Decoding Robotic Process Automation For The Indian Market
Sunil Aryan, Practice & Sales Lead, Verint Asia
In a growing economy and hyper-competitive market, Indian businesses are faced with the challenges of achieving greater customer experience, higher speed of execution, and reduced risk of errors & process compliance while reducing operational costs. Indian verticals like Telcos, BFSI and even PSUs have large back-offices, where bulk of the work gets done. India also happens to be one of the best back-office/ shared-services centre of the world. In back-offices typically, the work to be done by workforce are a series of tasks such as –
• Functions requiring intuition, perception, empathy, decision making, situation awareness, and even learning.
• Desktop level data-oriented functions that are rules-driven, repeatable and mundane. It is estimated that back-office employees spend almost 40 percent of their time in this type of data handling function. Robotic functions are much more adept at handling this part of work than humans, as they can continue to work 24x7 with least errors and without hints of boredom. Naturally, businesses have widely adopted Robotic Process Automation (RPA) to take-over this latter boring minion part of the work that the employees anyway despise.
RPA’s benefits generally tend to cascade beyond the tasks that have been automated. For example, processes with stringent maker-checker model can automate and reduce the ‘make’ team in the process. Since RPA does not generate errors and if they do, they can flag-out such tasks, it may be possible to substantially reduce the checker team as well. Indian back-offices have disparate back end IT systems. Since RPA works at desktop level, it can be used to unify systems at the presentation layer without requiring deep back end integration work.
Automating business process requires careful change management at People and Process level. Managing people and their expectations is key to acceptance & success of RPA in any business unit. It is natural for employees to be wary of the new digital workforce that works tirelessly 24x7 with minimal if not zero errors. While it is true that RPA would not be creating mass unemployment in near future and would actually help in creating new ones, certain short-term (collateral) damage is inevitable. For growing organizations, the extra FTE hours generated might just feed the growth. Most organizations with planned RPA endeavours need to be prepared for the change, and should plan to reskill & reallocate their employees. The new digital workforce can be ushered-in as a co-worker and maybe even as assistants for employees rather than as their replacement. Communicating that RPA role is to cover ‘task’ automation, which is only part of the process rather than entire process itself, would go a long way in avoiding negative sentiment among the employees.
A lot depends on the process side too that requires important decisions to be made in setting the right objectives, choosing the right process, and re-engineering of the process. Most organizations that have undertaken RPA projects seem to present results on the number of FTEs they have saved. Others have realised that RPA can have broader impact if envisioned as a tool for digital optimization journey with focus being business efficacy, with efficiency as an important by-product. Seeking operational savings from efficiency is not necessarily bad, but restricting vision to it might result in loss of broader gains.
Even getting to lower hanging fruit of efficiency requires processes to be broken into smaller tasks, where bots do the basic tasks, while employees focus on decision and execution. A shared service centre of one of the largest logistics company achieved almost four FTE worth returns per RPA bot by using RPA as part of their six
sigma project and pushing only the most basic tasks to automation. Employee productivity shot-up once these data collation and handling tasks were moved to bots. All became possible because they undertook process re-engineering as part of Six Sigma project and gave special attention to choice of process with tasks that were automation friendly.
Today, we have choice of how we would like to use the digital workforce. Any of the three approaches below can be used based on the nature of work, the complexity of the process tasks, and how much human factor is required for process efficacy.
• Unattended: In this mode, the RPA bots run as instances on centralized servers, where they execute the work allocated to them. Just like human workforce, this digital workforce resident on servers has multiple skill types to handle different types of work tasks. For effective utilization of this virtual digital workforce, the work tasks need to be presented to the RPA units in structured manner, and their output needs to be moved down the line to the next team with minimal delay for maximum returns. In a multi-task/ multi–stage process, bots generating errors or failing are likely to disrupt the line balancing and require close monitoring for their up time and outputs.
• Attended: RPA works alongside employees, offering assistance and awaits activation by the user or an application trigger. The attended bot can be triggered on demand to execute a sub task (doit mode) or show the human worker how a task is done (show-me mode). RPA here acts like an assistant or an on-the-job trainer. The attended bot can also help ensure employees’ compliance to rules. For example, it can monitor the work being done on the machine and instruct/obstruct the employee from approving transactions of values that are above their authorized limits.
• Hybrid: This model allows digital and human workforce to collaborate by seamlessly passing different work tasks among themselves. For example, while processing a loan, an employee might use the Assisted RPA to fetch data from multiple systems and take decision using approval criteria. With the decision made, the following tasks are sent to Unattended RPA pool for update in the requisite systems and generating the paperwork for dispatch. Employees can monitor the status of work that is delegated to the RPA units. This approach works great for processes with multiple business decision points and data driven subtasks. The pooling of digital workforce allows the robots and humans to work in their domains, contributing to process effectiveness and shorter work cycle times. The efficiencies obtained with this approach are generally larger than standalone RPA approach.
With multiple models of RPA going mainstream, for many early adopters, the benefits of this solution have already begun to flatten-out. They are already looking at Cognitive RPA (CRPA) as the next evolutionary step towards better service and cost models. While RPA focuses on replicating small tasks with capability restricted to structured information, cognitive part of CRPA can work with unstructured data, take decision forks and then execute, all the while learning from the process. While RPA in its many forms is able to execute tasks on our behalf, CRPA is capable of doing multiple tasks and can present us with inferences, alerts and opportunities from the work done.
Cognitive RPA can be seen as application of ‘narrow AI’ on top of the RPA function. We use the term narrow AI because the current crop of RPA have their cognitive functions focused on specific type functions like reading emails, handwritten forms, recognition of patterns like images & maps, and executing specific logic paths. Because the strength of CRPA is it being a learning system, it requires sufficient data sample to learn from, and this can in some environments be a shortcoming.
With the rising capabilities of CRPA giving it wider execution playfield, it has the potential to one day break-down the walls between the front-office and the back-office. The CRPA force sitting in corporate network might complete much of the work at the front office stage itself.
Today, we have choice of how we would like to use the digital workforce. Any of the three approaches below can be used based on the nature of work, the complexity of the process tasks, and how much human factor is required for process efficacy.
• Unattended: In this mode, the RPA bots run as instances on centralized servers, where they execute the work allocated to them. Just like human workforce, this digital workforce resident on servers has multiple skill types to handle different types of work tasks. For effective utilization of this virtual digital workforce, the work tasks need to be presented to the RPA units in structured manner, and their output needs to be moved down the line to the next team with minimal delay for maximum returns. In a multi-task/ multi–stage process, bots generating errors or failing are likely to disrupt the line balancing and require close monitoring for their up time and outputs.
Because the strength of CRPA is it being a learning system, it requires sufficient data sample to learn from, and this can in some environments be a shortcoming
• Attended: RPA works alongside employees, offering assistance and awaits activation by the user or an application trigger. The attended bot can be triggered on demand to execute a sub task (doit mode) or show the human worker how a task is done (show-me mode). RPA here acts like an assistant or an on-the-job trainer. The attended bot can also help ensure employees’ compliance to rules. For example, it can monitor the work being done on the machine and instruct/obstruct the employee from approving transactions of values that are above their authorized limits.
• Hybrid: This model allows digital and human workforce to collaborate by seamlessly passing different work tasks among themselves. For example, while processing a loan, an employee might use the Assisted RPA to fetch data from multiple systems and take decision using approval criteria. With the decision made, the following tasks are sent to Unattended RPA pool for update in the requisite systems and generating the paperwork for dispatch. Employees can monitor the status of work that is delegated to the RPA units. This approach works great for processes with multiple business decision points and data driven subtasks. The pooling of digital workforce allows the robots and humans to work in their domains, contributing to process effectiveness and shorter work cycle times. The efficiencies obtained with this approach are generally larger than standalone RPA approach.
With multiple models of RPA going mainstream, for many early adopters, the benefits of this solution have already begun to flatten-out. They are already looking at Cognitive RPA (CRPA) as the next evolutionary step towards better service and cost models. While RPA focuses on replicating small tasks with capability restricted to structured information, cognitive part of CRPA can work with unstructured data, take decision forks and then execute, all the while learning from the process. While RPA in its many forms is able to execute tasks on our behalf, CRPA is capable of doing multiple tasks and can present us with inferences, alerts and opportunities from the work done.
Cognitive RPA can be seen as application of ‘narrow AI’ on top of the RPA function. We use the term narrow AI because the current crop of RPA have their cognitive functions focused on specific type functions like reading emails, handwritten forms, recognition of patterns like images & maps, and executing specific logic paths. Because the strength of CRPA is it being a learning system, it requires sufficient data sample to learn from, and this can in some environments be a shortcoming.
With the rising capabilities of CRPA giving it wider execution playfield, it has the potential to one day break-down the walls between the front-office and the back-office. The CRPA force sitting in corporate network might complete much of the work at the front office stage itself.