Thank you for Subscribing to CIO Applications Europe Weekly Brief
How to Get Most Out of Robotic Process Automation (RPA)
By Joyjit Gorain, Operational Performance Management, RPA (Design, Architecture & Deployment), Zurich Insurance Company Ltd.
First the basics–what is RPA?
In few words and to put it simply, RPA is about mimicking human activity (including screen navigation, mouse-clicks, editing/copying/pasting of text fields, selection of field values, etc.) by an embedded software (call it Robot) in order to complete the same activity. These human activities are usually purely rules-based and needs defining with a set of steps in order to design the robot workflow. RPA makes data and information processing on applications faster, improves quality and reduces human overhead in repetitive tasks.
Naturally, it also comes with its own exception handling scheme, where incidence queues (cases or transactions which a Robot is unable to process) are managed in either of the two categories:
- Cases requiring partial manual intervention due to a missing data/information and can be re-run by the Robot
- Cases requiring full manual intervention due to data readability issues or previously unforeseen scenarios that has not been programmed into the Robotic workflow
The challenges of RPA
The fundamental block is often the lack of process documentation and SOPs, without which there will never be an eligible process to imitate in the first place (remember: a Robot in RPA context cannot read information from inside our brains, it can only read and record what we exhibit with our hands in forms of key strokes and navigations on screens). However, this is something that is increasingly being addressed in organizations by better training SMEs, enabling back and mid-office workers with recorder tools and documentation software, re-designing of tasks and activities in a structured/standardized manner.
Once a supposedly illegible character/text/ data is picked up by the OCR component of such a tool, it can be further processed by the machine’s cognitive capabilities to make sense and identify an appropriate construct out of those characters/texts
There are however bigger impediments to RPA which is more likely to hold back full realization of benefits and needs a more concreted strategic approach. To summarize:
1. Lack of coherent and strategic process architecture, leading to too fragmented process landscape–too many dissimilar processes across an eclectic range of systems and applications
2. Upfront data readability issues, leading to low automation feasibility of the rest of the process downstream
3. Continued usage of RPA for low-hanging (because it is easy to do) discrete task-oriented automation, rather than full end-to-end automation coupled with cognitive automation capabilities through like Artificial Intelligence and Machine Learning (AI/ML)
Addressing of RPA challenges
The solution to the first of the above challenges goes down to the adoption of a more strategic mind-set by organizations, where they need to fix processes themselves before putting a Robot on top. It needs a meticulous review of its entire process and application landscape, identification of process clusters which could be treated as “one and similar” and establishing of a standardized way of organizing the tasks and activities around such clusters. Coming from Insurance industry, one example could be to have ‘a single standardized way of processing claims for all retail Motor claims’, with acceptable notions of difference only (e.g. claims processing in Europe can be different than in Asia due to regulatory reasons) but even such differences should be clearly laid out as adjustable parameters. Such a clear modularization of “one and similar” processes exponentially scales-up the potential due to the easy replication of solution from one part of the business to another.
The second of the above challenges has both ‘tough’ and ‘easy’ ways out. The hard way calls for upfront investments into digitization of customer touch-points and modernization of legacy systems in order to ensure a more systematic and structured data at various channels of engagement. This of course needs strategic and systematic investments, because in the end it will not just be about enriching data and boosting RPA feasibility but about a much larger impact on customer front. The relatively easy way out would however be to augment data generation sources with intelligent character recognition tools, integrated with cognitive intelligence. Once a supposedly illegible character/text/data is picked up by the OCR component of such a tool, it can be further processed by the machine’s cognitive capabilities (trained with supervised or un-supervised training data-set) to make sense and identify an appropriate construct (e.g. words, sentences) out of those characters/texts.
The third and the final road-block has always been in using of RPA in small discrete process parts (task-focused approach) rather than truly scaling it up with a complete end-to-end philosophy by suitably augmenting it with other wider technologies and data science solution suits. Parts of a process value chain which are not a matter of simple binary or rules-based data processing can be processed through advanced deep-learning AI techniques and then fed into an RPA downstream. An AI/ ML augmented RPA solution like this can help organization attain true scale of automation, exponentially enabling its customers and internal workforce.