Process automation in investment banking

Challenge

An American investment bank wanted to have an automatic solution to process and monitor the content routinely sent to them in statements from their many holdings.
They needed to ingest the identification, status, and changes of the financial instrument into several record-keeping tools.
The customer wanted a solution based on an automated process of machine reading of subscription forms to identify the data required to register a subscription or a commitment.
They asked that the solution be based on the flow monitoring module with a simple user interface for follow-up and manual intervention in case of failure.

Solution

Warrants are filled manually by the clients of the management companies. Each management company then transmits them digitally to the investment bank. Unfortunately, the data is unstructured.
Octopeek uses several OCR templates given the different existing formats they receive. As there is very little labeled data and not enough data to train the models, we opted for proven trained OCR models.
Octopeek did Learning Transfer on already trained models of supervised ML, added “business” dictionaries to improve the reliability of the results.
A confidence score on the quality of the OCR is given at the end of the process. An employee then validates the result by sampling them in a software tool.

Why Octopeek?

The banking industry needs to move from ad-hoc experimentation with AI to enable everybody within the organization to use it to improve both operational efficiency and deliver business value.
Octopeek aims to become the platform that democratizes AI in the enterprise: this strategy makes possible a dynamic creation and provisioning of full-fledged enterprise AI applications, customizable by business analysts.