HSBC is building an AI-driven utility

HSBC is building an Artificial Intelligence (AI)-powered client intelligence utility (CIU) that is using 10 petabytes of corporate and institutional client data from its 1.6 million clients.
27 December 2018

A pedestrian walks past a HSBC UK bank branch in central London. (Photo by Tolga Akmen / AFP)

In what can be classified as a mammoth undertaking, HSBC is in the process of building an Artificial Intelligence (AI)-powered client intelligence utility (CIU) that is using 10 petabytes of corporate and institutional client data from its 1.6 million clients as a foundation.

The bank has created 22,000 physical tables from information it collected over the past year from across the 67 countries it operates in.

HSBC UK’s Chief Administrative Officer & Head of Transformation for Global Banking and Markets, Chuck Teixeira, is leading the effort and is seeking help from AI specialists to develop the utility.

Speaking to CBR, he said the bank is intent on delivering “Client 720”, a data asset, that will be a trailblazer as it will the first kind of creation that provides an automated bird’s eye view of client activity.

The bank intends to use the CIU to customize the structure of products and provide predictions into the effects of macroeconomic and geopolitical events that can have a bearing on a client’s global risk profile.

These novel techniques will aid the bank in their efforts to hedge exposure and automate compliance efforts. Loans, Foreign Exchange services or assistance in entering new markets will also be supported via the system.

The bank currently handles millions of transactions daily, which are sourced from 200 disparate systems in 66 jurisdictions that have 14 major product lines and a whopping 75,000 fields of data.

HSBC’s CIU could probably run on the cloud, according to Teixeira.

The data is now housed on an on-premise Hadoop environment as ‘real reusable data assets’. Data quality is measured on five different dimensions, namely accuracy, completeness, uniqueness, validity and consistency.

Teixeira mentions that Machine Learning (ML) is then used to link transactions across different clients. Initially, these techniques were utilized for ‘financial crime use cases’, but the next step is to optimize client services by harnessing this newfangled methodology.

Commonly, getting a project like this off the ground poses a range of challenges. Chief among them are compliance concerns, legal issues and struggles between data teams with regards to training of algorithms.

This was overcome by creating a physically shared space that housed the teams in charge of data science, engineering, IT as well as a data sharing, privacy, and ethics team.

Migrating this whole process to the cloud is being touted as the next evolutionary step, but Teixeira is aware that meeting the legal requirements will be the first hurdle to overcome to make this step a reality.

The bank is also working with CognitionX, an AI advice platform, as part of a request information process with a select few AI firms to evaluate their prowess in deconstructing data and whether they have what it takes to create machine learning software that is scalable.

Speaking to Forbes, Teixeira also cites the importance of data quality and cleaning to enable the full use of analytics and machine learning. “We’ve used machine learning itself to join up that data and link it all together at the most granular transaction level,” he said.

When quizzed about the ethical use of data, the cultural aspect and safety concerns that abound, he is adamant that reinforcing data science and engineering capabilities will not lead to any compromises linked to data sharing privacy and ethics.