Smart automation for retail powered by cognitive RPA

Arun Menon,

Manager, Intelligent Robotic Process Automation

Published: January 4, 2018

Innovation in retail banking tends to focus on the customer experience end where banks roll out digital banking platforms that enable customers to conduct multiple transactions online and integrate with online payments services, security service providers, and social networks to make the experience smooth and seamless. Banking back offices, on the other hand, still have multiple processes that are highly manual, cumbersome, and redundant. These inefficiencies rely on various fragmented systems, are document heavy, and have to comply with regulatory and compliance requirements.

These factors can have a significant impact on business. A slow onboarding process can lead to customer abandonment because today's customer currently does not have the patience or time to go through a lengthy process to open accounts or loan approvals. Errors in data entry can lead to time-consuming rework and lead to penalties when they have a regulatory impact. Inefficient customer servicing can lead to dissatisfaction and churn. Operational transformation initiatives need to focus on the end-to-end retail banking flow to derive optimal value.

The traditional approach has been to IT-enable operational workflows, by implementing automated rule-driven systems, digitizing paper-based content, and implementing business process management solutions. The disadvantage of this approach is that it may require huge investments and planning, and it is difficult to predict the success of an implementation, choose the right implementation partners, and the return on investment takes months, sometimes years, to realize. IT automation is also intrusive and impacts existing systems. Traditional IT-based automation is also deterministic. This means that the business rules have to be predefined and operational activities that are human-cognition driven are not partially or completed automated.

The challenge is to make the back office faster and smarter and achieve this with low investment. There are recent advancements in technology that facilitate rapid process automation.

Robotic process automation (RPA) is a recent trend that is rapidly gaining popularity across industries as a cost-effective, rapid means of integrating systems and automating workflows. RPA tools work with the different systems and applications that are part of a process workflow. It can capture and manipulate data, trigger applications, and communicate with systems in a non-intrusive manner.

Data science technologies are based on advanced statistical techniques and are designed to process huge volumes of data, analyze, and derive patterns. The data type can be numbers, images, text, or speech and the output is a probability-based recommendation. Data science platforms deal with large data sets, and the quality of output tends to improve as the availability of data on the problem domain improves.

Cognitive technologies are a set of technologies that attempt to emulate human intelligence using computers. In recent years, we have reached a high degree of maturity due to advancement in computer processing power and better ability of computer systems to process significant data.

In many cases, they are built on and further extend data science platforms. Cognitive technologies include voice and image recognition, optical character recognition, natural language processing, and machine learning. These technologies can complement or even replace most human operator activity in back office processing tasks and can lead to considerable improvements in process efficiency and throughput.

How do you combine these technologies to automate processes?

RPA with a cognitive technology is a powerful combination to achieve optimal end-to-end automation of a business process. A process ideally has a component where the rules can be easily defined and another component where the workflow is too complex to be clearly outlined. The initial part can be tackled by a business process rule-driven robot and the second part can be tackled by a cognitive engine to process complex, unstructured data. There is, of course, a human-driven, decision-making component that cannot be removed from the process.

Related content