The Truth About AI and Healthcare Payer Modernization

By now, it is safe to say that pretty much every executive in the healthcare sector has been tempted by AI. Payers — whether private, commercial, or government — are no exception. Like everyone else in the industry, they face shrinking margins and ballooning costs. They’ve got layers upon layers upon layers of complicated, convoluted, and critical data to manage and modernize. And they’re often trying to harness tomorrow’s technology while still saddled with yesterday’s IT infrastructure. 

When you’re stuck pushing a rock up that “digital transformation” mountain, why wouldn’t you leap at AI’s promises to automate-away your IT load?

I sympathize. But AI is no cure-all.

Reality bites

I’m not saying that the power of these new technologies, however over-hyped, is the stuff of fairy tales. There are huge savings in time and resources to be had in applying AI tools to some particularly sticky payer IT operations and workflow issues. 

For example, most established payers are at some stage of moving from legacy platforms to modern cloud models. But that process requires rationalizing the organization’s data — fundamentally reworking how it is managed and how the entire data stack functions — all without breaking some vital dependency and halting day-to-day business.

No matter how you slice it, modernization is a laborious and complex undertaking. I’ve had way too many meetings in just the past six months with payer teams who are stuck somewhere in the process. 

For legacy platform transitions, current AI tooling can, indeed, help clear a path forward. For example, it can be applied to speed the data rationalization and validation processes that consume 80% of the effort — which is fantastic! But that doesn’t get you over the finish line. It just frees a lot of bandwidth to focus on the 20% that is key to success: the complex business logic that you need to better define, rethink, simplify, and modernize for your new model. 

Sorry. There is no shortcut for strategy.

With so much frustration and so much hope hinging on technology that is the subject of so much hype, it is important to understand where AI actually helps at present, and where it most definitely does not.

Where AI works

To scratch the surface on how current AI technology can help payer organizations with IT modernization, consider Generative AI copilots. These tools aim to be like Ironman’s Jarvis and can be trained to understand particular work context and data to quickly perform a ton of the time-consuming preparations required to guide users through a task to completion. 

For legacy migrations, payer-applied GenAI copilots are already proving beneficial for migrating and integrating data (ETL/ELT, for example) by quickly performing statistics, data quality gathering, overlapping data identification and de-duplication, and automated code writing as opposed to having to write from a clean sheet. 

They’re also useful in tech evaluations for illuminating democratized reporting structures and actual use. Applied to a payer organization’s systems during transition, they can effectively build entity relationship diagrams (ERDs), explain data structures, and summarize dependencies to keep the process on track.

AI copilots are also already useful for optimizing a host of costly and time-consuming — but necessary — functions that are fundamental to payer operations. Consider the role of imaging in prior authorization. AI copilots applied to imaging data speed approval processes for run-of-the-mill procedures with clear identifiable standards. They are also adept at spotting fraud, waste, and abuse and can ensure images are all unique and not submitted to “game” the payer organization for unwitting approval. (Yes, that happens more than you think.) 

And due to their ability to ingest and organize tons of data and data relationships, AI copilots are also already effective for enhancing payer capabilities in care and case management, where they’re used in building delivery approaches for specific conditions or drug management methods. 

Where AI does not (and should not) work

When something is really effective for one thing, it’s natural to assume it may also be effective for other things. The use cases just mentioned show where AI works well, but there’s also a litany of examples where using it botched things up disastrously. 

Payers need to be careful about how they implement AI, and fully aware of its limitations and risks. For example, payers should never rely on AI for:

• Ethical or moral decisions 
• Rare or outlier condition management 

If you need me to explain why AI is not appropriate in these circumstances, please get out of the healthcare business. 

AI does not work even in appropriate areas when it is poorly implemented. Building or integrating an AI model and not keeping it monitored and maintained is a dangerous exercise in futility. Any complex data model without an attached human responsible for validation and interpretation will fail, as will relying on code to generate code without supervision, training, and prompts. 

Simply put: You cannot put AI copilots in the driver’s seat.

This is especially true for direct interface with your customers. In any industry, AI execution failures (even hilarious ones) erode human trust. Payers have broad reach into sensitive and serious matters of health for people, and they need to guard that relationship. Word travels fast, especially in health insurance where all of your customers are generally congregated by just a few community brokers. If even one of those brokers loses faith in your organization’s integrity, much less its ability to manage technology, it can impact your business for a long time. 

AI tools can be incredibly powerful and useful, but they are not magic. They are still just tools and require thoughtful use.

AI blindly implemented to satisfy an executive’s curiosity just because it is the buzzword of the year could be an extremely costly and wasteful endeavor. 

Photo: HASLOO, Getty Images

This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.