Artificial Intelligence is a disruptive technology that has become an important asset for companies, since it supports workers in performing their tasks, limiting their errors and increasing the quality of services. Accounting is a sector in which AI has already been successfully adopted to support business, as it enabled to improve the estimates of costs, the discovery of frauds and the creation of accounting reports. However, most of the accounting activities rely on the accuracy of the booking ledgers, whose filling is a task that is still mostly carried out by humans, especially in SMEs. Accountants who are involved in this process can benefit from the usage of AI for supporting their activities, since it could allow to produce more precise accounting records in a quicker fashion. Towards this end, the Master Thesis aims to combine NLP and ML to create a predictive model able to match banks’ transaction statements with accounting records correctly. To do so, the literature is explored in order to understand what are the techniques suitable for carrying out the project. After that, various models are trained and tested over datasets including the banks’ transaction statements and the corresponding accounting records. Furthermore, the performances of the models are assessed by comparing them against Contofox, a rule generator software, able to classify the accounting records very accurately. Finally, the model business value of the artefact is analysed by means of an interview with a domain-knowledge expert.