Skip to content


This blog is based on a conversation with Ricky D. Sluder, CFE, Head of Pre-sales SMEs & Customer Success Managers - Americas at Shift about the use of AI support in Health Plan SIUs for our video podcast, which you can find here.

If you’re in the market for a new solution to help your investigators detect fraud, waste, and abuse, you’ll undoubtedly find numerous solutions boasting about their Artificial Intelligence (AI), Machine Learning (ML), and Natural Language Processing (NLP) capabilities. And they aren’t lying — there aren’t many charlatans among the people trying to sell technology to people who literally perform investigations for a living. The problem is that many of these solutions require data organized in a specific way or that they’re only capable of detecting statistical anomalies, which is great data, but it’s difficult to use. 

Most AI-powered solutions on the market focus on offering insights that investigators can use in their manual processes. Shift’s approach is a little different. Shift Healthcare subject matter expert, Ricky D. Sluder, CFE points out three key differentiators that make Shift’s solution stand out from the competition in the health insurance fraud detection landscape.

1. Reinforcement Models

Shift’s AI improper payment detection solution uses supervised, unsupervised, and reinforcement learning models.

Supervised and unsupervised models are the types that are used by most industry practitioners — they’re great at finding statistics and identifying outliers that could benefit an investigator, but applying that data requires a lot of manual work.

Reinforcement learning helps train the model to think from a more linear perspective and present analysis in a way that’s more useful to investigators. Using this method, Shift teaches the model to not only look for patterns but also to automate the process of deriving meaning from them. This way, the model speeds up the traditionally labor-intensive process of identifying potential fraud, waste, and abuse cases.

Shift’s solution provides historical information to inform an analytic result, and external data integration automates the work an investigator would normally have to do by hand to come to a conclusion. Without reinforcement learning, AI models offer another tool for the box. When reinforcement learning is used, the model becomes a solution. 

When trained with reinforcement learning, Shift’s solution can even proactively detect new and emerging schemes, like telehealth or wound care fraud, even if human investigators haven’t caught onto it yet. Since one of the biggest challenges to stopping health insurance fraud is how quickly bad actors evolve their techniques, this capability is a game changer. 

2. Faster Detection and Prevention + Higher ROI 

When asked what value Shift’s technology specifically holds for Special Investigative Units (SIUs), Sluder highlighted two key aspects.

  1. ROI: Thanks to the reinforcement model described above, Shift’s solution can begin identifying potential fraud, waste, and abuse cases almost as soon as it’s implemented. Then it speeds up the process of detection and prevention by reducing manual work, increasing SIU effectiveness. 

  2. Do More with Less: Shift’s solution doesn’t require adding data scientists or tech specialists to a Health Plan’s roster, which can be extremely helpful for SIUs working on a budget. 

  3. Explainable Alerts: Gives investigators the reasoning, information and value associated with an alert for fast, meaningful investigations, rather than deciphering statistical outputs to conduct an investigation.

Because of the way Shift’s AI model is trained, it can be deployed quickly, and because Shift’s resources do the heavy lifting on the technical side, no specialized hires are needed to keep it running properly.

3. No Data Prep? No Problem.

Another advantage that Shift’s solution has over most of the market is that it’s simple to deploy.

As pointed out earlier, Health Plans don’t need to add data scientists to their team in order to use Shift’s solution; in fact, all a client needs to get it up and running is access to a web browser. The goal is to make health insurance fraud detection as easy as possible. 

Additionally, many solutions require clients to organize their data in a specific way so that their model can ingest it properly. Instead of forcing a client to structure their data in a particular manner, Shift can accept a client’s data as-is, with no format or pre-mapping requirements. Shift’s experts take client data in whatever form it’s being stored in and get it into an analytic state on their end, which ensures that existing technology and expertise aren’t a barrier to adoption. 

Health Plans have tons of data, and if they don’t have an existing AI or ML solution in place, they likely don’t have the resources needed to make their data machine-friendly. Implementing Shift’s solution only requires giving Shift access to existing data, regardless of what format it’s in. 


Sluder makes it clear that Shift’s goal isn’t to sell a widget, it’s to partner with Health Plans to increase the effectiveness of their investigators. Shift’s solution can help clients pivot from a rules-based or claim policy-level analysis to a more technically mature system without the need to change how their SIU functions.

By deploying this solution, Health Plans can realize more efficient investigations and fight health insurance fraud without restructuring the way they operate.

For more insights, watch or listen to the video podcast.