Illustrative engagement across five service pillars and five industries, reflecting representative Meruksha engagement patterns.
Industry: Health Insurance / Payer
Client Profile
The client is a regional health insurance payer covering approximately 2 million lives across commercial, Medicare Advantage, and Medicaid managed care lines of business. Several business units had independently begun piloting AI tools to support prior authorization review and claims triage over an 18-month period without central coordination.
Challenge
An internal audit uncovered that four different business units had deployed four separate AI tools touching prior authorization and claims triage decisions, each procured and implemented independently with no shared governance framework, no documented testing for bias in coverage determinations, and no consistent record of how or why an AI-assisted recommendation had influenced a given decision. This created serious exposure at a moment when state insurance regulators nationally were sharpening scrutiny of AI use in coverage and claims decisions, with several states enacting or proposing rules requiring disclosure, human review, and bias auditing for algorithmic decision-making in health coverage. If a denied claim were challenged and the payer could not clearly document how an AI recommendation was generated, validated, and reviewed, the organization faced both regulatory and reputational risk, along with potential member harm if any of the four tools carried undetected bias.
Approach
Meruksha designed and implemented an enterprise AI governance program aligned to the NIST AI Risk Management Framework and explicitly mapped against emerging state-level insurance AI regulations relevant to the payer’s operating states. The engagement began with a full model inventory, formally cataloguing all four existing AI tools, their data inputs, decision scope, and current level of human oversight. Each tool was then subjected to a bias and explainability testing protocol specifically designed for coverage-adjacent decisions, examining outcomes across demographic and geographic segments for disparate impact. A cross-functional AI governance committee was established with clear ownership, escalation paths, and authority to approve, restrict, or decommission AI tools going forward. Based on the testing results, one tool was decommissioned outright due to insufficiently documented bias testing, two were retained with expanded human review requirements, and one was cleared for continued use with only minor documentation updates.
Outcome
At its next scheduled state regulatory examination, the payer passed with zero AI-related findings. A peer organization operating in the same state market, by contrast, received a formal regulatory inquiry that same year over undocumented AI use in claims determinations — a direct illustration of the exposure the payer had proactively closed. The governance framework Meruksha built is now referenced directly in the payer’s public AI transparency statement, and the governance committee continues to review any new AI use case before deployment rather than after the fact.