Health care organizations are heavily investing in digital transformation to modernize operations and enhance service delivery. Reports indicate that 40% of these organizations are dedicating between $50 million and $100 million annually to digital technology. Furthermore, 66% are actively implementing new digital solutions. These investments aim to improve operational efficiency, patient service, and administrative functions.
Artificial intelligence (AI) has become a major component of this digital transformation. AI supports applications such as predictive analytics and clinical decision support. Predictive AI, a longstanding tool in health care, utilizes machine learning to predict future outcomes like readmission risks and early disease detection. Despite this interest, a readiness gap exists. Only 18% of health care organizations feel prepared for AI implementation, based on an HIMSS Market Insights survey. This gap underscores the challenge of transitioning from experimentation to enterprise use. Successful integration depends on robust infrastructure, governance, and operational alignment.
actAVA, a platform focusing on AI lifecycle management in health care, addresses these challenges. They have introduced Cura, a one-trillion-parameter model tailored for health care. Cura aids organizations in embedding their institutional knowledge, including workflows and policies, into proprietary AI systems. This initiative seeks to move organizations from renting external AI models to developing their own competent systems capable of clinician-grade communication and clinical reasoning.
Kevin Riley, CEO and co-founder of actAVA, believes enterprise AI will progress through developed systems owned and refined by individual entities. With health care’s emphasis on reliability, organizations must prioritize how AI is deployed and managed within their operations. This involves going beyond model capabilities to ensure dependable execution in real-world scenarios.
Frank Wang, CTO and co-founder, highlights the transition towards engineering dependable AI systems rather than merely developing capabilities. Building AI demonstrations is relatively straightforward, but achieving consistent performance requires detailed planning and continuous improvement.
To facilitate AI integration, health care organizations need systems connecting models, workflows, and governance processes. Agentic AI, able to manage multi-step tasks, offers automation opportunities, though safety measures are vital. actAVA focuses on managing AI agents through their lifecycle, ensuring policy compliance and adaptability within health care settings. Specialized language models tailored to health care workflows further support these efforts.
Ownership is a core theme, with health care organizations holding valuable knowledge within their data and processes. Future AI systems will enable greater control over these assets. Wang foresees a shift where 90% of enterprise AI tasks operate on common models for routine work, while 10% address novel cases with advanced models.
Weiran Yao, CAIO and co-founder, notes AI is transforming the interaction between expertise and technology. Organizations can leverage their accumulated knowledge as active capabilities while maintaining governance over AI deployment.
Benchmarking is crucial in evaluating AI performance in health care environments. actAVA’s collaboration with health care professionals and academic partners led to χ-Bench, a benchmark to assess AI agents in completing complex tasks. These tasks encompass processes like provider prior authorizations and payer utilization management.
The evaluation through χ-Bench showed the strongest agents succeeded in 28% of tasks at pass@1, with consistency dropping further in repeated scenarios. This highlights the importance of reliable deployment infrastructure.
AI systems must navigate multiple stages and adhere to policies within defined operational bounds. The χ-Bench findings underscore the need for thorough evaluation frameworks and governance in AI integration.
For health care organizations, advancing AI adoption involves establishing a foundation linking technological capabilities with operational accountability. As AI becomes more integrated, assessing risk management and workflow adaptation will be just as vital as technological sophistication. The growth of health care AI will rely on merging advanced models with governance structures and evaluation frameworks, supporting practical and reliable applications in health care settings.
Leave a Reply