Leading enterprises across the world are willing and keen to adopt Robotic Process Automation (RPA). However, they encounter many roadblocks in their journey. More often, three major challenges affect effective implementation of RPA in an enterprise
- Fallouts of automation journey, specifically ‘Automatable Exception’
- Manual scheduling leading to lower bot utilization
- Absence of alert mechanism to prevent bot idle time and bot code defects
This has resulted in the need to optimize some of the steps of RPA journey in order to ensure efficient implementation and faster outcomes.
Hurdles in the RPA journey
In many organizations, most processes have exceptions that are not pre-identified and well documented. These organizations also may not have detailed standard operating procedures (SOPs) of their processes and sub processes drawn at a keystroke level. Therefore, they fail to provide relevant documentation to the bot developers. This leads to incomplete design of the code where exception scenarios are not captured – leading to bot errors. Such fallouts tend to be very high, which in turn compels the bot governance team to ask for accurate process maps of the exception scenarios. The governance team is entrusted with the task of increasing the overall bot utilization, process amenability and efficiency on a continuous basis.
Lack of detailed low-level SOPs requires RPA consultants to draw keystroke maps of every process step. The objective is to capture all the steps of the process along with respective timestamps. Then, these keystroke maps are manually assessed for possibility of automation. In most cases, the system and workflows have fields containing both structured and unstructured data; the task of identifying the data type at each step is also a manual process.
Once the process is automated, the task of allocating or deallocating runtime bots is assigned to agents known as bot runners. The task of bot configuration is manual. The run time of the bot is constrained by the availability of human agents and unavailability of bot runners. Therefore, when the bot finishes processing the transactions in one queue, it remains idle until the human agent manually switches the bot over to another queue. Thus, there is a need for a controller engine to bridge this gap by identifying the available or free bot and scheduling them to manage work queues automatically.
Lack of auto allocation and auto de-allocation of bots across queues compels businesses to either create custom code for a controller engine, or assign the responsibility to human agents for running bots as per volumes in the workflow.
Typically, as human agents are entrusted with multiple tasks, they do not sit in front of the console to monitor the performance of the runtime bots continuously. Therefore, when the bot code is defective, the runtime of the bots becomes skewed significantly. The buggy bot may end up in an infinite loop, or its faulty memory management design might cause the system to freeze. If there are no auto- identification of such faulty situations with auto- email alert mechanism, the bots remain silent. However, a trigger – like an email, SMS or tweet – alerting the human agent to reboot the bot machine immediately helps address the memory freeze. Otherwise, bot utilization suffers even though the bot is readily available for reuse.
Agile route to automation
In order to accelerate process automation, the traditional RPA deployment approach has four phases: