AI readiness within Medicaid Enterprise Systems Modular Transformation
Sierron provides AI and modernization advisory in support of the Maryland Department of Health's Medicaid Enterprise Systems Modular Transformation, focused on the data foundation and governance conditions that determine where artificial intelligence can be applied responsibly.
Program context
The Maryland Department of Health administers Medicaid for approximately 1.5 million Marylanders, close to a quarter of the state’s population. The Medicaid Enterprise Systems Modular Transformation, known as MMT, is a multi year program replacing a legacy Medicaid Management Information System more than three decades old, structured according to the modular approach the Centers for Medicare and Medicaid Services requires of states. CMS provides oversight and the majority of program funding.
Modular transformation changes the nature of the technical problem. Rather than a single replacement system, the program delivers a set of independently procured and certified modules that must interoperate, satisfy CMS certification criteria, and align to the Medicaid Information Technology Architecture framework. Program progress has included migration of MEDFIN files to MD THINK document management and onboarding of EDITPS and LTSS with EDI claims processing in both real time and batch modes.
Maryland has also established a formal position on artificial intelligence in state government. Executive Order 01.01.2024.02, issued in January 2024, created the Maryland Artificial Intelligence Subcabinet and directed development of a statewide AI action plan. The Maryland Artificial Intelligence Governance Act took effect in July 2024. The 2025 Maryland AI Enablement Strategy and AI Study Roadmap, submitted to the General Assembly, identifies strengthened data foundations and sector specific governance frameworks among its stated priorities.
Challenge
A modular program presents a data problem before it presents an artificial intelligence problem. Modules are procured separately, implemented on different schedules, and maintain their own data structures. Information describing the same beneficiary, provider, or claim exists in several places, each accurate for its originating purpose, with differences that remain invisible while experienced staff mediate between systems.
Medicaid data carries protected health information and personally identifiable information, with handling requirements that constrain where and how processing may occur. Eligibility and claims determinations affect access to care, which places any system influencing those determinations in a category demanding explainability and retained human authority.
The practical difficulty is that artificial intelligence initiatives in this environment frequently begin before authoritative sources are resolved, and a retrieval system exercises no judgment about which record is correct.
Sierron’s role
Sierron leads artificial intelligence readiness for the MMT program, assessing the data foundation, prioritizing use cases, and aligning governance to CMS requirements and the State’s published AI strategy.
Approach
Sierron directs early artificial intelligence effort toward operational work rather than beneficiary facing capability. Operations within a Medicaid program, including claims review support, correspondence handling, document intake, provider record maintenance, and case triage, present four conditions that materially improve the probability of a successful result.
The task recurs at volume, so improvement compounds. Documented procedure establishes a definable correct answer against which performance can be measured. Staff review precedes consequence, which contains the failure mode. Baselines already exist, because operations groups are managed on throughput, cycle time, and backlog.
The fourth condition determines whether an initiative survives its first funding review. Programs unable to demonstrate a before state struggle to justify continuation regardless of the improvement produced.
Ahead of use case selection, Sierron assesses the data foundation. In a modular environment, authority must be resolved before retrieval architecture inherits ambiguity. Establishing which module governs a given data element is a governance decision rather than a technical one, and organizations can defer it indefinitely until an artificial intelligence initiative forces the question.
Access control receives specific attention. Conventional controls assume users reach data through an application enforcing authorization rules. Retrieval based approaches can bypass that enforcement, surfacing content from multiple modules to a user who would not have reached it through the originating system. Authorization must be enforceable at retrieval time and must satisfy protected health information handling requirements.
Sierron establishes, in advance, which determinations will not be delegated to or influenced by automated systems. Documenting that boundary explicitly prevents more downstream difficulty than subsequent testing resolves.
Outcomes to date
The engagement is ongoing. Work delivered to date includes:
- Prioritization of operational use cases with existing baselines ahead of beneficiary facing pilots
- Identification of authoritative source conflicts across modules before inheritance by retrieval design
- Definition of protected health information boundaries within proposed artificial intelligence workflows
- Documented eligibility and claims determinations that remain human decisions
- Alignment of approach to Maryland AI Enablement Strategy priorities on data foundations and sector specific governance