Shift Technology helps hundreds of insurers fight claims fraud using Azure OpenAI Service

Acting quickly with AI and machine learning to support better, faster insurance decisions

Ask anyone at Shift Technology what makes them proud to work for the company, and they’ll probably answer that the work they do positively affects their customers and the global insurance industry on a daily basis. The company’s AI-powered decision-optimization solutions and fraud-detection tools help the world’s leading insurers mitigate fraud and risk across policy life cycles, streamline claims processing, increase operational efficiency, and deliver superior customer experiences.

For this to be possible and sustainable, AI and machine learning have always been part of Shift Technology’s DNA, starting with its initial use case for claims fraud–detection. Insurance companies regularly struggle with costly and complex fraud-detection challenges, especially as fraudsters continue trying to outsmart systems and standards. While technology evolves to help detect fraud, so does the technology that enables fraudsters to cause more harm—often ​​billions of dollars a year.

Shift Technology prioritizes staying at the forefront of AI and machine learning, and it finds engaging and innovative ways to drive positive outcomes with ​​evolving fraud schemes and other obstacles for insurance companies and other clients using its platform. Because so much of its business begins and ends with data, ​​the company’s data scientists work closely with customers to understand how they operate and help them get the best out of its platform. Insurers can then make smart, informed decisions based on their data—and most importantly, Shift Technology’s analysis of it.

To provide an easy on-ramp for clients, Shift Technology standardized its platform on Microsoft Azure. The result is strong, effective, and lean cloud operations with which engineering and data science teams can take full advantage of new capabilities, regardless of where they are. “Given the global footprint of Azure and its overall capabilities, the core of the Shift platform runs with a Microsoft technology stack that allows us to operate and innovate simultaneously,” says Marc Jones, Chief Technology Officer at Shift Technology.

A prime example of this innovation is the company’s use of Azure OpenAI Service, found within Azure AI Foundry, and Azure AI infrastructure to hone its generative AI capabilities and help make insurance decisions quick, accurate, and transparent. “Because Microsoft is on the forefront of enabling generative AI adoption and we’re on Azure, we can use Azure OpenAI to move very quickly from research to true proof of concept and then straight to production,” adds Jones.

Consolidating weeks of policy, claim, and documentation digestion into days

With each of its solutions, including recently introduced tools for automating ​claims​, mitigating underwriting risk, and improving subrogation recovery, Shift Technology processes and generates mass amounts of documents supplied in varying formats, languages, and structures. In fact, it’s already analyzed more than 2.6 billion policies and claims and their supporting documentation. Historically, staff would face weeks of data labeling and model training to provide useful outputs. “A fair amount of unstructured data comes into our platform as ​​​​documents, so being able to automate information classification and extraction with a high degree of accuracy is critical,” says Jones.

Using machine learning–based optical character recognition (OCR) with cloud-based ​​​​Azure AI Vision, also found in Azure AI Foundry, and both OCR and layout detection with Azure AI Document Intelligence, Shift Technology simplifies and streamlines document classification and can quickly extract pertinent information. “Reducing weeks of work to days, we’re now able to maximize Azure OpenAI and focus on the prompt engineering that allows us to extract data from those documents,” continues Jones. “We can then fine-tune prompts to get the output that we had previously in a way that’s faster and better.” In doing so, it’s built a library of prompts for future prompt engineering, and it continuously fine-tunes the prompts and applies them to other document types.