Natural language processing (NLP) is an area of computer science and artificial intelligence (AI) that aims to enable computers to process large amounts of natural (or human) language data.
NLP could significantly empower the audit, as it would enable auditors to analyse unstructured data. Structured data – found in spreadsheets and ledgers – can already be comprehensively analysed using our data and analytics (D&A) and automated capabilities. But more than 80 percent1 of data today is in unstructured formats such as contracts, emails, PDFs and other documents. A key battleground is to develop digital assistants that can read this data and identify key information. Having a bot, for example, analyse the accuracy of one of those unstructured files. The development of NLP capabilities to read emails is another example. By using the processing power of intelligent machines, we can use correlation theory to extract data from unstructured sources.
Once the technology has been instructed on what to look for by the auditor, NLP could read emails and other documents to search and identify information (also utilising optical character recognition technology that can ‘read’ documents such as PDFs). The difference between the technology doing this and a person is scale – an NLP application could read thousands to millions of documents in a fraction of the time it would take a human to perform the same task.
The benefit of NLP capabilities when combined with robotic process automation (RPA) technologies can be monumental. As one example from KPMG in the US, the ability to analyse 100 percent of revenue or purchase transactions through RPA allows the auditor to make judgments about the areas of risk, as well as help in the identification of potential outliers and exceptions.
This, by itself, does not complete the audit process. This is one critical step in the audit process; however, auditors also have to consider source documents (e.g. invoices) to determine if the reliability and relevance of the underlying data were supported by the facts. The role of an auditor vouching the details of a financial record to the source file has been a central principle of an auditor’s procedures.
The marriage of NLP technologies allows machines to read those unstructured data sources and create the digital files that can be compared to the transactional records. The speed and effectiveness of such procedures will allow auditors to move quickly through those transactions determined to be supported and those that are not, allowing even more time to focus on areas requiring more critical judgments and time.
Many forms of new technology are optimised when working in conjunction with each other, which is the case for NLP. NLP would find and extract key data from documents, and then put it into a format that a machine-learning tool could analyse.
The full extent of NLP’s capabilities is still being realised, and the potential it holds to enable an auditor to look for and find audit evidence in emails and other documents is potentially hugely significant.
NLP, working in combination with other new technologies, could have an important role in improving audit quality. The power of NLP, together with robotics, machine learning and, in time, deep learning will mean that an audit may become deeper and further-reaching than ever before, based on increasingly granular and sophisticated analysis of data. Combined with the integration of our applications into KPMG Clara and its flexible deployment in the cloud, KPMG is well placed to develop ever more powerful capabilities – helping drive consistency across the network and enhancing service to cross-border businesses.