Omniscience: Empowering Underwriters with Automation

CIO VendorSunil Rawat, Founder & CEO
Underwriting is a fundamental block in the insurance value chain. Yet, according to Sunil Rawat, CEO of Palo Alto-based Omniscience, this area is very technologically sparse and heavily dependent on legacy rules-based systems and human underwriters.

“Rules-based systems suffer from the same issues that most expert systems do: they depend on the interview with the expert, and the expert is typically only able to articulate the heuristics that are top of mind for them, never the intricate patterns they have developed as a result of decades of experience. The result: low STP ratios, lots of time maintaining rules, and high refer to underwriter rates. Human underwriters are no panacea either, with high costs to hire, significant time to proficiency, and high variance in decision making,” says Rawat.

The status quo of rules-based systems and human underwriters is being challenged in the 21st century by urgent global demographic, economic, and technological changes. Developed nations with aging populations are seeing an aging out of expertise. In North America and Western Europe, the market for key insurance types like Life Insurance is in a see-saw growth pattern of +/- 3 percent a year, forcing a need for right-sizing the cost envelope of underwriting operations. The challenge is even greater in burgeoning economies in Asia, where the market is growing in double digits, and penetration rates are very low. To serve the masses of uninsured requires transplanting underwriting expertise from the West through AI systems.

Technological advances force underwriters to look at a milieu of new data from the smart car, smart home, smartwatch, smartphone, social media, plus government data, satellite data, search, the list goes on. But give a human more data sources, and you drastically slow decision making.

Insurance enables people to take risk and new economic activity, we use AI to help insurers take risks intelligently

Omniscience’s roots are in distributed computing and machine learning, having built production grade, internet scale, revenue generating machine learning systems for Comcast, AOL, Hewlett Packard, Nokia, and the US intelligence community. In 2016, Omniscience brought their distributed data mining and machine learning technology to the global insurance industry, building Omniscience Underwriting Automation, an AI-based suite comprising of STP Engine, Auditor, Digitization, Underwriter Advisor, and MegaMeld.

The STP Engine learns from insurer and reinsurer data and provides rapid, consistent, accurate decisions in under a minute per case, and at a fraction of the cost of a human underwriter. Auditor provides detailed explainability for regulatory compliance, while Digitization provides template-free OCR capabilities even on low-resolution forms in Asian languages. Advisor provides decision support to augment human underwriters if an insurer’s risk team still wants a human in the loop, and MegaMeld provides access to a growing library of >4 exabytes/day of data.

Underwriting Automation is just one part of Omniscience’s vision for the new AI-enabled central nervous system for banks and insurers. The company also provides a Capital Management solution, using brute force compute to eliminate replicated portfolios, best estimate liabilities and policy groupings and provide true stochastic-in-stochastic computation for compliance with IFRS 17.

After success in the US, Canada, and Japan, the Omniscience team is looking forward to expanding in other prospective European and Asian economies such as France, Hong Kong, Taiwan, Singapore, and Indonesia. The company is active in Life Insurance, Workers Compensation, Commercial P&C, and Contingent Business Interruption, with R&D efforts focused in Fire, Catastrophe, and other asset classes. “We partner closely with our customers to understand their requirements and add their asset class into our underwriting system to help our customers build new, AI-powered billion-dollar lines of business,” concludes Rawat.