Over the last two decades, my experience developing sophisticated AI technologies-from university research to large-scale industry applications-has revealed the cyclical nature of AI breakthroughs. My career started amid the “AI Winter,” a period marked by heavy investments in expert systems that ultimately failed to meet expectations. Today, while large language models (LLMs) signify a remarkable advancement, their widespread use through prompt engineering often overpromises, essentially functioning as complex rule-based systems cloaked in natural language.
At Ensemble, a frontrunner in hospital revenue cycle management (RCM), we tackle the inherent shortcomings of LLMs by pioneering the next frontier in AI: integrating LLMs with neuro-symbolic AI to anchor outputs in factual accuracy and logical reasoning. Our dedicated AI incubator brings together top-tier AI researchers and healthcare specialists to create agentic AI systems that combine the intuitive strengths of LLMs with the rigor of symbolic logic and structured reasoning.
Addressing the Challenges of Large Language Models in Healthcare
LLMs are exceptional at grasping subtle context, performing instinctive reasoning, and generating human-like dialogue, making them ideal for intelligent agent applications that interpret complex data and communicate effectively. However, in healthcare-where precision, regulatory compliance, and adherence to clinical protocols are critical-symbolic AI plays a vital role. The healthcare domain is governed by extensive structured resources such as taxonomies, clinical guidelines, and regulatory frameworks that demand exactness beyond what LLMs alone can reliably provide.
Our hybrid AI architecture merges LLMs and reinforcement learning with comprehensive knowledge bases and clinical logic. This fusion not only enhances intelligent automation but also significantly reduces hallucinations, broadens reasoning capabilities, and ensures every decision aligns with established medical standards and enforceable safeguards.
Building a Robust Agentic AI Framework
Ensemble’s agentic AI strategy is built on three foundational pillars:
1. Comprehensive and High-Quality Data Assets
Managing revenue operations for hundreds of hospitals nationwide grants Ensemble access to one of the most extensive administrative healthcare datasets available. Our team has invested decades in data aggregation, cleansing, and harmonization, creating an unparalleled environment for developing advanced AI applications.
Our agentic systems leverage over 2 petabytes of longitudinal claims data, 80,000 denial audit letters, and 80 million annual transactions, all meticulously mapped to industry-leading performance outcomes. This rich data ecosystem powers our end-to-end intelligence platform, EIQ, which supports over 600 distinct steps in the revenue cycle with structured, context-aware data pipelines.
2. Synergistic Collaboration with Domain Experts
Our AI researchers work hand-in-hand with revenue cycle specialists, clinical ontologists, and data annotation teams throughout the innovation process. This close partnership ensures that AI solutions are designed with a deep understanding of regulatory requirements, payer-specific rules, and the intricate workflows of revenue cycle management.
End users embedded within the system provide continuous feedback post-deployment, enabling rapid identification of friction points and iterative improvements. This triad of AI scientists, healthcare professionals, and frontline users fosters a system that mirrors expert human decision-making while delivering the speed, scalability, and consistency of AI, all under vigilant human supervision.
3. World-Class AI Talent Driving Innovation
Our research incubator attracts AI experts typically found only in leading technology firms. Our team includes PhD and MS graduates from prestigious institutions such as Columbia University and Carnegie Mellon University, with extensive experience at FAANG companies and AI startups. Within Ensemble’s mission-driven environment, they pursue cutting-edge research in LLMs, reinforcement learning, and neuro-symbolic AI.
Unlike traditional tech companies, our scientists have privileged access to vast, sensitive healthcare datasets and state-of-the-art computational resources, enabling them to experiment with novel ideas and push AI boundaries while delivering tangible improvements in healthcare outcomes.
Real-World Applications: Transforming Healthcare with Agentic AI
By combining elite AI expertise with comprehensive healthcare data and domain knowledge, Ensemble is successfully deploying AI models that generate measurable benefits across hundreds of health systems. Here are some key implementations:
Enhancing Clinical Decision Support
We have implemented neuro-symbolic AI integrated with fine-tuned LLMs to bolster clinical reasoning. Clinical guidelines are translated into a proprietary symbolic language and validated by experts. When hospitals face payment denials for appropriate care, our system analyzes patient records, converts clinical journeys into symbolic representations, and deterministically matches them against guidelines to identify valid justifications.
An LLM then crafts denial appeal letters grounded in clinical evidence. This AI-driven approach has increased denial overturn rates by over 15% among Ensemble’s clients. Building on this success, we are piloting similar capabilities for utilization management and clinical documentation improvement, analyzing real-time records to detect documentation gaps and recommend compliance enhancements that reduce denial risks.
Streamlining Accurate Reimbursement Processes
Ensemble is testing a multi-agent reasoning framework to navigate the complexities of securing accurate reimbursements from insurers. Autonomous agents collaborate to interpret account details, extract necessary data from disparate systems, determine next steps, automate resolutions, and escalate complex cases to human specialists.
This innovation aims to shorten payment cycles, reduce administrative burdens on hospitals, and ultimately enhance the financial experience for patients.
Optimizing Patient Interaction Through Conversational AI
Our conversational AI agents manage inbound patient calls with natural dialogue, seamlessly transferring to human operators when needed. Operator assistant agents provide real-time call transcriptions, surface pertinent information, suggest optimal next steps, and streamline follow-up procedures.
According to Ensemble’s client data, these AI enhancements have cut average patient call durations by 35%, increased first-call resolution rates, and boosted patient satisfaction scores by 15%.
The future of AI in healthcare requires a commitment to precision, accountability, and meaningful impact. By anchoring LLMs in symbolic logic and fostering close collaboration between AI researchers and healthcare professionals, Ensemble is advancing scalable AI solutions that improve outcomes for providers and patients alike.
