This paper describes the use made of AI/ML at Banca d'Italia in the management of statistical and supervisory reports. It summarizes the results of the studies carried out to improve data validation and enrichment, as well as to increase the efficiency of operational processes, most of which have been implemented in the actual pipelines. The lessons learnt are also thoroughly presented and the areas of current and future research are briefly outlined.
Applying AI/ML to regulatory reports management requires specialized know-how, diversified teams and the necessary IT infrastructure. At the beginning, it is useful to stimulate the creativity of the staff involved in the operational processes. Attention to transparency in communicating the results is also key and a careful evaluation of costs and benefits should always be considered. In a nutshell, the approach is: start small, get results, invest in enabling factors, and then scale up.
