Our customers have referred to the Infinitus AI agent as the best performer on their team. And while the AI agent – we call her Eva – may seem like magic to the payor representatives she speaks to, it is in fact the Infinitus knowledge graph that powers her capabilities. 

Those capabilities are anything but simple. Eva needs to acquire accurate information for Infinitus customers, and to accomplish that she must have a huge amount of knowledge about the specifics of healthcare and health insurance plans. 

Thanks to our knowledge graph, our AI agent can determine what kind of insurance policy a plan categorizes into and run live calls more efficiently. For our fully digital offering, AI Connect, we can mine rules from our knowledge graph to determine if a particular treatment is covered by a patient’s plan.

Below, a deeper dive into the challenges the Infinitus knowledge graph must solve for, as well as an example of what it looks like in action.

There are healthcare-specific challenges …

Consider this: There are 33 independent Blue Cross Blue Shield companies, and no two of them operate the same. All told, there are thousands of medical and pharmacy insurers in the US. While the top five payors own 50% of the market share, there is a longtail of additional insurers patients rely on. And again: No two operate the same. 

Within each payor, there are different plan designs – commercial medical is perhaps what most are familiar with, but there are also government medical plans (e.g., Affordable Care Act plans, Medicare plans), government pharmacy plans, short-terms plans, and supplemental plans (e.g., dental or vision). The plan type and plan design determine a patient’s cost of care, in the form of premiums, co-payments, co-insurance, and/or deductibles. This information is critical for healthcare providers to understand, and extremely important for the AI agent to get right, given the extremely high costs medications can have. No one wants to be stuck with a surprise bill. 

Every day, the Infinitus AI agent needs information from hundreds of plans, and through our rule-mining algorithm, can determine what kind of policy the plan categorizes into. As an example, this helps our AI agent run live calls more efficiently.

… and there are technical challenges

Unfortunately, benefit verification and prior authorization information is not available in a centralized database. Not only that, but many plan rules and policies don’t even exist in an accessible digital format – some, even in 2024, are stored inside spreadsheets and PDFs. 

And beyond the complexity of health insurance plan types and plan designs, there’s the challenge of the sheer numbers game. In 2024, for example, the average medicare beneficiary had over 43 plans to choose from. And even members enrolled in the same commercial plan provided by the same payor, may have different step therapy guidelines depending on the state in which they reside. 

On top of all that, rules change frequently. The Affordable Care Act dramatically changed healthcare in the US, and the Inflation Reduction Act (IRA) will change much about Medicare in the US in the next few years. Navigating these rules, and staying up to date on how payor plans change, could be a full-time job itself. For this reason, the Infinitus knowledge graph has constantly refreshed data, so that it is able to keep up with changes better than any human caller possibly could. 

The Infinitus knowledge graph in action

We use historical data from over three million calls that feed into our rule-mining algorithm to come up with accurate, precise, and logically true statements to help our AI agent run calls as smoothly as possible with live payor agents. The payors we call offer thousands of different plans and we are able to capture the variations through our knowledge graph. We also ensure that the rules that our algorithm generates are based on a minimum sample size to ensure accuracy and remove anomalies.

An illustration of the Infinitus knowledge graph workflow process.

This workflow runs on a daily basis – new, accurate rules are added and outdated rules are removed from our system to keep the data produced by our knowledge graph real time and of high quality. The knowledge graph powers all of the models in our AI system, and it is constantly being updated by our AI agent experience, payor integrations, and third-party data sources. 

As mentioned above, the Infinitus knowledge graph powers all of our products, from our AI agent to our AI Connect eBV+ offerings, allowing us to both double-check payor representative utterances, and return digital data to our customers. All of the AI agent use cases, and our recently released Medicare Part B eBV+ solution, rely on the intelligence of our knowledge graph. If you or anyone on your team is interested in learning more about Infinitus please contact us for a demo.