More than RPA — the synergy between RPA and AI

According to a report by the Economist (2019), 61% of US companies executives already use Intelligent Automation and Hyper Automation in their business processes extensively. We previously explored how Conversational A.I. Office Assistants could help to improve engagement with staff in the workplace and improve on overall efficiency and productivity. In this article, we will dive deeper into both Robotics Process Automation (RPA) and AI as the technologies driving the automation landscape.

RPA, sometimes also referred to as software robotic, is a form of business process automation technology that uses rule-based logic and structure to automate mundane and repetitive business requirements such as invoice processing and data entry tasks. Gartner describes RPA as one of the fastest growing enterprise solutions in the software industry, forecast to reach $1.89 billion revenue in 2021 and projected to have double digit growth through to 2024. RPA’s success has attracted more than 50% of Fortune 500 companies to experiment with automation, with companies like Dell and IBM leading the trend by adopting automation as a core future strategy.

However, adopting RPA isn’t always a smooth ride, and many companies have experienced mixed success when first starting their automation journey, with Ernst & Young reporting that as high as 30–50% of initial RPA projects will experience difficulties. E&Y further finds that the primary root causes for this is a wrongful methodology or general misunderstanding of RPA as a technology.

So first, let’s understand some limitations of RPA. RPA, in its current form, is designed and suitable for automating repetitive and fixed processes like managing payroll data, but generally speaking they are not suitable for tasks related to knowledge and cognition such as writing an article, or where the process itself is subject to frequent change like handling customer service enquiries. In short, RPA may be able to speed up a process, but it is not able to fix the process itself, therefore design and implementation must be clearly defined to have a reasonable chance of project success. Unfortunately, many companies do not have a clearly defined working processes or they are frequently changing with new administration personnel — such as adding new approval procedures to existing processes — and consequently maintaining RPA systems can become complicated.

Furthermore, while up to 80% of business processes can be automated without great difficulty, the final 20% of tasks relating to knowledge and cognition can prove to be a costly challenge. The rule-based logic framework used within RPA simply makes the technology unsuitable for tasks requiring complex logic or where the data is not in structured formats. Attempting to automate these types of tasks, the final 20% of processes, is likely to result in complicated designs and the project cost would increase as much as five times on average. It is also debatable that setting up such a complex system would ultimately increase the post-deployment management and maintenance cost, quickly eating into the cost saving benefit of adopting such a system in the first place. According to a study in 2019 by UiPath about RPA employee experience, 93% of the respondents said that they already struggle to understand the different RPA deployment options without any added complications.

Although, according to another research on RPA implementation, over 40% of RPA projects failed to deliver the expectations on implementation time, implementation cost, cost saving and providing enough analytics (Dilmegani, 2021), the next emergence of new generation RPA technologies leveraging AI solutions is largely predicted to solve or greatly reduce some of the existing challenges For example, process mining powered by AI can help companies to identify tasks that are good candidates to automate, thereby improving implementation time and costs.

Here at Kami, we focus on how we can apply Natural Language Processing (NLP), or Conversational AI, to empower Intelligent Automation. Primarily, NLP is increasing the scope of what processes can be easily automated by providing a considerable boost to the data processing capabilities for RPA systems, particularly within unstructured text formats — such as a customer service chat sessions. Many companies are starting to explore chatbots as a customer service automation solution, a process that was once thought to be impractical to automate due to complex consumer expectations. In 2019, American Fidelity Insurance combined RPA with NLP to scan emails, classify them, and to route them to the correct departments. A task that used to take 45 hours to manually read through 9,000 emails was reduced to 3 seconds of scanning (Boulton, 2019). NLP, and AI in general, is starting to unlock what is the new “possible” within Intelligent Automation.

It is also worth noting that one of the key emphasis of Conversational AI is to create a better understanding, and to humanise, the relationship between humans and computers. Conversational AI makes it possible for software robots to understand human languages, which ultimately can be leveraged to create much more natural interfaces and user experiences for human use. By making these software robots easier to interact with and to communicate our intents and requests, it is possible to reduce the training time and cost for using the system, and in some cases, to eliminate them altogether.