Five-point strategic blueprint for AI and generative AI in healthcare | Viewpoint

Navigating the myriad perspectives on artificial intelligence (AI) and Generative AI (GenAI) can be overwhelming for a healthcare executive.

MJ Stojak

Rather than revisiting these technologies' capabilities and limitations that have been extensively covered, it may be more valuable to discuss strategic approaches for embracing and leveraging AI and GenAI within a healthcare organization.

Given GenAI's nascent nature, it is understandable for leaders to want to tighten governance processes or even block access to GenAI or ChatGPT. But such restrictive measures often lead to workarounds or missed opportunities for improving outcomes or saving lives - a lose-lose situation.

Instead, healthcare organizations should apply some strategic approaches to embracing AI and GenAI.

Consider the following to minimize adverse outcomes, as well as promote transparency, innovation, and agency within your organization.

1. Identify the problem and develop a use case

Clearly define the business, clinical, or operational problem you believe GenAI can solve. Develop a use case to ensure the right technology (AI, GenAI, Machine Learning, etc.) is applied, and factors such as clinician and patient risk, data security, ethics, equity, and privacy are considered during the evaluation process and solution design.

Healthcare organizations have been using AI broadly to optimize patient flow, replenish supplies, identify revenue cycle leakage, validate ICD-10 coding, address plan staffing, and assess patient risk, among other areas. We can all expect that the deployment of AI will eventually expand to other areas and continue to improve overall.

Developing a well-defined use case for your healthcare organization is important because it ensures targeted, secure, and effective application of the technology with articulated outcomes such as improving patient outcomes, streamlining operations, or enhancing decision-making.

2. Prioritize cost-saving solutions

Focus initially on use cases to drive the most cost savings for your organization. This will make the impact powerful and support the bottom line. The AI deployment areas cited above save organizations millions of dollars and drive operational efficiency gains. Going after the low-hanging ROI projects is an approach that will make it easier to secure funding and, over time, allow for the inclusion of essential but harder-to-quantify use cases as processes and frameworks for AI exploration continue to mature.

3. Ensure data quality and consistency

Data fidelity may be the most critical issue to address over the coming years. Even if you believe your data quality is excellent, and let's say it is, it probably isn't good enough for future AI use cases.

To start, evaluate the quality and consistency of your organization's data. All flavors of AI will leverage this data, so it is imperative that your organization begins with a single source of truth and shared values for metrics across teams. This will provide a baseline and ensure the trustworthiness of decisions and information based on the output of these capabilities.

Don't overlook the importance of data quality, especially as we move toward the future of AI. We currently rely upon human judgment to compensate for data gaps. Still, with the new models, the volume and complexity of the data will make it impossible for humans to intervene. Very soon, data gaps won't be acceptable. Accuracy and completeness need to be as close to perfect as possible. Parallel path the data integrity work so that you can keep the train running as you build the track. In addition to getting your house in order, you'll also need to invest in monitoring tools to provide ongoing validation of the data and models that your organization uses.

4. Strengthen governance processes

Review and revise your governance process to include clear evaluation criteria and guiding principles for AI or GenAI. This will enable your leaders to consistently determine risks (security, privacy, legal, clinical, patient, etc.), consider ethical and equitable impacts, review success criteria, and measure success at appropriate points.

A top governance tip is to leverage "frameworks" over "controls." Frameworks help guide decisions and trust people to act responsibly to leverage and evolve the efficacy of these conclusions. This approach minimizes the frequency and impact of adverse outcomes while also promoting transparency, innovation, and agency within the organization.

To this end, two initiatives have been launched publicly to help shape and refine the governance process:

• Microsoft and 18 healthcare leaders announced the creation of a consortium called the Trustworthy & Responsible AI Network (TRAIN), which aims to operationalize responsible AI principles to improve the quality, safety, and trustworthiness of AI in health.
• Late last year, President Biden issued a landmark executive order to establish new standards for AI safety and security across all industries focused on protecting Americans' privacy, advancing equity and civil rights, standing up for consumers and workers, promoting innovation and competition, and more.
• Past colleagues at Seattle Children's Hospital created a data ethics checklist to address the lack of standards supporting decision-making when determining and deploying new data products in a hospital.

Refer to these three resources and share them with your teams to help establish frameworks for your healthcare organization.

5. Foster a continuous improvement mindset

Embrace a continuous improvement mindset to manage the output provided by AI, GenAI, and ChatGPT. Human evaluation and clinical oversight of all output used for significant decisions will be crucial. Review success criteria with an open, experimental mindset, acknowledging that some solutions may not meet the criteria and must be canceled.

We recommend convening a committee of clinical, operational, financial, legal, security, compliance (privacy, risk, etc.) and ethics leaders and experts to review each project's defined success criteria and determine whether it has been met. If not, evaluate other learnings, such as patient/provider/clinician feedback, and determine whether the project still has merit. If so, advise the team to revise their approach and success metrics. If the project doesn't meet the success criteria and doesn't offer value in other ways, be prepared to cancel the initiative.

With over 25 years of experience collaborating with various industries on digital, data, analytics, and AI strategies, I have observed that the critical factor for success is seldom the technology itself but rather the shifts in workflows and processes that individuals must adapt to and adopt. This fundamental truth remains unchanged in the era of AI and its various forms.

In the healthcare sector, there is an amplified necessity for a deliberate focus on supporting the workforce who will interact with and be directly influenced by this technology. By addressing the five strategies outlined above, you have a blueprint to embrace and harness the power of AI and GenAI in ways never imagined.

MJ Stojak is the managing director of the Data, Analytics & AI practice with Pivot Point Consulting, a healthcare IT consulting firm.