While more IT investment is being directed at robotic process automation (RPA) than ever before - even more than is spent on cloud, IOT and analytics - this could significantly increase as RPA becomes more intelligent and moves towards cognitive automation.
This is the view of Rajeev Mishra, process automation offering lead at Johannesburg-based systems integration and business process firm, Ovations.
He says although most industries are becoming increasingly willing to invest in RPA because of the benefits the technology offers, there is still a long way to go. McKinsey has estimated that at present, only about 37% of most typical business processes have been digitised.
According to Mishra, most RPA adopters are driven by the prospect of the technology driving down operating costs as the costs of automating processes and maintaining them are usually lower than those of human resources. Aligning middle- and back-office operations to improve customer experiences and business process efficiency is also a strong driver.
In addition, RPA can boost employee experience by freeing human resources from having to do repetitive, boring tasks to rather tackle more value-adding tasks, upskilling them and giving them an opportunity to advance within the organisation.
It can also improve process compliance by precisely executing prescribed rules and thus reducing the risk of mistakes; allow for greater scalability of solutions which can be expanded or reduced to meet variable demand far more quickly and easily than trying to adjust human resources to such changes; enable comprehensive insight into and analysis of business processes and better predict process outcomes; allow for changes in business processes to go live far more rapidly than would be possible with human resources.
From acting to thinking RPA
However, Mishra warned that not all business processes were suitable for RPA. Effective RPA implementations were usually limited to processes that are repetitive and time-consuming; have clear, established and documented rules; require the use of multiple systems, applications and data sources that are often technically challenging to integrate; and utilise structured data that does not involve subjective (human) interpretation.
"In other words, the most suitable processes for RPA are those with low complexity or volatility, are repetitive and require a large number of people to execute; while highly complex, non-repetitive processes that generally require only a few individuals to execute are not suitable for today's basic RPA which could be termed RPA 1.0," he explains.
"Because bots - the software that is programmed to follow prescribed rules in order to carry out business processes - can't deviate from those rules or ask questions, RPA 1.0 can only carry out business processes and tasks that can be presented unambiguously. RPA 1.0 does not learn from mistakes, nor can it adapt to or recognise changes, or use new interfaces."
Mishra points out that if bots could learn, they would require less up-front effort in RPA deployment. Thanks to advances in applied artificial intelligence (AI) and machine learning algorithms that have the ability to detect patterns and make predictions and recommendations, bots do not have to receive precise programming instructions to adapt to changes in business processes.
"This means bots will be able to be used to automate a far wider range of business processes than is currently possible, which could drive demand for the technology. However, this is not going to happen overnight. The process automation journey that moves a process from one that involves acting (RPA 1.0) to thinking (RPA 2.0), will require careful planning.
"There are numerous steps that will have to be followed in order to move from early automation practices towards the implementation of intelligent processes," Mishra concludes.
Share