In 2024, the San Francisco Fed and the Federal Reserve System Innovation Office launched the EmergingTech Economic Research Network (EERN) to support a better understanding of how new technologies like GenAI are shaping the economies of today and the future. As part of the EERN initiative, we often hold roundtable discussions to hear from industries and sectors at the front lines of digital adoption to ask questions such as: How are artificial intelligence (AI) tools being used? Is AI adoption still in the early stages or has it been fully integrated in certain areas? What are expected or realized impacts on firm productivity and employment? How is sensitive data protected?
To address these and other questions on the adoption of AI in the healthcare industry, we recently convened a roundtable of executives within the Twelfth District, with representation from an integrated managed care provider, a not-for-profit multi-state healthcare system, a regional not-for-profit health plan provider, and a pediatric hospital.
The executives met with Mary C. Daly, President and CEO of the San Francisco Fed, Sunayna Tuteja, Senior Vice President and System Chief Innovation Officer for the Federal Reserve System, and Kevin Ortiz, Deputy Chief of Staff and co-head of EERN.
AI Adoption and Integration
Kicking off our roundtable discussion, we asked the participants how AI is being used within their organizations and the level of integration so far. One executive noted four key areas where AI is being used: 1. clinical decision support; 2. patient engagement; 3. medical education and training; and 4. administration and operations.
One executive noted that administration and operations, with tasks such as billing, procurement, coding, and system management, are seeing some of the greatest efficiency gains from AI to date. However, use of AI to assist in clinical decision support is still in the early stages, and it is primarily used to support medical professionals rather than as a decision-maker. Other executives said their organizations are also using AI to enhance patient experiences, specifically assisting clients with finding information directly or reaching the right person to speak with. They remarked that call center service has also improved as workers have faster and more comprehensive access to key information to perform their jobs. One executive noted that AI is being used to steer patients to either urgent care or the emergency room (ER) by assessing their condition and providing instructions. They shared that this greatly improves the flow of patients to the right places which reduces bottlenecks in the ER, relieves strains on the staff, and enables the most important emergencies to be seen faster and more effectively.
An executive also emphasized the benefits of ambient note-taking systems. Other executives agreed that manual note-taking had been contributing to burnout for physicians and nurse practitioners in recent years. They said ambient AI note-taking has transformed that work, greatly improving productivity, strengthening staff retention, and enhancing patient experience. They shared that doctors and nurses can more attentively interact with their patients when assisted by AI, instead of having to look down and focus on their notes.
The executives also discussed a helpful framing device for understanding AI’s impacts in the sector, although the device applies in other areas as well. Specifically, the “5 A’s Framework” shows that AI can help with Assistance, Augmentation, Automation, Amplification, and Acceleration.
Complimenting Human Skills
Each of the executives noted they largely see AI as a tool that will augment human skills rather than replace them and said that this is particularly evident in the clinical decisions space. They noted that physicians and nurse practitioners always review AI- compiled data, and that the AI essentially functions as an assistant, not a decision-maker. Several executives also noted that AI is being used to more effectively bring together and process information from existing, disparate databases. One executive noted they have connected 200 outside databases over the last 10 years which has been particularly helpful with unusual diagnoses. The executive believes AI will further improve the ability to aggregate and process this type of critical information, leading to higher quality decisions.
Another executive characterized certain AI applications as akin to a medical device. Practitioners will need to be skilled in the AI tool and supervise its use, just as they would with other medical devices. Other areas of AI complementation included marketing efforts, which can now customize content for people in the modality they prefer, rather than the non-personalized broad-based advertising model. Accelerations in research and diagnostics were also mentioned as benefiting from AI.
One executive noted that external pressures, such as changes in funding models, may necessitate organizations to leverage AI to enhance productivity and efficiency. The executive noted that AI tools help to boost medical professional work quality and satisfaction.
Data Protection
Balancing the benefits of AI-enabled innovation with potential data security risks was a chief concern during the conversations. According to the participants, while healthcare providers use some public data, much of it is proprietary and highly protected through regulation. An executive added that AI tools need to be trained on local data to ensure outputs match realities on the ground. One executive stressed the importance of protecting the data at each stage of development and integration, emphasizing the need for the right safeguards.
The executive also mentioned the importance of patient confidence in the systems, referencing the initial skepticism over virtual medical appointments when they first rolled out, which have now become more normalized. AI is likely to endure similar reticence before becoming more widely accepted, in the executive’s view. Another executive mentioned security processes, bias reduction, and quality assurance programs as key requirements for their AI applications. This was echoed by an executive who said they have a full task force focused on governance of diagnostic models to ensure accuracy, quality, and data protection, and to monitor for algorithmic bias.
Related:
The views expressed here do not necessarily reflect the views of the management of the Federal Reserve Bank of San Francisco or of the Board of Governors of the Federal Reserve System.