As large amounts of information are being processed from multiple different sources, manual methods of obtaining data are becoming increasingly challenging for firms. It's more likely that information obtained will be erroneous, complex and unstructured, meaning that pinpointing key information and gaining insight is difficult.
We look at an example of our latest automated system that we developed to deal with these data challenges using various advanced tools and techniques.
We wanted to strengthen our corporate governance review process by automating the collection of annual reports and providing a streamlined method of visualising the data. The process of automation covered a range of factors to enhance our systems, including using structured data-processing and a centralised storage unit that could be easily integrated into our current infrastructure.
We also wanted to highlight the key outcomes of improving our data automation tools. Through a range of testing methods and analysis, we identified the main factors that explained our overall approach:
Our previous method of completing an annual corporate governance review required answering 200+ questions on the latest annual reports produced by each FTSE 250 company. The data would then need to be collected and answered manually by our team and took approximately a day to complete per annual report. They found that this method was time-consuming and had difficulty extracting the strongest data.
Our solution was to automate the collection of annual reports and visualise the results. The dashboard broke down the information by different categories, such as inclusion and diversity, environmental, social and governance (ESG), and audit risk.
To improve the process, we identified a series of solutions that would automate the data and could be integrated easily into our current systems. This required using robotic process automation (RPA), machine learning (ML) and artificial intelligence (AI) to automatically gather the information from the annual reports and extract the latest reports from each company.
The information would need to be stored in an internal network, so the solution was to create an in-house web app to host and manage this information that could be integrated into our current systems. Identifying the key strengths and weaknesses of each company was also necessary to get the most out of the information, so we developed a dashboard that clearly indicated the key metrics and allowed the client to easily distinguish data.
By establishing an automated system of collecting data, we found a range of benefits across the board. Our solution enhanced the metrics by providing more accurate data at a faster rate and, in most cases, the information was processed in five minutes. Incorporating a web application into this process created an improved functionality that allowed the storage of the latest annual reports and made it easier for us to review the information through a centralised method that could be used by the whole team.
Previously, the corporate governance process required a manual system of going through the information individually and using resources to answer the questions. Our feedback showed that an automated system returned a more accurate automation model that gave better information and required fewer overall resources to answer the questions. Our data collation impact was 40 minutes saved per report and 20 minutes saved per report QA, which totalled 32 days of manual work saved across all reports.
Making the most of the system required a clear method of visualisation. The original UI of the corporate governance system was a basic dashboard that had limited visuals and no roll-over for future years, which slowed the process of obtaining data and required more manual observation to distinguish key metrics. Our updated system enhanced the dashboard with a complete picture of our key areas and a method of data comparison to competitors in the industry. This sped up the process of obtaining the relevant data and clearly highlighted the findings of the reports.
Machine Learning Engine
Using AI and Data Science techniques, extracted meaningful insights form the annual reports, Data Science module to automatically identify the Contents Page of the annual Reports using Natural Language Processing. Extracting the contents of the Annual Reports to parse information at different sections of the report.
A combination of Object Detection, Optical Character Recognition (OCR) and Named Entity Recognition is applied to extract domain-specific information (eg Board of Directors page).
Having access to the strongest data is necessary to improve your processes. Building systems that obtain the strongest metrics on a centralised platform is an important step to building an effective infrastructure and streamlining the overall process of reporting.
We can support firms looking to enhance their systems by automating their existing processes and integrating a tailored solution that will provide more accurate data.