Health care’s quest for an enterprisewide AI strategy

The increase in investments isn’t surprising, as 90% of the health care leaders surveyed believe that AI initiatives are important for their organizations to remain competitive in the market. When asked about their organization’s approach to technology innovation, 80% self-reported that they are either edge experimenters (organizations that tend to be first adopters of new technology or first to try new approaches and test unknown use cases) or fast followers (organizations that typically are next in line to adopt after some experimentation).

As adoption of AI increases, so can the risks. AI’s potential risks make it even more important for health care organizations to establish appropriate governance and oversight of algorithms and data (see sidebar, “Deploying initiatives to tackle AI bias in a trustworthy way,” for more information).

Our survey shows that health care organizations are better prepared to manage AI’s potential risks compared to 2019. Survey respondents reported doing a better job compared to 2019 on risks arising from cybersecurity vulnerabilities, unethical AI systems, and perceived job losses resulting from AI automation (figure 2).

One of the health plan interviewees mentioned that they’re solidifying data governance through a central repository to have one version of truth. They’re also strengthening their data management practices and investing in what they call “explainable AI” to address unconscious bias. For example, instead of providing a social security number for member matching purposes, they’re now tokenizing or deidentifying the data they share with external entities. Another example is an academic medical center in the United States that’s tackling AI bias by continually evaluating the purpose of the data and reasons behind using AI over the course of its lifetime to account for and correct data biases.³

The barriers health care organizations face in achieving AI maturity

Making integrated data more accessible: Our interviewees mentioned that AI succeeds in its purpose when “there is a rich trove of data” and when “multiple datasets can talk to each other.” However, health care organizations often are challenged by siloed, unstructured, and at times, incomplete and inaccurate data. To add to this, patient-level data is sensitive and highly regulated, which makes it more difficult for AI applications to gain access to it. Lack of access to clean, integrated datasets hinders the ability to train high-performance AI models and deploy them at scale. According to one health plan interviewee, the organization’s leaders are focused on generating AI-powered insights to drive outcomes, improve quality, and reduce costs. To achieve this goal, a key is building an integrated health care dataset, which includes health plan data (e.g., enrollment and claims data), provider data (e.g., electronic medical records), and third-party data (e.g., health equity and ESG-related data).

To help ensure that AI algorithm models are robust and equitable, health care stakeholders recognize the importance of including data around drivers of health. For example, the Biden administration formed the Equitable Data Working Group to establish equitable data practices by generating disaggregated statistical estimates to depict experiences of historically underserved groups using survey data, increasing non-federal research and community access to disaggregated data, and conducting robust equity assessments of federal programs. Using equitable data in AI models can highlight opportunities to improve outcomes for underserved communities.⁴

Establishing an enterprisewide AI strategy: Our survey findings showed that just one in three executives strongly agree that their organization has an enterprisewide AI strategy, their leaders communicate the vision effectively, and AI differentiates them competitively (figure 4). Our interviewees discussed how, at times, AI loses out to other organizational priorities, and may not be featured in the enterprisewide strategy. As one of the interviewees pointed out, “there are a growing number of AI-powered point applications across the enterprise, but most are utility-based, and there isn’t a unified organizationwide strategy on AI.”

Keys to achieving AI maturity for health care organizations

Our interviewees and survey respondents highlighted three principles that health care leaders should embrace to achieve enterprisewide AI-led transformation:

1. Gaining senior leadership support: An executive champion is essential for crafting and translating this vision into achievable action steps and milestones, according to our interviewees. They discussed how the cultural characteristics set by executive leadership—namely the ability to make tough decisions, manage change, and partner with the workforce on key decisions—are as much as or more important to achieving AI maturity than the organization’s core technology capabilities. Health care organizations should build an AI culture with leaders driving support through communication and transparency at all organizational levels. While critical, gaining leadership support is just one of the factors that help ensure success. In fact, most survey respondents named leadership support, strong data, and technical capabilities the top three factors most critical to success of AI implementation.

2. Elevating talent beyond technical skills: The survey respondents believe that their organizations need a mix of technical (AI expertise, data science, project management) and professional (critical thinking, adaptability) skills to achieve broader success of AI initiatives. The survey results also showed that health care organizations are undertaking significant changes in workflows and workforce to scale AI initiatives and achieve greater maturity (figure 5). For instance, four in five respondents said that their organizations are changing how they create teams, roles, and workflows to take advantage of AI. The same number of respondents said that their organizations are investing in change management and AI readiness by creating new AI jobs and by training the workforce to better integrate AI into their workflows. One of our interviewees expressed the need for a training program or curriculum that, in a simplified way, teaches their typical health care staff just enough about AI, without overwhelming them with the science behind it. This can enable them to have a meaningful conversation about AI applications they use.

3. Pursuing the right ecosystem partnerships: To adopt an enterprisewide AI strategy, organizations should consider collaborating effectively and making strategic decisions on build versus buy. More than one-third of the survey respondents believe that cloud vendors, IT professionals, and consulting firms are the key ecosystem partners for their respective organization’s AI success. Interviewees mentioned that health systems are still at a nascent stage when it comes to developing the technical architecture needed to train, integrate, and validate AI models at scale. Health systems aren’t yet designed to deploy AI at an enterprisewide level. To address this issue, health care technology vendors offer product packages that help enable health systems to quickly train and validate models, which allows them to focus more resources on designing workflows and improving clinical outcomes. Platform-enabled ecosystems can bring together ecosystem participants on a digital network and help health care organizations improve and expand their services by providing opportunities to expand customer reach, access new capabilities, and increase revenue.

The path to an AI-fueled future for health care organizations

Comparing the survey findings shows that health care organizations have made progress on their AI initiatives since 2019. The pandemic, in part, has provided avenues for health care organizations to quickly turn AI pilots into full-scale implementation in many functions, according to our interviewees. Here’s how the path to an enterprisewide AI transformation moves through the health care value chain (figure 6):

• Operations: AI-enabled digital authorization for fast and frictionless decision-making
• Consumer: AI-enabled personalized services that help in omnichannel engagement for consumer needs
• Clinical: AI-enabled care that focuses on prevention, monitoring, and delivery of care
• Performance: Autonomous monitoring that helps in real-time anomaly and trend identification
• Workforce: AI-enabled smart workforce management by optimizing resource and talent allocations

When health care organizations apply AI across the value chain, they can improve consumer health and well-being and support better outcomes while also boosting organizational efficiency and reducing costs.