This article was originally written for Forbes. View the article here.
Ainsley MacLean, MD, is the Former Chief Medical Information Officer (CMIO) at the Mid-Atlantic Permanente Medical Group, Kaiser Permanente.
Artificial intelligence is increasingly playing a role in healthcare, such as in breast cancer screening and diagnosis and medical transcription. In fact, in a 2024 survey by McKinsey, “more than 70 percent of respondents from healthcare organizations—including payers, providers, and healthcare services and technology (HST) groups—[said] that they are pursuing or have already implemented gen AI capabilities.”
At my organization, we’ve rolled out AI in several key areas. Some of our physicians are leveraging AI for support with documentation, radiology dictation, breast cancer analysis in radiology and pathology, diabetic retinopathy screening, flagging urgent messages from patients and finally, identifying patients who are likely to become seriously ill. In each instance, physicians are treating AI as an assistant, using the technology to streamline their workflows or add additional analytical power that enables them to deliver even better care to their patients. They never use AI as a replacement for physician expertise. We’re still piloting and researching AI usage; these use cases are not yet present across our entire network. However, along the way, my team and I have learned important lessons about how healthcare institutions can strategically, ethically and responsibly implement AI.
Why A Strategic Approach To AI Implementation Is Crucial In Healthcare
At the core, AI involves content generation, be it an article draft, a line of code or a medical analysis. Whenever content is generated, especially in healthcare where patients’ lives are at stake, it’s imperative to carefully consider what inputs led to the creation of that content and who is responsible for overseeing that content.
In healthcare, as with any profession, AI tools are only as good as the data they are fed. Even if healthcare leaders have evidence that good data is being input into the AI tools they implement, in order to safeguard accuracy, there must still be people tasked with overseeing outputs. When healthcare leaders sit down and map out how they will bring AI to their organizations step-by-step, they’ll be more likely to put processes in place that minimize the chances of inaccurate or biased results.
Key Steps To Strategically Implementing AI In Healthcare
To strategically implement AI in their organizations, healthcare leaders should take several key steps.
First, healthcare leaders should make sure their physicians, nurses and other medical professionals are ready to embrace AI. While AI has come a long way in the public’s perception, there is still skepticism around it. To find out how ready their medical teams are to use AI, healthcare leaders should conduct surveys and meet with staff face-to-face to get their thoughts. These informal conversations can help leaders identify opportunities to pilot new tools with early adopters. Simultaneously, healthcare leaders should educate their medical teams about AI’s medical use cases. Given that AI is still in its early stages in healthcare, leaders should treat it as an optional tool for medical professionals, not a mandatory one.
The next step stakeholders should take is to think carefully about how they want to approach AI as an organization, what their operational goals are as an organization and what problems they want to solve with AI. For instance, do they want their organization to be a leader in AI implementation or a later adopter? Do they want to use AI in a certain department first? Is there a clunky workflow within the institution that AI could help streamline? The answers to such questions will impact how healthcare leaders navigate AI implementation.
Then, healthcare leaders need to ensure they have the right governance processes in place. Those processes will help everyone on the team remain accountable as AI is piloted and potentially rolled out throughout the organization.
From there, healthcare leaders need to decide which AI system they’re going to use so they can start building the pilot program. They could turn to existing solutions on the market or, if they have the budget, develop solutions internally in partnership with outside vendors. No matter which route they choose, they must think through how existing technologies will interface with the proposed AI system. At a minimum, any AI tool that’s implemented should smoothly integrate with electronic medical records (EMRs). Moreover, stakeholders must also keep data security and privacy top of mind. They should also devise standardized workflows that will help everyone who will use the AI tool do so in a correct, responsible manner.
In my experience, it can take many months of planning for an AI pilot program to be rolled out in a healthcare organization. Rollouts should, in my view, be multi-pronged. For instance, leaders could start with a technical rollout, then move on to a test rollout where a few doctors who are highly trained in the technology use it for a few months. As time goes on, the AI solution can be scaled and introduced to more medical team members—with leaders capturing feedback every step of the way to evaluate if things are going in the right direction.
The Importance Of Measuring Results
Feedback is crucial even after the rollout phase of an AI tool is over. Healthcare leaders, physicians and other medical staff should carefully monitor the implemented AI system to catch any biases and errors. Specifically, they should continuously check inputs and outputs. In healthcare, my team and I have found that AI works best when it’s trained on data from a diverse set of patients—and when it’s used to treat a diverse set of patients.
In terms of KPIs, they will vary depending on the goals of a healthcare organization. However, regardless of what the exact KPIs are, there must be processes in place to regularly evaluate them and take action accordingly.
Ultimately, by strategically implementing AI in their organizations and treating it as an ongoing, evolving effort, healthcare leaders can help physicians, nurses and other medical staff treat patients with greater precision.
