Revolutionizing Claims Operations With AI and Agentic Workflows

Errors, inefficiencies, manual handoffs, workarounds, and silos all describe traditional claims processing. Each also represents an obstacle to optimal payer performance as health care’s reimbursement models change and regulatory requirements increase. Most payers struggle to achieve high accuracy and auto-adjudication rates under fee-for-service models. Bundled pricing, outcomes-based reimbursement, and reimbursing for nontraditional services—such as transportation and food shopping—will make claims operations even more complex.

Fortunately, accelerating advances in artificial intelligence (AI) tools and technology can help payers significantly enhance claims operations relatively quickly and cost effectively. Autonomous AI agents and multiagent workflows are 2 key developments enabling payers to revamp current operations while building flexibility for adapting as health care evolves continues to shift toward outcomes-based models.

AI Agents and Agentic Workflows

An AI agent is an AI-enabled digital worker trained to make human-like decisions and continuously learn from feedback. Agents are built using AI models, business rules, and data. In claims operations, AI agents may be trained in prior authorization, edits, adjudication, appeals and grievances, coordination of benefits (COB), credentialing, and other specialized functions and processes.

When multiple AI agents work together in real time, they form an agentic workflow. Agents in an agentic workflow behave like a team of skilled claims professionals, collaborating on tasks such as data extraction, policy verification, member validation, provider validation, pend resolution, COB validation, and payment integrity checks. They share information faster than humans, enabling rapid issue resolution.

AI agents and agentic workflows enable payers to infuse their claims operations with a range of next-generation capabilities, including the following:

Intelligent prioritization and auto-adjudication: Agentic workflows dynamically route claims through the most appropriate channels based on factors such as accuracy, complexity, urgency, and more. For example, an agentic workflow can verify member data, check edit codes, and evaluate the claim’s odds of successful adjudication. If an AI agent is not confident about a claim’s accuracy, it will route the claim to a human examiner for review.

AI agents also can predict and resolve potential pended claims during adjudication, perform real-time claim edits, use member life events and claims data to predict potential COB, flag potential appeals and grievances, and check for over- and underpayments.

Adaptive claims operations: Agentic workflows supercharge claims processing, making it dynamic and essentially teachable. Payers can swiftly adapt to new regulations or launch new lines of business using agentic AI capabilities. An agent built on generative AI can digest and summarize new information for other AI agents in just a few minutes, such as a new contract or a regulatory policy. The agents in a workflow can then adjust procedures as appropriate to comply. Human professionals can also give new instructions in plain language to AI agents equipped with natural language processing capabilities.

Seamless integration across functions and applications: Agentic workflows may also span different systems and applications, orchestrating data sharing across core administrative platforms, proprietary systems, quality systems, customer relationship management applications, etc.

This enables payers to automate more complex processes, such as prior authorizations and appeals. For example, an agentic workflow can review and triage prior authorization requests. AI agents can validate the request against member and provider records, members’ benefit plans, payer guidelines, and standard operating procedures. The agentic flow can summarize medical notes and present that information and recommended actions to clinical reviewers. If additional supporting information is required, the agentic workflow generates an appropriate letter to the provider or routes complicated requests to human agents. One payer has reported spending 46% less time reviewing authorizations via an agentic workflow.

Continuous learning: Agentic workflows improve daily as they learn more about a health plan’s specific mix of benefits and members. Each agent in the workflow learns from each processed claim, authorization, and appeal, continuously improving decision-making and efficiency.

Time to Transform With AI and Agentic Workflows

Traditional claims processing is increasingly ill-suited to the demands of modern health care. AI agents and agentic workflows enable health plans to transform claims operations today to reduce administrative costs, improve payment accuracy, and deliver superior experiences to members. Agentic workflows will enable payers to adapt to new payment models, rules, and requirements almost as swiftly as they emerge. Payers that adopt AI-driven workflows will now have the flexibility and power to meet the future’s demands as they unfold.

About the Author

Deepan Vashi is the EVP & Head of Solutions for Health Plans and Healthcare Services at Firstsource with over 27 years of experience in health plan IT, business operations, and consulting. He is renowned for his expertise in developing member-centered digital solutions and building cross-functional teams to ensure successful implementation. In his role at Firstsource, he spearheads solutions and strategy for health plans, including Intelligent Back Office, Health Tech Services, and Platform-based Solutions (BPaaS). Deepan has extensive knowledge of innovative technologies such as Process Mining, Digital Twin, AI, and Blockchain.