Ingestion Redefined: Accelerating the Underwriting Workflow Through the Combined Power of AI and Human Oversight

Boosting Insurance Risk Assessment with Consolidated Analytics

Today’s commercial P&C underwriters are under significant pressure to make better decisions, increase efficiency and write more of the right business – all of which require easy access to accurate, high-quality data. Unfortunately, the current state of submission ingestion processes at many insurers makes this very difficult to achieve.

At most organizations, these processes remain highly manual – they require significant effort to manage documents, identify the relevant data points and input them into the right systems. The result: underwriters, underwriting assistants, underwriting technicians and operations staff spending far too much time making sense of complex information trapped within email inboxes, PDF readers and Excel spreadsheets. It’s a laborious, error-prone workflow that pulls vital resources away from the more strategic aspects of underwriting – particularly when initial tasks, such as clearance, are treated the same as more involved steps, such as poring over loss data. Previous efforts to evolve this process, such as leveraging optical character recognition (OCR) technology to help with basic templating, have fallen short of solving the problem due to the high degree of variance and variability in how the content is structured and presented.

We firmly believe that humans have a key role to play in the underwriting workflow – but submission ingestion is one area where that role can be greatly reduced. By leveraging a variety of AI techniques, insurers can automate much of the initial extraction, enabling insurance professionals to go beyond that first step and serve as quality control artists as opposed to data entry specialists. By turning a slow, variable process into a fast, accurate one, these organizations can increase productivity and ensure all downstream decisions are based on accurate data, giving them a trustworthy source of business intelligence.

How can insurers achieve that vision? For Insurance Quantified and our Groundspeed intelligent document processing solution, it all comes down to years of proven experience in the commercial insurance industry, our advanced technology and our human-in-the-loop process.

Combining Human and Machine Intelligence

Just like how underwriting professionals who rely on manual workflows cannot compete with modern systems, fully automated extraction solutions cannot currently address the full breadth of submissions while maintaining a very high degree of quality, resulting in an overall lack of trust in the data. For success in today’s ever-shifting technology landscape, organizations need the best of both worlds. Humans remain essential to fostering trust in data and decisions – it’s just a matter of matching those human resources to the appropriate need.

That requires the system to have a keen understanding of where processes may have gone sideways. In our case, Groundspeed performs in-production quality checks and generates human tasks for unexpected outcomes that need review. For example, our system automatically classifies, annotates, cleans and structures data from relevant documents – but if a given document does not align with expectations, such as a loss run document that appears to be missing certain fields or business rule checks that don’t pass, the discrepancy is flagged for human operators, who can look to see if there’s something amiss and make any necessary corrections. This exception management process brings people into the workflow only where they are needed.

AI automation and smart human-in-the-loop processes are a big part of why Groundspeed intelligent document processing is an industry-leading solution, but our ability to fit into the underwriting workflow at multiple points is crucial as well. Rules-based processing allows us to accelerate the entire submission lifecycle by tailoring the data extracted to the relevant stage of underwriting. For example, rapidly processing the most critical fields for clearance using pure automation enables the removal of irrelevant or out-of-appetite submissions from consideration as soon as possible. More detailed ingestion work around ratings and loss run data occurs later on, and only for the right opportunities. By complementing data extraction with data triage, organizations can account for differing requirements around the speed of data delivery, optimizing workflows and costs.

We’re continuously auditing our models for accuracy, making regular enhancements to ensure our automated submission ingestion is as comprehensive as possible. We also flag lower-confidence data and encourage team members to perform their own audit processes and make corrections, with all interventions then incorporated into our models. While our system performs the crucial grunt work, its processes are overseen and refined by human experts and end users at every step.

All of this requires significant industry- and domain-specific knowledge since there’s no generalized solution for extracting data from documents of this complexity (yet — we’re working on it!). Developing a comprehensive, production-ready solution requires extensive foundational work and deep familiarity with the relevant processes, workflows and document types, as well as enrichment with third-party data. That’s why Groundspeed handles everything from PDF forms and natural language emails to highly technical and bespoke Excel spreadsheets with tables and visualizations, with all processes tailored to the world’s most high-stakes commercial P&C industries.

Through this powerful combination – AI-based automation, intuitive human workflows and domain expertise – our submission ingestion supports the rapid conversion of data into accurate business intelligence that organizations can confidently translate into action. 

Making an Impact

Why is this important? With the workflow complexities outlined above, it can take a very long time for a submission to go from the initial email to full completion, leading to missed opportunities. Hurdles like inaccurate data or wasted operational work on out-of-appetite risks only magnify the challenge. Through a consistent process defined by speed, quality and user-friendliness, insurers can drive superior results while better supporting their employees throughout the underwriting chain.

For the end users of the data – underwriters, underwriting assistants, operations groups and the like – it’s about better process efficiency and visibility. These employees are under constant pressure to write good business without sacrificing speed, so giving them the tools they need to find their edge is crucial. For underwriting managers and department heads, it’s about process transformation – improving metrics around ingestion time, conversion rate and more are essential for reaching their strategic goals. Having trust in submission data is the first prerequisite for every step in this process.

The pressure to innovate is mounting. As insurers increasingly embrace technological advancements and workflow optimization, the threat of losing business to competitors that are better prepared to rapidly process submissions will only grow. Redefining ingestion is the surest path to going on the offensive – driving greater efficiency throughout the underwriting lifecycle and leading to more wins.

Want to learn more? Drop us a line.