Software bots can successfully automate routine and repetitive tasks to increase business productivity, but on their own, are unable to provide depth or insight into what tasks are actually being performed. Using the latest in machine learning, robotic process automation is breathing new life into bot capabilities and opening up new doors for enhanced business productivity.
Across the enterprise, robotic process automation (RPA) is increasingly handling routine and time-consuming tasks and challenging how businesses operate. This form of digital transformation is already showing a significant return on investment for early adopters, and the switch from legacy business systems to the integration of RPA technology allows enterprises to become more competitive, efficient and flexible.
RPA and machine learning
In the past, RPA tools were successful at executing specifically defined tasks, but limited in the sense that they could not adjust to changing conditions or learn from experience. Machine learning, on the other hand, applies artificial intelligence (AI) capabilities to lend business context to tasks executed by RPA systems, enabling the latter to make better decisions and be more productive overall. For example, when extracting field values from unstructured data, RPA can extract values based on the rules set. Machine learning, on the other hand, “learns” the most common labels for fields, while working with a human trainer to confirm what is being learned, which can then be applied to future scenarios. This results in a 10x faster path to automation, because explicit programming isn’t needed for quick improvement gains. By contrast, RPA without this learning capability requires a human to explicitly program these improvements, defeating the purpose of automation all together.
Leading technology offerings now use a suite of machine learning algorithms to automate complex steps in business processes. By directly embedding RPA into workflows and other lower operating software bots, this technology provides the ultimate flexibility in terms of triggers, logic and pre- and post-processing actions. By creating tools that can think, process and flexibly learn from their tasks, a more seamless experience can be provided to users.
The ability to use RPA and machine learning in parallel establishes a symbiotic link of continuous improvement between execution and analysis. The transactional data generated by RPA tools also provides a steady stream of analytical fuel to drive AI capabilities forward and enable a deeper level of understanding. That deeper understanding, in turn, can be applied to expand RPA adoption even further.
Emerging analytics tools provide increasingly enhanced visibility and transparency into business events and data records. In this context, RPA, AI and analytics are less impactful when used in a sequential manner or as individual components. Rather, the process should be fluid, where RPA deployments generate data to refine AI capabilities in an ongoing basis. Those capabilities can then be applied to conduct ongoing and increasingly targeted, relevant and effective data analysis. The result is a wide range of new possibilities and enhanced business value in terms of cycle time reduction, scalability, innovation and ongoing productivity gains.
RPA at work
When businesses adopt RPA, it’s often because they’ve identified problems with their legacy systems that are losing them time and money. For example, when a business has an inefficient manual process for purchase order fulfilment -- where human agents have to manually query the system for new orders and physically validate each one, the process is not only time-consuming and tedious, but prone to human errors. With RPA software, the process could instead operate like the following:
1. The RPA system automatically pulls data from the customer system, checking for new purchase orders, reducing fulfilment times and increasing productivity.
2. Once a purchase order is downloaded it can be immediately pushed into the legacy system.
3. The agent acts as a “human-in-the-loop” and manually validates the order for accuracy based on the customer contract.
4. RPA software then uploads the purchase order into a database where discounts are automatically applied based on customer agreements.
5. Agents then spot-check the fulfilled order, ensuring quality control and human touch.
6. With added machine learning capabilities, RPA software can start to learn and adapt to this process for even greater gains in efficiency, that over time, will require less human interaction to ensure accuracy.
By adding RPA to this legacy system, the fulfilment process can be significantly improved and bottlenecks in productivity, especially during busy seasons, can be eliminated entirely resulting in higher customer satisfaction.
Let’s look more closely at the banking industry. RPA systems can effectively perform many tasks associated with loan origination and account management. However, RPA typically can’t determine if the person making the inquiry is who they say they are. By analysing unstructured data (e.g. say, reviewing a scanned passport image and matching it against a customer’s account record), machine learning is then able to create a connection between doing and thinking in an automated environment.
Other applications of RPA and machine learning working in tandem include insurance claims and customer service. For auto insurers, sending claims agents out to review fender benders is expensive and inefficient. Today, many of these providers are exploring the use of computer vision applications with AI capabilities that can assess how an accident happened for fast approval of minor claims. For customer service departments deploying chat agents, sentiment analysis technology can detect anger, dissatisfaction or sarcasm conveyed by customers via text, and then flag at-risk customers and escalate issues to a human for proactive outreach.
The RPA market today
While the RPA market today is still relatively small, the innovation from leading RPA vendors to incorporate AI and machine learning technology into these offerings is pushing the industry forward at a rapid pace. In fact, in a recent report by Market Research Engine, projects that the RPA industry will reach $13B by 2022 and is gaining in support as users turn away from more traditional business process automation (BPA) systems that often take hundreds of work hours and millions of dollars to deploy. RPA by comparison is an easier and faster system to implement, which can be rolled out more cost-effectively through small incremental, very specific projects and then expanded to large-scale implementations.
Early adopters from industries such as banking and finance, insurance and healthcare, are seeing the benefits of RPA implementation grow exponentially. From lowering operating costs and error rates, to improving service and compliance, to the ability to scale on-demand, the application possibilities continue to broaden.
As enterprises begin to further explore and implement RPA technology, we will see bots’ abilities grow beyond automating routine tasks. RPA combined with advances in AI and machine learning is just the stepping off point for enterprises as they move away from legacy processes and work towards the integration of this new technology that has the potential to disrupt how we think and do work across all industries.
Abhijit Kakhandiki, vice president of product, Automation Anywhere
Image source: Shutterstock/Sarah Holmlund