Refined and ready: Introducing Risk Data Refinery—from raw submission to EDMs with AI

Authors: Cihan Biyikoglu, Managing Director - Product and Technology, Moody's; Alok Kumar, Managing Director - Analytical Services, Moody's

Underwriters know that precision is vital; get an address slightly wrong, miss a contract term, ignore an important exposure attribute, and you could find yourself creating expensive mistakes further down the line.

That’s why for many in the insurance industry, extracting, validating, and enriching data from unstructured or semi-structured sources has been a long-term issue. Insurance slips or Statement of Values (SoVs) move through underwriting, catastrophe modeling, and portfolio management systems, all traveling between you, your broker, and your reinsurance partners.

This all requires several sophisticated and time-consuming data transformations, and each stage is critical to get right. The work has traditionally involved large teams of internal experts to prepare the data, sometimes manually or in semi-automated ways, or requires large outsourced teams to help fulfill the need.

For everyone in this situation, the good news is that all this is about to change for good. We have been building a new AI-powered application on Moody’s Intelligent Risk Platform™ (IRP) called ‘Risk Data Refinery,’ which converts your raw submission data into Exposure Data Modules (EDMs), our structured data container holding detailed exposure, financial, and contract information.

The core AI processing in Risk Data Refinery has been purpose-built by underwriting data experts who have done data cleanup and transformation nearly a million times over, for more than 15 years. That means we were able to incorporate a huge list of edge cases that many others don’t consider.  

Risk Data Refinery is available now as a preview for you to experience and try on your own slips and SoVs.

Introducing Risk Data Refinery on the Moody's Intelligent Risk Platform

What is Risk Data Refinery? It is our new agentic application that allows users to structure complex property insurance submissions into an analytics-ready format. Just upload your slips or SoVs into Risk Data Refinery, and it turns them into structured data ready for property underwriting and modeling analytics. First-pass interpretation accuracy is ~95%, and with an expert in the loop, it can get near 100% within minutes.

How does Risk Data Refinery compare with other solutions used to process SOVs and insurance slips? We believe that Moody's Risk Data Refinery stands out by delivering superior and more efficient processing, driven by three key factors:

1. Training data: Risk Data Refinery on the IRP is powered by multiple AI agents built utilizing decades of work from Moody’s Analytical Services. Our Analytical Services (AS) team has been processing more than 50,000 complex data submissions each year, and is trusted and known by clients for their data quality and precision. The AS team has seen every type of submission across all format types. Training our AI agents on a large and diverse real-world dataset enables Risk Data Refinery to achieve 95% accuracy in data conversion during testing.
2. Continuous learning: Perhaps more importantly, the AI agents utilized in Risk Data Refinery continue to learn from our experts in the Analytical Services teams, who deliver AI-assisted services to our insurance customers. Where Risk Data Refinery comes short on a unique new ‘edge’ case, Analytical Services' continued feedback is translated into improvements for the agents utilized by Risk Data Refinery.
3. Platform integration: By integrating slip and SOV processing into Moody’s Intelligent Risk Platform—whether through a standalone application, Risk Modeler™, UnderwriteIQ™, ExposureIQ™, via an API, or Model Context Protocol (MCP) server—customers gain seamless access to critical data in the format that best suits their needs. This flexibility reduces the burden of manual data handling, enhances workflow efficiency, and enables greater automation, empowering users to focus on higher-value tasks.

How does Risk Data Refinery work?

Risk Data Refinery takes in your slips and SoVs in formats such as Microsoft Excel, text, CSV, or PDF, and across multiple languages—not just English. Then, using various AI agents such as an address agent, slip agent, SoV agent, data enrichment agent, policy coding agent, etc., all extract structured details from the details you uploaded. Risk Data Refinery then uses Moody’s Enhanced Risk Data to build rich analytics-ready data in EDM format for input into property underwriting and portfolio management workflows. 

The process takes just seconds to minutes to get clean, structured analytics-ready data. Risk Data Refinery can either be fully automated or allow you to supervise. In the supervised mode, you can inject your own changes and corrections. However, our early testing suggests that, in most cases, you won’t need to supervise, as the success rate for our Analytical Services teams today exceeds 95% accuracy.

When can I try Risk Data Refinery with my own data?

You can try today, as Risk Data Refinery is now available as a preview. We expect to deliver the generally available capability shortly.

Try the new Risk Data Refinery preview in your own Moody’s Intelligent Risk Platform instance under Risk Labs. Just ask your Moody’s representative to enable the experience for you, and we’ll switch it on in your environment. Once Risk Data Refinery is enabled on the Intelligent Risk Platform, you can try your own slips and SoVs to see the experience and observe the accuracy on your own data. 

With Risk Labs, you can use your login on the Intelligent Risk Platform to access the previews. Please note that Risk Labs provides a safe experimentation surface for us to deliver early previews of agents, applications, and experiences to collect customer feedback. It strictly isolates your production data and operations from Risk Labs data and operations, so your live production data isn’t accessible directly to Risk Labs preview and vice versa.

Hear more about Risk Data Refinery and other AI applications and agents we are developing at Exceedance 2026 (June 1-4, Fort Lauderdale, FL), find out more and register here.