Blog: Five Criteria for Choosing Initial Payer GenAI Use Cases

By John Pyhtila, PhD, Chief Product and Strategy Officer and Gokul Varadharaj, Chief Revenue Officer & Co-Founder

The generative artificial intelligence honeymoon phase is drawing to a close. While the technology’s potential for transformation is undoubtedly real, few initial efforts by companies in healthcare and other industries have yet to move beyond pilots to scalable models that deliver enterprise value.

For payers, the business opportunity in making that leap is immense. The right payer GenAI solutions could deliver net savings of 13% to 25% in administrative costs and 5% to 11% in medical costs plus boost revenues by 3% to 12% according to June 2024 McKinsey estimates.

Getting to scale

One of the fastest ways to achieve significant impact is leveraging open-source models and customizing them with the payer’s own proprietary data for its specific use cases. This strategy tailors the model to an organization’s unique needs without the cost and time required to develop a new LLM. Read the first in our GenAI Health Planning blogs, Payer-Customized Generative AI Models Deliver Strategic Impact.

To build successful payer-customized large language models (LLMs) and large multi-modal models (LMMs), organizations need to follow four development requirements: Trust, Security, Scalability and Data Usability. These requirements will ensure models are built on a solid and scalable foundation. Read our second GenAI Health Planning blog, Four GenAI Development Requirements for Payers to Realize Strategic Value.

With this development framework in place, payers can then focus on picking the best initial use cases.

Five criteria for initial use cases

1. Start with the business problem, not the tool.

Instead of becoming enamored of GenAI technology, payers should determine where they have the most significant opportunity to improve. Historically, to find improvement opportunities, payers have looked at their performance vs. industry benchmarks. But in the new age of GenAI, if McKinsey’s estimates on improvement opportunities are accurate, looking at ‘rear-view mirror benchmarks’ is no longer valid because those benchmarks indicate what good performance looked like prior to harnessing GenAI to drive transformative improvements.

While generative AI is not the best solution for all improvement efforts, GenAI can have significant impact in use cases that include:

  • high human capital costs associated with synthesizing information and/or creating work products, including prior authorization review, network design and contracting, RFP response creation, IT software coding, and Analytics.
  • improving member, employer, and provider engagement, such as call center operations, case and care management, account management, and provider relations.

An effective GenAI governance approach requires both tech and business executives to decide uses cases. At this stage, it’s important to move beyond experimenting and secure some near-term wins.

2. Drive tangible business benefit.

GenAI has the potential to substantially impact payer revenue, medical costs, and administrative costs. A key factor for success depends on measuring GenAI’s impact and measuring the investment required to achieve that impact.

With this measurement framework in place, payers can quickly double-down on areas where ROI is generated and cut losses where value isn’t being achieved. These measurements enable senior management to justify further investments and change-management efforts to help teams to trust, accept and use the solutions.

Payers will benefit from starting with more targeted use cases to quickly prove out the value, learn from these initial cases, and harden the technology, processes, and approach to preparing the workforce for GenAI to pave the way for bigger impact in the future.

3. Mitigate risk.

Given the high level of skepticism and even fear surrounding generative AI and the admitted hallucinations and other inaccuracies at this early stage, it’s critical to steer clear of developing fully autonomous solutions.

Instead, payers should look to drive efficiencies with humans, not completely replace them. Additionally, payers should not focus on a single use case to ‘prove the value’ of GenAI, instead they should work on two to three use cases in different parts of the organization in parallel to accelerate learning on how best to capture value through GenAI.

For example, a payer could look to drive efficiency in the IT team through GenAI-augmented coding, improve a call center rep’s ability to respond to member questions by using GenAI to answer questions about benefit design, and drive efficiency in prior authorization review by using GenAI to make recommendations to reviewers supported by clear citations.

Through this approach, companies can mitigate suspicion surrounding GenAI. Employees won’t be relinquishing control and will have final review before decisions are made or actions are taken. Additionally, by tackling multiple use cases, the flywheel of GenAI value creation will start spinning faster, which will result in accelerated ROI.

4. Put fundamentals in place.

Usable data is essential for realizing the full value of GenAI models. One key to accelerating business value, then, is prioritizing initial uses cases that have a foundation of data that is well-understood and usable—accurate, complete, timely, relevant, versatile and suitable for any use case or applications–and thus trustworthy.

It’s also important that the models are contained in a secure cloud environment to foster greater trust. Isolating the development platform, data, and end application from the public internet ensures the query or data, including personal health information, personally identifiable information and/or company confidential information, does not leak into the public domain.

Meeting these guidelines could mean initial use cases rely only on data already vetted in the secure payer environment and don’t draw on external sources unless the payer has the capability to make all data usable.

5. Maximize value with tool + redesigned workflow.

While payers could potentially plug GenAI models into existing workflows, scaling the impact and maximizing outcomes requires new workflows and employee behavior developed through training and upskilling. To fully realize the value of GenAI models, payers need to simultaneously rethink the entirety of the workflow and understand the impacts on the employees, including job descriptions, required skillsets, and necessary training to support new processes.

For example, call center reps could leverage a GenAI solution which ‘listens in’ to the conversation and automatically generates responses for the rep to use. The reps no longer need to be experts at finding information, but instead will need to quickly understand and validate the GenAI content and weave that into their conversations with members. Or a software coder needs to be more proficient at articulating the required output of the code and ensuring all edge cases are considered versus being a technical expert at writing code.

Because of new workflows, payers will need to put major effort into change management, supporting employees in getting comfortable in a GenAI world. This focus cannot be underestimated as many otherwise good initiatives have failed due to a lack of focus on change management.

Compounding effect of GenAI leadership

McKinsey has found that leaders using AI to transform their organizations see a compounding benefit and are pulling farther ahead of AI laggards. AI leaders are doing this by:

  • Bringing business, technology, and operations more closely together to digitally innovate.
  • Upskilling their organizations.
  • Building a distributed technology and data environment to empower hundreds, if not thousands, of teams to digitally innovate, day in, day out.

Payers that choose to follow “could find themselves in a position where industry leaders have 10% lower medical costs or 20% lower administrative costs, a daunting prospect. So the ability to be a fast follower will be critical,” McKinsey reports.

Focusing on payer-customized LLMs and LMMs, sticking to the critical development requirements of trust, security, scalability and data usability and following the five criteria for choosing early use cases are all key to putting payers on the path to GenAI success, value creation, and market leadership.

Contact us at if you’d like to learn more about getting the usable data you need to power GenAI and strategic initiatives.