How AI Will Take Robotic Process Automation to the Next Level
RPA, even when empowered by AI, is not a Terminator-like robot. It’s not cloning humans. It’s just an advanced set of software that follows rules to perform business actions.
Join the DZone community and get the full member experience.Join For Free
For the last couple of years, many companies of virtually all sectors of the business market — including finance, government, retail, telecommunications, utilities, energy, and transportation — have been enthusiastically considering the evaluation and management of robotic process automation projects. Some of them even consider this technology as the cornerstone of their digital transformation journey.
Robotic process automation (RPA) provides advanced software robots taking the place humans whenever complex processes or routine tasks can be automated. That being said, how can artificial intelligence and related technologies empower it? As we enter the the digital transformation era, our industries are reporting that their task forces are operating about 80% of their IT processes manually, lowering their performance and motivation. At the same time, they estimate that at least 50% of these tasks could be automated.
While some companies might just think about automating their IT processes, RPA also aims to help them reinventing the way they do business, heightening their customer satisfaction and strengthening employees' work value.
RPA uses software and methodologies that are capable of taking advantage of the latest technologies including artificial intelligence, machine learning, voice recognition, and natural language processing to take automation to the next level. That makes it a must for companies of all industries that want to convey their business all along the digital transformation journey.
According to Gartner, the growth of RPA could exceed 40% from one year to the next by 2020. TMR (Transparency Market Research) states that at a global level, the RPA market will reach up to $16 billion by 2024.
According to Accenture, “a successful RPA implementation can yield a 40 to 80 percent reduction in processing costs and up to an 80 percent reduction in processing time.”
Among all the questions that arise concerning RPA, one of the most repeatedly asked is to know what can be automated. McKinsey points out that it should be considered not only from a technical point of view. Beyond the technical feasibility, cost of automation, relative scarcity or abundance of resources, required skills, cost of human workers who are alternatively doing the job, and expected benefits also need to be taken into consideration.
Robotic process automation software and services are able to run applications the way a human operator would. Based on rules, the workflow operates automatically complex tasks. RPA brings a whole variety of benefits such as:
Continuing service: When it comes to running real 24/7 service, software robots emerge as obvious in that they do not have to take breaks.
Scalability: The processes specified for one software robot can be expanded to any number of other robots and conversely, robots can be decommissioned of a process to work on another one.
Truthfulness: Once assigned tasks, robots are designed to faithfully comply with the instructions without failing.
Audit trail: The robots' modus operandi involves the generation of output data. This data aims to ensure compliance and leads to improved processes.
Cost: A robot costs at least 20% of a human does.
Time: While it takes years to implement traditional projects with humans, it only takes weeks with robots.
To complete this non-exhaustive list, we must not forget that people tend to ignore the fact that RPA increases humans' work value. By definition, the tasks earmarked to robots are those not requiring creativity, emotional intelligence, or complex decision strategies. In this way, the human workforce is highlighted.
RPA Success Factors
Although the time required to build an RPA project can be seen as a matter of weeks, Accenture describes six elements that are mandatory to successfully achieve it:
An organizational commitment.
A comprehensive approach to standards, identification, and development.
Expert teams with deep knowledge of best practices.
Buy-in from individuals and groups affected by automation.
Recognized targets and benchmarks.
In most industries, the average employee spends up to 80% of their day on repetitive tasks that don't require creativity or deep thinking. These mundane tasks are meant to be automated. To illustrate this, the following are a few of practical business cases:
Fraud detection: Robots can assist human bank employees performing background checks and time-consuming fraud investigations while the employee can focus on customer satisfaction.
Form-checking: Robots can handle tedious customer order-checking to prepare the delivery process. It decreases the required time and at the same time reduces the margin of error.
Claim processing: Robots can review customer claims and identify who will end up with a refund without requesting any aid from a human.
Fax categorizing: Robots can convert fax images to machine-readable text and then extract data and categorize faxes.
Artificial intelligence empowers RPA. Many various business cases for RPA are being realized within innovative companies from many different industries. Use cases include accounting, billing management, customer onboarding, data validation, customer service inquiry routing, inventory list updating, loan qualification, risk assessment, and official document validation. RPA promises to be able to run 24/7 with no stops, no breaks, no sleeping time, no vacations, and no sick leave, without forgetting, omitting, misunderstanding, or underestimating errors and without encountering any problems.
However, this is a theoretical assumption. In reality, it would be difficult to put it into practice RPA and meet this level of service. There are exceptions, unexpected events, and particular cases.
This is where cognitive technologies and machine learning become all the more important. RPA platforms empowered with AI technologies tend to automate the emotional- and judgment-based process. To achieve this, they need to integrate cognitive capabilities including natural language processing, machine learning, and speech recognition. Then, these automated processes can integrate a human response into their workflow. They can learn from human actions and be sure that they will be able to take the required action autonomously. The purpose is to learn, ingest, and modelize data so that the RPA platform will able to keep the intervention of a human to a minimum.
This combination of RPA and ML is called IPA (intelligent process automation) or CRPA (cognitive robotic process automation). In cases like this, artificial intelligence value added consists of being able to acquire and aggregate complexe data from heterogeneous sources (text, voice, natural language...) and to exploit this data just as easily as traditional data. In addition, RPA workflows can be empower by advanced algorithms in order to analyze weak signals, detect patterns, recognize models, and associate events in order to make predictions. Finally, very advanced tasks can be achieved intelligently by combining RPA with refined software mechanisms and algorithms including ML, voice recognition, and more.
That is how RPA will go far beyond the basic task automation, taking advantage of the potential and the capabilities of AI and widening the sphere of possibilities — which means taking its productivity to the next level.
A short list of what RPA does in comparison to AI:
RPA: Mimics human activity through UI in a non intrusive way and can handle structured and semi-structured data. The spectrum of possible actions is determined by predefined rules so that the behavior of the robot is deterministic.
RPA + AI: Mimics human activity through machine vision, speech recognition, and pattern detection capabilities and can handle structured, semi-structured, and unstructured data. Machine learning lets robots learn how to process and also improve processes that leads to probabilistic behavior.
These few examples give us a glimpse of how AI can broaden the scope of RPA and extend its activities. AI technologies, data analysis, prediction, perception, prescription, and reasoning solutions constitute genuine improvements in areas including data mining, revenue, or expenses predictions, sentiment detections, contextual reasoning, purchase behavior, fraud prediction, next-best action, and more.
What if robots take our jobs? According to Erika Morphy, "It's not a matter of if, but when." She predicts that "workers of the future will learn to deeply cultivate and exploit creativity, collaborative activity, abstract and systems thinking, complex communication, and the ability to thrive in diverse environments."
As a matter of fact, RPA, even when empowered by AI, is not a Terminator-like robot. It’s not cloning humans. It’s just an advanced set of software that follows rules to perform business actions. Humans are requested to build the rules, handle exceptions, bring emotional intelligence, and deal with complex and unpredictable issues.
For my final point, let’s consider that while companies are reinventing themselves in an effort to embrace digital transformation, RPA — and even more likely, CRPA — will let them redeploy employees in a more interesting and more valuable mission in order to "never assign a human to do machine job."
Opinions expressed by DZone contributors are their own.