PLAN TYPE
Regional Health Plan
PLAN SIZE
350K to 400K members
GEOGRAPHY
Northeastern U.S.
80%
Claims queues automated
95%
Agentic AI accuracy
>99%
Human-in-the-loop accuracy
10-15%
Shorter cycle times with zero added headcount
Issue
A regional health plan faced an uphill battle with claims processing across commercial and Medicare Advantage lines of business (LOBs) due to fragmented intake processes and inconsistent document formats. Claims teams had to extract and verify data manually, only to spend additional time reconciling member, provider, and clinical information across assorted multipage documents. This extra work led to quality and member and provider experience risks.
Seeking to scale claims processing capacity, client leaders enlisted Sagility to explore the potential of Straight Through Processing (STP), the “gold standard” of end-to-end (E2E) automation. They requested a solution that would:
- Maintain auditability and trust
- Ensure strong compliance and human oversight
- Guarantee data and information security
- Maintain or exceed quality and production metrics
- Demonstrate resilience under changing business or process variables
Action
Sagility’s STP intervention was two-pronged. Early diagnostics using AI Agents revealed that potential E2E automation could stall on “last mile” nuances, such as referencing providers by logos, splitting invoices across pages, and mismatching names and dates.
To eliminate these hurdles, Sagility’s team combined Azure OCR, the “eyes” that read printed or handwritten text out of an image and the client-approved GPT model, which served as the “brain” that formulates answers from the extracted data. The result: a modular, API-ready architecture that could be integrated methodically into the client’s existing IT system without compromising compliance or human oversight. The structured execution plan allocated responsibilities across the client’s LOBs and Sagility for document access, user interface (UI) finalization, whitelisting, dashboards, and posting of AI-aligned rules.
Impact
Sagility’s structured plan moved from proof-of-concept to production in 5 months, at which point most in-scope claim queues could flow straight through the system. Claims reviewers focused on nuances related to clinical, contract, eligibility, and other high-level activities, while Agentic AI handled repetitive data extraction, validation, and rule-based processing.
Quantitative outcomes:
- Approximately 80% of the in-scope claim queues were automated, with high-volume categories reaching around 85% and complex categories about 76%
- More than 95% Agentic AI accuracy with a human-in-the-loop model ensured >99% reliable data extraction, enabling faster and more accurate downstream claims processing
- Shorter E2E cycle times, 10% to 15%, resulted without increases in headcount
- Modular, API-ready architecture accelerated time to deployment and scale-up to other processes and LOBs
Qualitative outcomes:
- Links to the source of each extracted field improved claim reviewer trust, reducing back-and-forth queries
- Post-AI rulesets provided a clear audit trail and stable, versioned decision logic
- Claim reviewers could more easily accept, edit, or pend claims via UI features showing claim context, line-level details, and deep links
- Dashboards showing production, exceptions, rule hits, and claim details enhanced QA capability
- Failure modes were called out and routed to targeted fixes for continuous improvement
The value delivered exceeded client expectations of STP by boosting processing speed and accuracy through Agentic AI-enabled automation; reinforcing enterprise readiness with built-in compliance and observability; and prioritizing human expertise for complex, context-heavy decisions.
