This guide is a brief introduction on implementation of Artificial Intelligence (AI) in Robotic Process Automation (RPA). If you are unaware of technological advancements related to adoption of robots and AI in business processes, then this guide can give you a head start. You will learn what the future of AI and RPA holds for you, your employees, or your business. It does not matter what industry you are from, this guide can equally benefit you.
This guide is divided into four key sections. In the first section, you will be introduced to Robotic Process Automation. It will help you learn why your business needs RPA, important aspects worth considering when implementing it in your business, and some use cases of RPA.
In the second section, you will be introduced to Artificial Intelligence. It will help you learn the extent of development in this area and AI technologies that are implementable so far. You will also learn the difference between General AI and Narrow AI.
In the third section, you will be introduced to implementation of AI in RPA, also called Intelligent Process Automation (IPA). You will learn why AI should be implemented in RPA; four key capabilities that are worth considering during implementation of AI enabled RPA; process that you should adopt in your business when you decide on implementing AI and RPA; and some use cases of IPA.
In the fourth and fifth section, some external useful resources are listed for you. You can use it to further explore and learn about AI and RPA.
Effort has been made to keep this guide bereft of technical jargon that can be difficult for you to understand. However, adequate explainers are provided for technical terms and phrases that could not be done away with.
Robotic Process Automation helps you in automating activities in your business that are repetitive. RPA interacts with User Interface (UI) of your computer applications — same as humans would — and complete required business activities. Let us take an example for better clarity.
Suppose you receive hundreds of low amount invoices on email every day. So you have kept an employee who performs following activities whenever a new email arrives:
- If email has the subject “Invoice”, and an attachment, then he opens it.
- Reads the invoice
- Copies invoice number, quoted price, name, and bank account information of service provider
- Pastes copied information in an excel sheet and saves on the computer locally
- At the end of the day, he sends the collected information — final spreadsheet — to the finance department for payment processing
Here, RPA can mimic this employee’s activity every time an email arrives. It can follow through these steps without missing a single step, and send the completed excel sheet to the finance department on time. Also, if this employee consumes around 10 minutes to process one email, then RPA can do it in a minute or less. Moreover, unlike your employee, RPA can operate 24×7 without additional cost to the company.
So you can see why RPA is important to save time and cost of your business. Businesses around the world are implementing RPA and saving substantial amounts of cost and time. A 2018 study by Grand View Research stated that the RPA market is poised to reach “$3.11 billion by 2025” . RPA reduces challenges associated with humans and brings efficiency and accuracy in work processes. A 2016 study by Deloitte stated that in comparison to other business processes, RPA has gained higher position in speed and predictability . Also, it brings negligible disruption to the existing processes (see image 1).
You can implement RPA in your business by following through a three stage process. In the first stage, you identify the process that you find suitable to automate. In the second stage, map the activities conducted by your employee. In the third stage, configure your RPA software as per the activity map.
Now let us see some use cases of RPA in various businesses:
- Maintain consistency in customer data in banks. RPA can consider bank statements of the customers as a reference point and update their data wherever required.
- Errors in accounting are unpardonable as minor mistakes add up to huge losses. RPA can automate data entry activities to eliminate the possibility of human errors.
- RPA can activate customers’ bank cards after going through compliance, updating data in all departments, entering data manually, and other needed activities.
- Processing emails and preparing reports.
There are hundreds of other uses of RPA. If you can do something on a computer, and if it is repetitive, then RPA can do it. Though RPA has several benefits, it has its own limitations as well. If a business process is not deterministic (following predefined rules and regulations), RPA implementation will be tough. RPA uses GUIs, and so an A/B test from an application provider may stop RPA from working. RPA in its general form cannot do things that are novel. It cannot think on its own like humans.
According to the father of Artificial Intelligence, John McCarthy, “AI is the science and engineering of making intelligent machines, especially intelligent computer programs.” Theoretically, AI can do all activities that a human can do such as think, learn, touch, recognize voice, and others. This is called General AI. However, we are yet to build AI that can do all these exactly as humans. AI has excelled humans on some of the capabilities, but individually. For example, Deep Blue AI can beat top chess players and AlphaGo can beat the world’s best Go players. These are individual systems that can perform better than humans on a single activity, and so they are called Narrow AI.
There are various subsets of AI that are implementable so far in businesses. They are:
- NLP (Natural Language Processing): It works by converting text to data, and facilitates the interaction of humans with computers. NLP helps humans to interact with computers in natural language. Some popular examples are Alexa (AI virtual assistant by Amazon), Siri (AI virtual assistant by Apple Inc.), and Google Assistant (AI virtual assistant by Google).
- Computer Vision: It extracts information from images and scanned documents. It uses deep learning models and collection of digital images and videos to understand the surrounding. One example of Computer Vision is OCR (Optical Character Recognition) that detects texts from scanned or handwritten pages, and converts to editable texts. Image Recognition is another example of Computer Vision that identifies objects in an image.
- Machine Learning: It consists of algorithms that use historical data to get trained to provide intended outcomes. It learns and improves gradually without being programmed.
- Predictive Analysis: It uses combinations of machine learning and statistical algorithms along with data for prediction.
- Recommendation engine: It uses historical search data of customers to recommend items.
AI is in gradual adoption in various industries whether it is finance (use of Cyborg by Bloomberg to understand complex financial reports), healthcare (virtual nurse assistants), manufacturing, transport (route recommendations in Google Maps), and customer service (intelligent chat bots). A 2018 study by International Data Corporation predicted that by 2022, the market size of AI may reach USD $77.6 billion .
Full potential of AIs are yet to be recognized as they are still in their development phases. However, the inclusion of AI has begun in several industries and can be witnessed in RPA as well (see section 3).
In this section we will look at a few areas related to implementation of artificial intelligence in robotic process automation.
RPA has limited capability, can only do what is predefined, and lacks depth and flexibility. Any novel changes throttle its further actions. However, if you club capability of AI with RPA, you can multiply its performance.
RPA works only through UI and API (Application Programming Interface). However, AI enabled RPA recognizes speech, identify documents using OCRs, and detect patterns as well. It can work seamlessly with unstructured data. Moreover, unlike RPA, AI enabled RPA continuously learns and improves itself. Therefore, the more it works with structured or unstructured data, the better it gets. Companies in the future are likely to face a huge amount of unstructured data, which increases the need for AI based RPA .
The combination of AI and RPA is called Intelligent Process Automation or Cognitive Automation. Mohanty and Vyas, in their book “How to Compete in the Age of Artificial Intelligence”, said that there are five technology capabilities that form IPA. They are: RPA, Smart workflow, Machine Learning, AI capabilities, and Cognitive agents .
Here, we can club AI capabilities, machine learning, and cognitive agents together and call them AI. Smart workflow can be termed as Business Process Management (BPM). So basically there are three technology capabilities that we can see here that constitute IPA. They are: BPM, RPA, and AI.
Further, Mohanty and Vyas also mentioned the concept of system integration but did not consider it as the capability. However, integration of the systems is crucial for the functioning of all the existing systems. Therefore, we can consider System Integration as a separate capability as well.
In the following subsections, we will understand these capabilities in detail:
BPM helps in efficiently coordinating people within the organization with the systems and data to achieve intended outcome. It provides clarity in how things are being managed in a complex business process environment. You can observe several actions being taken by all the employees in tandem, whether it is data entry, making critical decisions, controlling actions, generating and storing data, or anything else. BPM ensures that things take place without duplicacy and overlap. BPM takes care of structured as well as unstructured data.
As we understood in the first section, RPA helps in handling repetitive tasks that are done with predefined rules.
As we understood in the second section, AI helps in building systems that can reason like human beings and learn new things continuously.
3.2.4 SYSTEM INTEGRATIONS
An effective business process cannot function in isolation. One activity flows into another. Therefore, you need effective and efficient system integrations. During IPA implementation, you have to ensure that all the tools and activities are integrated together and helps achieve what it intends.
In 2017, Pascal Bornet, who was the head of AI, Automation and Digital Innovation at McKinsey, suggested a stepwise process for organizations to adopt AI and RPA . He suggested that the companies should consider implementing traditional RPA prior to any other advanced tools. The suggested steps are given in the diagram below:
According to Bornet, there are several factors because of which the above model has been suggested. As you will move up the stairs in the above model, you will find that:
- Implementation cost and time increases
- Number of processes to which technology can be applied reduces
- Higher technology still needs major improvements
- Majority of benefits can be realized in earlier adoptions
Additionally, Bornet asserts that if an organization begins implementation starting with traditional RPA, then it helps in building foundation upon which future implementation improvements can stand.
Further, RPA has continuously proved its capabilities. It is easy to access and fast to implement. On the financial ground, RPA has higher ROI (Return on Investment) and it helps you visualize returns you can get from your business in short or long term. RPA brings in quick cost savings, which further builds the confidence in the system.
Though Bornet has suggested this model, he also stated that organizations can follow different sequences. However, from the time and cost point of view, only organizations with deep pockets can afford to play with other sequences. For smaller and medium sized companies, it would be best to traverse from bottom towards top.
There are several use cases of IPA. Some of them are listed below:
- OCR in RPA for account opening: Banks using OCR for reading scanned documents of customers such as address proof, age proof, signed checks, and others. Scanned and converted data can be further mapped as required.
- Identifying buttons: AI enabled RPA can identify buttons on screen that should be clicked. Thus, even UI changes will not affect the workflow.
- Recognizing images: Customers’ images can be recognized by the system and mapped to their account despite the absence of customers’ name in the page.
- Claim processing: System can review the claims made by the customer, verify his/her past history, recorded interaction behaviours, and make decisions for refund.
- Fax categorizing: Converting received fax images to text, extract data, and place in appropriate category.
- Email sorting: Understand the context of the email and understand the intention. Also, read through unstructured emails from customers and identify and pick relevant data and fill in the form.
Given below are few resources that can help you further in building your understanding about AI and RPA:
- Grand View Research (2018). Robotic Process Automation (RPA) Market Worth $3.11 Billion by 2025. Retrieved from Grand View Research: Available here. [Accessed 5 Nov. 2020]
- Deloitte (2016). Next generation automation Transform your business processes with robotic and intelligent automation. [online] Available here. [Accessed 5 Nov. 2020].
- D’Aquila, M., & Shirer, M. (2018). Worldwide Spending on Cognitive and Artificial Intelligence Systems Forecast to Reach $77.6 Billion in 2022, According to New IDC Spending Guide. Framingham, MA
- CiGen | Robotic Process Automation | RPA. (2020). Intelligent Automation: When Robotic Process Automation (RPA) Meets Artificial Intelligence (AI) — CiGen | Robotic Process Automation | RPA. [online] Available here. [Accessed 5 Nov. 2020].
- Mohanty, S., & Vyas, S. (2018). How to Compete in the Age of Artificial Intelligence: Implementing a Collaborative Human-machine Strategy for Your Business. Apress, pp. 132–133.
- Bornet, P. (2017). Intelligent automation is about creating synergies between RPA, cognitive, chatbots and AI. [online] Linkedin.com. [online] Available here. [Accessed 5 Nov. 2020].
- Marr, B. (2020). What Is Unstructured Data And Why Is It So Important To Businesses? An Easy Explanation For Anyone [online] Forbes. [online] Available here. [Accessed 5 Nov. 2020].