Digital Operations: Robotic Process Automation
“Looking to the future, the next big step will be for the very concept of the “device” to fade away. Over time, the computer itself—whatever its form factor—will be an intelligent assistant helping you. We will move from mobile first to an AI first world.” – Sundar Pichai, CEO Google
Software agents or Robotic process automation (RPA) is becoming a mainstream topic at leading corporations as C-Suite execs look at new automated strategies to do more with less.
Digitally Powered Business Process Automation is taking center stage again. Outsourcing, offshoring strategies are reaching the point of diminishing returns so a new frontier enabled by a virtualized workforce of software robots is emerging.
The focus is not just on simple tasks like answering the phone in a call-center but on managing complex data-heavy business outcomes, such as predictive maintenance on instrumented aircraft engines or preventing fraud in a large bank.
In the past year, I have seen a massive uptick of interest in digitizing work – automate key processes and increase efficiency – via robotic process automation. Large corporations like Citibank are implementing this trend with vendors as they race to cut operating costs further.
The market opportunity of AI across industries and business processes has been expanding rapidly, with analyst firm IDC predicting that the worldwide content analytics, discovery and cognitive systems software market will grow from US$4.5 billion in 2014 to US$9.2 billion in 2019, with others citing these systems as catalyst to have a US$5 trillion – US$7 trillion potential economic impact by 2025.
Digital robots ∼ Apple Siri, Amazon Echo/Alexa, Microsoft Cortana, IBM Watson, Google Home/DeepMind, Facebook ChatBots, drones and autonomous driverless cars ∼ are now mainstream. What most people are not aware of is the rapidly advancing area of enterprise robots to create a “virtual FTE workforce” and transform business processes by enabling automation of manual, rules based, back office administrative processes.
This emerging re-engineering of key back-office and front-office operations is called Robotic Process Automation (RPA). Machine Learning (ML), guided ML, NLP and graph processing are becoming foundations for the next wave of advanced bot use cases. Speech recognition, image processing, translation have gone from demo technology to everyday use in part because of machine learning.
RPA – What?
“Robotic automation refers to a style of automation where a machine, or computer, mimics a human’s action in completing rules based tasks.” Blue Prism
RPA is essentially the novel application of analytics, machine learning, AI and rules based software to capture and interpret existing data input streams for processing a transaction, manipulating data, triggering responses and communicating with other enterprise applications (ERP, HRMS, SCM, SFA, CRM etc.).
RPA is not a question of “if” anymore but a question of “when.” This is truly the next frontier of business process automation and enterprise cognitive computing. Immediate impact is being seen around self-service processes, customer facing processes, call center interactions, finance and accounting processes.
Industrial robots are remaking factory and warehouse automation by creating higher production rates and improved quality. RPA, simple robots and complex learning robots, are revolutionizing the way we think about and administer business processes (e.g. customer service), workflow processes (e.g., order to cash), IT support processes (e.g., auditing and monitoring), and back-office work (e.g., data entry).
I strongly believe that as ML becomes mainstream, RPA is going to impact process outsourcers (e.g., call center agents) and labor intensive white collar jobs (e.g., compliance monitoring) in a big way over the next decade. Any company that uses labor on a large scale for general knowledge process work, where workers are performing high-volume, highly transactional process functions, will save money and time with robotic process automation software.
Business Impact of RPA – Where?
Apple and Samsung supplier Foxconn has replaced 60,000 factory workers with robots. One factory has reduced employee strength from 110,000 to 50,000 thanks to the introduction of robots — BBC
RPA is already being applied to a wide range of industries to improve speed, quality and consistency of service delivery of digital work.
Similar to manufacturing robots, software robots mimic the actions of human users. They can:
- Learn from natural language interactions in order to solve customer problems and respond easily to a wide range of queries
- Automate data and rules intensive activities like HR, procurement, invoicing, billing. Now it is possible to create complex cross-enterprise apps (xapps or composite apps) like order-to-cash automation.
- Orchestrate other application software apps through the existing APIs or user interface
Workflow and Process automation
Clerical and white collar labor is replaced by software.
Best projects for robot automation are bulk repetitive rules based procedures. Process automation can expedite back-office tasks in finance, procurement, supply chain management, accounting, customer service and human resources, including data entry, purchase order issuing, creation of online access credentials, or business processes that require access to multiple existing systems.
Technologies like BPM software – a technology that mimics the steps of a rules-based, non-subjective process without compromising the existing IT architecture – are able to consistently carry out prescribed functions and easily scale up or down to meet demand.
Automated agents and assistant
Large call centers are going to get restructured. The people answering simple queries will be replaced by 2020 with software bots.
As in voice recognition software, IVR or automated online assistants, developments in how machines process natural language, retrieve information and search mean that RPA can provide answers to self-service customers without human intervention. I can see demand reducing systematically for armies of low-cost labor offshore that do simple tasks like status checking…. query multiple systems and respond; data entry…input into multiple systems and error check.
Voice driven self service bots are going to transform call centers. Siri and FB Chatbots are the precursors of what’s coming. Translation is another example. Recently combined translation with computer vision and doing it all on the phone, where you can take a picture of a sign that say “Exit” and have it translated into another language.
The bot engines rely on NLP and machine learning. It means that you can feed the bot sample conversations so that it can handle many different variations of the same questions. The potential is quite big as developers could improve their bots over time. So for instance, you could open up a conversation with a Movie bot and casually ask questions about movie showtimes, ratings and more. It will be like talking with a human agent.
Monitoring support and management
‘Human only’ processes will shift as machine learning and data-driven decision making evolve further.
Activity, fraud and risk monitoring is going thru some changes. Automated processes in the remote management of IT logs, audit trails, security, and other risk related areas can consistently monitored, flagged and exception handled faster. In IT function specifically, RPA can improve service desk operations and the monitoring of network devices.
KPMG, for instance, is leveraging IBM Watson in improving Audit, Tax processes. One current initiative is focused on employing supervised cognitive capabilities to analyze much larger volumes of structured and unstructured data related to a company’s financial information, as auditors “teach” the technology how to fine-tune assessments over time. This enables audit teams to have faster access to increasingly precise measurements that help them analyze anomalies and assess whether additional steps are necessary.
This example highlights how cognitive technology is further advancing improvements to sampling processes, in which auditors review subsets of data to analyze thousands or millions of actions to draw conclusions. Cognitive technology helps allow for the possibility of a larger percentage of the data to be analyzed, providing KPMG professionals the potential to obtain enhanced insights into a client’s financial and business operations. At the same time, cognitive-enabled processes allow auditors to focus on higher value activities, including offering additional insights around risks and other related findings.
Many of professional services rely heavily on judgment-driven processes. Adding RPA and cognitive technology’s massive data analysis and innovative learning capabilities to these activities has the potential to advance traditional views on how talent, time, capital and other resources are deployed by professional services organizations.
How is a Software Robot Trained?
- A robot is trained through a flow chart of the procedure. This flow chart is managed and audited to document how well the robot follows the procedure.
- Management information (e.g., log files) is gathered automatically as the robot operates. All processes generate statistical profiles as a by-product of doing the action. This allows tuning and development of a process in light of real data.
- Modern robots systems come with failover and recovery inbuilt as core capabilities. It means that if changes take place, or downstream failures occur a “smart” response can be trained into the overall system.
- Software robot platforms have full audit and security authorization meaning that all changes and all access is recorded and regulated. Back-up process steps are managed, roll-back and recovery, as well process change-highlighting, are all automatically captured by the robot platform.
The robots are coming to digitize work! Enhanced scalability, greater accuracy, digital integration with APIs, improved compliance and reduced cycle times to deploy – as these improve… RPA adoption will take off.
Analytics enabled Robotic process automation (RPA) will drive improvements in accuracy and cycle time and increased productivity in transaction processing (e.g., healthcare claims processing) while it elevates the nature of work by removing people from dull, repetitive tasks.
RPA is in early days. So, sometimes the hype can get ahead of the reality. But this is an area where I am going to be digging deeper in subsequent blog posts.
The digital workplace is about a fundamentally different way of working. Influence, networks, and dynamic decisions become much more important than power, hierarchies, static decisions, processes, and rules that made sense in a slow-moving, traditional environment.
- Predictive Analytics 101 (quick overview)..To make confident, data-driven decisions, robots need predictive insight.
- Business Analytics 101 (quick overview)
- Executing Analytics 101 (quick overview)
- Machine Learning 101 (quick overview from Google)
- RPA vendors include: Blue Prism; AutomationAnywhere, Deskover, Jacada, UIPath, IBM
- Institute for Robotic Process Automation – http://www.irpanetwork.com/
- O2 implements RPA
- Amazon Robotics ~ Robots in the warehouse…. interesting vision statement… “At Amazon Robotics, we are continually reimagining what now looks like. We see the big picture, imagine a better one, and make the connections that turn complex problems into elegantly simple solutions. Our drive toward a smarter, faster, more consistent customer experience fuels Amazon – and the industry – forward, now. With a fearless resolve to achieve the improbable with real solutions, we meet tomorrow’s challenges today. We Reimagine Now.”
- RPA is an Analytics use case. See more on Use Cases in Analytics here.
- IBM Watson
- Professor Leslie Willcocks and Professor Mary Lacity of the London School of Economics are doing interesting academic work in this area.
RPA is powered by Machine Learning. What is Machine Learning?
Machine learning uses massive amounts of observational data and focuses on automation. ML focuses on the algorithms, such as a random forest or gradient boosting, to automatically handle things like missing values, finding interactions and so forth.
Shown below is a medical diagnostic and treatment algorithm used by IBM Watson. Central to machine learning is the idea that with each iteration, the algorithm will learn from the data. To measure whether or not you’re improving performance, you look at an objective function, such as minimizing a loss function. The algorithm iterates through the data until a convergence criterion is met. Holdout data is used to see if you are overfitting.
The key types of machine learning include: • Supervised learning. • Unsupervised learning. • Semi-supervised learning. • Reinforcement learning.