Artificial intelligence (AI) is the hottest topic in the corporate world and it is affecting a plethora of areas in our lives. Tasks that once took days to perform manually are now being replaced with automations, reducing cycle times to hours or even minutes.
Operational risk management (ORM) is no exception. FinTechs and financial institutions are introducing AI applications in ORM in a limited way. These applications are also used in other areas of operations that typically require vast amounts of data being handled manually.
Capgemini is engaged in work on an operational risk project within a global financial institution. To our experience, ORM is often deemed as an administrative task and is required due to internal audits or external regulatory pressures from various domestic and international regulators.
In one of the work streams, we leveraged AI technology to help our client build new taxonomies for their risk and process libraries. Historically, every region of this financial institution had their own version of risk and process taxonomies, due to lack of a centralized governance. The overall operational risk library contained a redundancy of specific risks and processes created at the local level. With the help of natural language processing (NLP), we aimed at rationalizing and standardizing the risk and process library so that every region would utilize and apply the same risk and process libraries going forward.
There was a total of 5,000+ risk taxonomies and 6,000+ process taxonomies across the different geographies and not all taxonomies were documented in English. The NLP technology first translated all the taxonomies from 13 different languages into a single language (English). The system then cleaned up the taxonomies database by identifying similarities among taxonomies. Based on the level of similarity, the AI solution created three different categories:
- For taxonomies having more than 80% similarity, the system automatically deemed those taxonomies identical and did not need require human intervention.
- For taxonomies between 50% and 80% similar, the AI categorized those taxonomies as needing some human intervention. It highlighted the differences and asked the human to make the final judgment.
- For taxonomies below 50% similarity, the system identified those taxonomies as completely different and required them to be checked thoroughly by a human.
As a result, the NLP solution narrowed down the list from 5,000+ risk and 6,000+ process taxonomies to 250 and 190 respectively, which subsequently served as the basis to conduct the risk control self-assessment (RCSA).
AI has been continuously used in the same project and another example was to link the existing stock of internal and external historical incidents (HIs) (10,000+ HIs globally) with the new standardized risk taxonomy defined from the example previously mentioned. The format of a HI is very similar to a risk event as it is also described in a few lines of wording. Same as the above example, depending on the level of similarity, NLP technology can detect the common words for both HI and risk event taxonomy and make the linkage between the HI and risk event based on the similarity. If there is a high number of relevant words that are the same, the NLP will automatically link the historical incident with the corresponding risk event. Through this link to the historical incidents, this will significantly improve the efficiency and help our client identify and prevent similar risks in the future.
The above examples and the success of continuous application of AI technology rendered in this “non-client facing” area demonstrated the increasing use of artificial intelligence (AI) outside the purely customer experience focused parts of an FS organizations. NLP, machine learning, and other AI technologies will start to impact and change areas such as operational risk management, also changing the way financial services organizations will manage their risk going forward. In the near future, front office sales, traders, asset managers and risk managers will be able to leverage artificial intelligence (AI)-based platforms to monitor the operational risk exposure of the entities they face more efficiently by sweeping through the internal database and internet for all types of risk information. With help from AI and machine learning technology, they will be able to identify, analyze, monitor, and prevent risks in the blink of an eye.
To learn more, feel free to get in touch with me on social media.