After minutes of training, algorithms were able to use 250,000 records of previously unseen raw data to predict regulatory reporting output with degrees of accuracy exceeding 99%.
Wolters Kluwer and PwC have completed a proof-of-concept showing that machines can learn how to take over any end-to-end regulatory reporting process for financial institutions and regulators – across every jurisdiction – by using historical source data and its corresponding regulatory submissions.
“As global regulators impose ever more rigorous reporting obligations on financial institutions, regulatory reporting has become more onerous, with an increased risk of potential error,” the paper says, adding that emerging regulations require more prescriptive and highly granular data sets, as well as increasing reporting frequencies.
“Financial institutions are therefore looking to new technologies, such as ML [machine learning], to relieve these regulatory reporting burdens.”
As outlined in a new whitepaper, the proof-of-concept found that it is possible to build predictive models with high accuracy and flexibility that complement human judgement and oversight, making it likely that production reporting mechanisms will incorporate machine learning in the near future.
The proof-of-concept was trained on two separate end-to-end regulatory reporting processes – MAS 610 and APRA ‘s EFS. After minutes of training, 250,000 records of previously unseen raw data were used to predict regulatory reporting output with degrees of accuracy exceeding 99% in many cases.
“Our PoC shows that machines can indeed learn to take over any end-to-end regulatory reporting process for any financial institution and any regulator in any jurisdiction,” says Wouter Delbaere, Director of APAC Regulatory Reporting for Wolters Kluwer FRR, and one of the authors of the report.
“AI has the potential of disrupting today’s regulatory reporting landscape; rather than taking the traditional approach of explicitly creating deterministic logic, financial institutions can instead adopt machine learning to replace any existing regulatory reporting process with significantly reduced time and effort.”
The paper also covers PwC’s ‘‘Responsible AI Toolkit’, a suite of customisable frameworks, tools and processes designed to help organisations apply AI in an ethical and responsible manner.
The full paper is available here.
Credit: Google News