I've noticed over the last few years that among health care professionals, there's a lot of discussion around artificial intelligence (AI): how it works, what it can do for the industry, and how to implement it strategically.
It's that last piece — figuring out the most effective way to start an AI initiative — that can really trip people up.
Too often when we talk about AI today, people jump to, "How do we use AI to drive automation?" In this framing, AI is like an all-knowing Magic 8 Ball. Its sole purpose is to automate decision-making, thus reducing costs and the need for human judgment.
You can certainly target those areas with artificial intelligence. But the ability to do so usually doesn't surface until much later in an AI strategy, after you've developed sound foundational capabilities and workflows. (For a step-by-step guide to getting the most out of AI, read our recent article on harnessing AI for human care.)
Given that, how do you actually use AI to build solutions for better health outcomes? It starts with an understanding of the problems and issues best suited to an AI solution. For instance, AI is great at organizing related data points into a single model in order to streamline complex processes that rely on many pieces of information.
But it's important not to choose focus areas that are too complex or that would benefit from human judgment and creativity (for example, diagnosing a rare disease).
What you're looking for are those "just right" issues. Here are four examples:
- Natural language processing - As a health care professional, you might have no problem remembering long, multisyllabic medication names or surgical procedures. But most non-clinicians do. That can sometimes result in important information getting lost in translation between patients and providers. AI can comb through chat logs, emails and other communications. It learns how to map layman language like "sugar pills" or "heart surgery" onto their corresponding medical terms. You can also train your AI model to identify state-specific language around programs and all the different ways customers and clients refer to them. With a robust set of natural language terms, you can build a system that allows you to match words and phrases to very specific customer questions that can be answered conclusively.
- Image data identification - We've all heard the jokes about the legibility (or lack thereof) of doctors' handwriting. But humor aside, messy handwriting in health care settings can lead to all manner of miscommunication. Or it might require time-consuming follow-up to interpret the scribbles. AI systems can quickly scan digital images, whether they're medical records, prescription notes or marked-up x-rays. It then "translates" illegible sections by drawing on their data stores of similar images. AI can also review images for typical appearances and flag suspicious or abnormal areas for human review (for example, spotting a possible tumor on an MRI scan).
- Eligibility and prior authorizations - Determining whether a customer is eligible for a particular program or pre-certified for a specific procedure can involve combing through vast numbers of policies, procedures and patient records. Until recently, these complex tasks could only be managed by human operators and demanded countless hours of meticulous work. Even worse, human errors in this area, like mistakenly denying coverage, can lead to huge hassles and stress for customers. AI systems are perfectly suited to pulling information from numerous places to paint a complete picture of an individual's eligibility status. This allows human workers to make final determinations using a more accurate data foundation.
- Customer engagement tactics - Think fast: What's the optimal time of day, day of the week, and communication channel to engage your customers? If you're not sure, don't worry. This was something of a trick question. The fact is, your customers aren't a monolith, so there's no single best way to engage them. Depending on who you're reaching out to — working parents, elderly diabetes patients, adult caregivers, and so on — you might find that engagement dips and rises wildly depending on when and how you communicate. AI can surface these kinds of insights. It allows you to craft and deliver messages to micro-demographics that have a far greater chance of getting a response.
Building an AI competency within any organization requires significant energy and resources. So starting with careful consideration of the types of projects and challenges you want to tackle first is very important. Remember, AI isn't a one-time event or a Magic 8 Ball. It's a commitment to using technology to personalize the way providers, members, patients, caregivers and other access services.
Whether your organization is fully behind AI or still in the contemplation stage, I want to hear from you. How does AI impact your organization today? What are some areas of your work that might benefit from AI?