Home > Research > SortSpoke: A Recipe for Turning Unstructured Documents Into Operational Data

SortSpoke: A Recipe for Turning Unstructured Documents Into Operational Data

A different approach to machine learning answers a longstanding problem in financial services – how to efficiently extract critical data from inbound, unstructured documents at 100% data quality. SortSpoke is a machine learning software service that sets up in an afternoon and allows tens of people to do the work of hundreds of manual document reviewers, in less time and with fewer errors.

The data that underwriters and bankers need to process new business applications or review important contracts is sprinkled throughout complex freeform documents, and extracting it manually slows the course of commerce.

SortSpoke founder and CEO Jasper Li saw this problem over and over when he was a management consultant, but no solution on the market, besides “hiring a whole bunch of people,” could meet his clients’ requirements:

  • Quality: 100% data accuracy – 90% accuracy is not good enough when you are processing a mortgage or insurance policy.
  • Flexibility: Able to handle different types of documents in a given business process, with unlimited variations, in multiple languages, while being able to add new fields in the future.
  • Agility: Sets up in an afternoon without consultants or data scientists spending months developing custom models.
  • Continuous learning: Constantly improves through user feedback (rather than a frozen model).

The current practice of using data scientists to “artisanally” develop machine learning models to solve this problem has not provided practical business solutions, as they overpromise on accuracy and are slow to adapt and expensive to maintain.

Li’s flash of insight in 2016 about a new form of neural network, along with his background in machine learning and management consulting, led to a quickly growing company with a disruptive presence in financial services, delivering tangible results rather than overselling AI dreams. SortSpoke helps major banking and financial institutions in the US and Canada from its base in Toronto.

Our Take

In a market full of AI hype, SortSpoke is one of the few practical, business-driven applications of machine learning to solve a very defined and real-life need in financial services. It was designed deliberately to address the key barriers to enterprise adoption of AI: quality and compliance, speed to value, and simplicity of implementation. This is why SortSpoke is attracting attention around the world.

This unique approach, which combines the best of AI with human oversight, gives companies a powerful opportunity to reduce costs, improve customer turnaround time, and address one of the biggest sources of data quality issues.