What changed
FACT: CMS/HHS published a final rule, 'Medicaid Program; Community Engagement Requirement for Certain Individuals' (Federal Register, 2026-06-03, corrected 2026-06-29), compelling a defined class of Medicaid beneficiaries to document qualifying activities (work/volunteer/education) or lose coverage, and requiring every state to build verification/reporting systems. FACT (separate source): Google released computer-use/agentic browser control in a cheap, low-latency Gemini 3.5 Flash tier, making per-person UI automation economically plausible at population scale. HYPOTHESIS: those two make per-member automated evidence collection viable.
Why now
The rule creates a compelled-filer class with a deadline (FACT), and the cheap agent tier is newly available (FACT). CAUTION/HYPOTHESIS: Medicaid work/community-engagement requirements have a long history of litigation and injunctions; a final rule of this shape carries real risk of being stayed or vacated, which would evaporate the whole demand thesis. The provided documents list EFFECTIVE 'n/a', so the actual compliance clock is unverified here.
Converging signals
Three signals meet: (1) a federal mandate forcing beneficiaries to document qualifying activities, (2) fragmented pay/volunteer/state portals where that evidence lives, (3) a cheap agentic browser tier able to collect it per person. Convergence is real but the monetizable buyer (the MCO) sits one step away from the compelled party (the member).
Customer pain
HYPOTHESIS (not evidenced in input): MCOs lose per-member capitation revenue on every disenrollment, so coverage-loss churn is a real cost center. This is plausible and industry-known but NO figure, complaint, job posting, or existing-spend evidence was provided — demand on the buyer side is inference, not fact.
Who pays
Stated: Medicaid managed-care organizations (MCOs) and enrollment navigators. The member is the beneficiary, NOT the buyer. Selling to MCOs means selling into large, HIPAA-regulated health plans with procurement, security review, and BAAs.
Solved today
States build verification systems; MCOs use care-management/outreach staff, existing member-engagement vendors, and navigators. Members self-report through state portals. No evidence provided of a dedicated automated evidence-assembler.
Why current solutions are bad
HYPOTHESIS: manual outreach doesn't scale to the caseload and members drop for paperwork reasons. Reasonable but unproven from the input.
Proposed product
A consented agent that (with member authorization) logs into pay/volunteer/education portals, extracts qualifying-activity proof, runs a deterministic checker mapping evidence to the rule's criteria per state, and emits a gap report + pre-filled state-form packet. Sold white-label to MCOs/navigators as a churn-prevention layer.
MVP version
Single-state, single-MCO design-partner pilot: consent/authorization flow, one state's qualifying-criteria checker, agent connectors for the 3-4 most common evidence sources, gap-report + form-packet output. Run the KILL TEST: 20 real members end-to-end; measure human-takeover rate.
30-day build
Confirm the rule's actual status and effective dates; map ONE state's qualifying criteria and verification/reporting spec; secure a signed MCO or navigator design-partner and a BAA/data-handling plan BEFORE touching real member PII; build the deterministic checker on synthetic data.
60-day build
Build consented agent connectors for top evidence portals; run the 20-member kill-test cohort; measure straight-through rate. If human takeover is heavy, stop.
90-day revenue plan
Convert the design partner to a paid per-member-per-month or per-packet pilot contract; document churn-prevention ROI to open a second state/MCO. Realistically first revenue is a healthcare-procurement pilot, not day-90 cash.
Distribution path
Direct to MCO care-management/retention leaders and to enrollment-navigator organizations; industry associations. This is relationship/enterprise-adjacent healthcare selling, not self-serve card payment.
Pricing hypothesis
Per-member-per-month retention fee or per-completed-packet; pilot as a fixed monthly contract with a design partner.
Technical difficulty
High. Reliable consented browser automation across many heterogeneous, auth-gated portals at production reliability is hard; per-state rule mapping multiplies scope; handling member PII/pay data demands security controls.
Legal / regulatory risk
High. Handling Medicaid-member PII and financial documents implicates HIPAA/BAA obligations and state privacy law; automated login on the member's behalf and consent scope must be airtight. The underlying rule may itself be litigated/enjoined.
Platform dependency
Government/state portals are not a deplatform risk, but reliance on the Gemini Flash computer-use tier and on portals that can change their UI or block automation is a real dependency.
Founder fit
Mixed. The government-mandate/forced-filer SHAPE matches his proven FMCSA edge and primary thesis — but the buyer here is a large, HIPAA-regulated health plan (enterprise healthcare procurement + BAAs + medical-adjacent data), which directly hits his stated avoids: heavily regulated medical products and deep enterprise/relationship sales. His edge is filing FOR the compelled party per transaction; here the compelled party (member) isn't the payer, breaking the pattern.
Breakout potential
If the rule stands and one state/MCO works, 50-state replication is large. But that upside is fully contingent on the rule surviving litigation and on the pilot clearing the human-takeover kill test.
Final recommendation
REVISIT / RESHAPE, don't build as specified. The mandate is a genuine forced-filer signal, but the MCO buyer makes this enterprise healthcare sales with heavy HIPAA/PII exposure — a poor fit for this founder despite the government-portal shape. If pursued, reshape toward the compelled party's own agents (enrollment navigators, community benefit orgs, or a member-facing consented tool) where the founder can sell demonstrated value, and gate everything on (a) confirming the rule's legal status/effective date and (b) passing the 20-member human-takeover kill test before writing production code.
Next action
Verify the rule's current legal status and effective/verification dates on Federal Register and any pending litigation; only if it is live and unenjoined, approach 2-3 enrollment-navigator orgs (lower procurement bar than MCOs) to test willingness to pay before any build.