is only dispatched to enrollees whose ex parte did not complete - the exact population most at risk of procedural loss.
3. Permissioned outbound call. AI places the call using the state's published caller ID, identifies the program ("This is an automated call from the [State] Medicaid program about your annual renewal"), and asks consent to continue in the enrollee's preferred language.
4. Identity confirmation with minimum necessary data. AI verifies identity using the minimum HIPAA-compliant data elements (date of birth, last four of case ID, or ZIP plus one knowledge-based check). AI never reads sensitive data out loud to an unverified caller.
5. Guided renewal conversation. AI walks the enrollee through renewal in conversational form: address verification, income change, household composition change, other insurance. Questions are phrased in plain language, not MAGI jargon.
6. Channel choice. If the enrollee prefers to complete the full form online or on paper, AI texts the portal link, confirms the mailing address, and schedules a follow-up call. If they want to finish by phone, AI completes the allowable data capture and submits directly to MMIS where program rules allow.
7. Warm transfer for complex cases. Any case that requires caseworker judgment (disability determination, long-term services and supports, appeal, hardship) transfers to a live eligibility worker with full context.
8. Confirmation and follow-up. AI confirms the submission or next step, texts a written confirmation, and schedules reminder calls before the deadline if needed.
9. Writeback and reporting. Every call outcome - contacted, voicemail, wrong number, completed, transferred, opted out - writes back to MMIS and into the state's unwinding/redetermination dashboard in real time.

Call Types AI Handles End-to-End

Pre-Deadline Reminder Calls

The highest-volume call type. Enrollees 30, 14, and 7 days from redetermination deadline who have not yet responded. AI confirms the deadline, offers to complete renewal on the call, or routes to the online/mailed form. A single well-run reminder pass typically raises renewal completion by 15-30 percentage points on the reached population.

Address and Contact Update Calls

Many procedural disenrollments are simply address-staleness. AI places targeted calls to verify and update address, phone, and preferred language before the renewal packet mails, preventing return-to-sender failures.

Missing-Information Follow-Up

Renewal form was returned but is incomplete. AI calls to collect the missing fields (income verification, household composition, proof of residency), confirms identity, and submits the completed record to the caseworker queue.

Ex Parte Failure Outreach

The state's automated ex parte renewal did not complete. AI reaches out to gather only the specific data elements the ex parte could not verify, closing the gap without forcing the enrollee through a full 20-page renewal form.

90-Day Reconsideration Outreach

CMS requires states to offer a 90-day reconsideration period after a procedural termination. AI runs outbound campaigns to procedurally disenrolled individuals within that window, reinstating coverage without requiring a new application when appropriate.

Managed Care Plan Continuity Outreach

When an enrollee's MCO assignment is changing or a new plan year requires confirmation, AI handles the plan-choice call, explains options, and routes enrollment decisions back to the state enrollment broker.

Continuous Eligibility Confirmation (Children, Adults)

States operating 12-month continuous eligibility for children (now required under federal law) or continuous eligibility waivers for adults use AI to confirm continued enrollment and surface changes that would affect eligibility mid-year.

Inbound Overflow for Renewal Questions

When the state's eligibility call center is on hold, AI answers the overflow, handles straightforward renewal status questions, schedules a callback with a live caseworker, or completes the renewal in-line.

Multilingual Outreach

AI places outbound calls in the enrollee's registered preferred language - Spanish, Mandarin, Vietnamese, Haitian Creole, Somali, Arabic, Russian, Tagalog, Hmong, and 50+ more. Language-line latency and interpreter cost drop to near zero on routine renewal calls.

Post-Disenrollment Reconnection

For enrollees who lost coverage procedurally, AI runs reconnection campaigns during the reconsideration window, explains the path to reinstatement, and, where applicable, confirms re-enrollment without a fresh application.

SNAP, TANF, and CHIP Renewal Campaigns

The same recertification pattern applies directly to SNAP, TANF, and CHIP - and in most states, these programs share an integrated eligibility platform, a shared caseworker pool, and overlapping enrolled populations.

• SNAP recertification. SNAP requires periodic recertification (typically every 6 or 12 months depending on household composition) plus an interim report at 6 months in many states. AI handles SNAP recertification interviews for straightforward cases where federal and state policy allows phone interviews, captures the required verification, and transfers complex cases to a live SNAP eligibility worker. SNAP's interview requirement varies by state; AI is configured to respect each state's interview policy.
• TANF recertification. TANF has stricter work-requirement reporting and case management touchpoints. AI handles routine reporting calls (hours worked, job search verification where applicable, continued eligibility confirmation) and routes any case-management conversation to the TANF caseworker.
• CHIP renewal. CHIP renewals often run on the same schedule as Medicaid and share the eligibility platform. AI coordinates the Medicaid/CHIP transition when a child ages or a family's income changes, confirms continuous eligibility where applicable, and prevents gap coverage that forces families back through full application.
• Cross-program renewal coordination. Many households receive Medicaid, SNAP, and CHIP simultaneously. AI can run a single combined renewal conversation touching all three programs where the state's integrated eligibility system supports combined renewal, dramatically reducing the number of calls and forms a family has to manage.
• Stakeholder outreach for block grant reporting. State TANF administrators facing federal reporting deadlines use AI-captured structured data to feed ACF TANF Data Report and TANF participation rate calculations without back-end chart scrubbing.

MMIS, Eligibility, and MCO Integrations

A recertification campaign only works if the outcome of every call writes back to the system of record. The enrollee confirms a new address on the AI call; the caseworker must see that address when they open the case ten minutes later. The enrollee asks for a paper form; the print queue must pick that up overnight. The enrollee transfers to a human; that human must open the case with the full call context.

• Medicaid MMIS. Gainwell (HealthInteractive), Conduent (Texas, Connecticut, Missouri, others), Optum (multiple states), DXC, CNSI (now Acentra Health in many states), HPE/Accenture-built MMIS. Standard integration is REST API for near-real-time, SFTP for batch, HL7 where the state publishes clinical eligibility segments.
• Integrated Eligibility Systems (IES). Deloitte (California CalSAWS, Colorado PEAK, Michigan Bridges, Florida ACCESS, others), Accenture (Texas TIERS, New York NYBEAS), Wipro, CGI, CNSI, custom state builds.
• SNAP / TANF / CHIP platforms. Many run on the same IES. States with separate SNAP platforms use AI integration specific to those systems.
• Medicaid Managed Care Organizations. MCO member services systems (UnitedHealthcare, Centene/WellCare/Ambetter, Anthem/Elevance, Molina, CareSource, Humana, BCBS plans) integrate with AI for plan-choice calls, continuity outreach, and plan-year transition campaigns.
• Enrollment brokers. Maximus, Public Consulting Group (PCG), Conduent, and state-operated brokers for plan selection calls.
• Case management and 1099 worker platforms. For LTSS programs with self-directed services, AI integrates with case management platforms (Therap, Sandata, CareBridge) to coordinate renewal across medical and LTSS eligibility.
• State 211 and community partners. AI warm-transfers to 211, navigator programs, and community-based organizations (CBOs) where the enrollee needs in-person enrollment help.
• Real-time dashboards. Every outcome feeds the state's redetermination dashboard, CMS monthly reporting extract (CMS-416, CMS-64, unwinding data), and the agency's own executive scorecard.

TCPA, HIPAA, and State Confidentiality

Outbound calling on behalf of a state benefit program sits at the intersection of telecom law, health privacy, and state administrative law. AI voice agents deployed for recertification run with all of the following controls on by default.

• TCPA and FCC government calling rules. Outbound calls made by or on behalf of federal, state, and local government agencies for benefit administration are treated under a distinct framework from commercial telemarketing. AI honors do-not-call preferences captured in the call or in the state's system, provides clear opt-out, and respects state calling-hour restrictions.
• HIPAA and state Medicaid confidentiality. Medicaid renewal calls involve protected health information. AI operates under a Business Associate Agreement with the state or its contractor, encrypts data in transit and at rest, applies minimum necessary disclosure, and authenticates the enrollee before sharing any PHI.
• 42 CFR Part 2 where applicable. For Medicaid enrollees whose case includes substance use disorder treatment records, AI respects 42 CFR Part 2 consent controls on any outreach that touches SUD-covered data.
• State benefit confidentiality laws. SNAP has federal confidentiality rules at 7 CFR Part 272; TANF under 45 CFR Part 205; state laws add additional protections in California, New York, Illinois, Washington, and many others. AI enforces the stricter of state or federal rules.
• FedRAMP and StateRAMP. AI platform hosted on FedRAMP-authorized cloud (AWS and Azure government regions available), with StateRAMP alignment for states requiring it.
• CJIS where routing touches law enforcement. Not typical in benefit recertification, but integrated systems that include child welfare or dependency casework may require CJIS-compliant handling.
• Accessibility. ADA, Section 508, and LEP (Limited English Proficiency) Executive Order 13166 obligations - AI provides IVR-bypass for callers who need TTY/RTT, auto-routes to the preferred language, and can hand off to a live ASL interpreter via video relay for in-person follow-up.
• Call recording, transcription, and retention. All calls recorded and transcribed per state retention schedule (commonly 3-7 years for benefit case files). Retention and destruction automated.
• Audit. Full call-level audit log exportable for CMS, state auditor, and HHS OIG review. Coverage of PII access and system actions is end-to-end.

What States Are Measuring

Two metrics matter more than any others: procedural termination rate (reported monthly to CMS) and renewal completion by deadline. Every other metric is instrumental to those two. An AI campaign that raises right-party contact rate but does not move procedural terminations has failed its actual purpose.

How to Procure This

AI voice agents for state benefit recertification can be procured through several pathways. The right one depends on the state's existing contract portfolio and timeline.

• MMIS or eligibility system modernization. When the state is re-procuring MMIS or its integrated eligibility system, AI voice can be scoped into the new contract as a subcomponent. Longest path but captures the largest transformation.
• Managed care contract amendment. MCOs can include AI outreach as part of care coordination or member services, funded under existing capitation. Faster than a state procurement.
• Enrollment broker contract. States with enrollment brokers (Maximus, PCG, Conduent) can amend to include AI outbound on renewal and reconsideration populations.
• State cooperative purchasing. Texas DIR, NASPO ValuePoint, and state cooperative contracts frequently carry cloud and contact-center line items that can host AI voice deployments. BetaQuick delivers Texas DIR work through partner Compass Solutions, LLC (DIR-CPO-6057, active through October 2030).
• Innovation or pilot procurement. Many states have established innovation procurement paths for technology pilots of 6-12 months, which then inform a full procurement.
• CMS Innovation and unwinding-extension funding. CMS has made enhanced FMAP and targeted technical assistance available for states working to reduce procedural disenrollment. AI outreach is an eligible category.
• SNAP/TANF federal funding streams. SNAP administrative match, TANF administrative funds, and SNAP E&T program dollars can fund AI where the use case is clearly tied to program administration.

Frequently Asked Questions