Using artificial intelligence to optimise robotic business process automation
The interest and use of robotic process automation (RPA) has rapidly increased to support repetitive, labour-intensive and transactional business processes. For many enterprises, RPA has emerged as a best-fit alternative to making huge IT investments in order to make business processes more efficient. Recent studies have found that the service delivery automation (SDA) software market size has doubled in the last two years. Additionally, 78% of shared services organisations are implementing or planning to deploy RPA.
The most common processes enabled by RPA today include back-office administrative processes and workflow within IT, finance, procurement and human resources functions, as well as industry-specific operations across banking, insurance and mortgage. Any business process that has digital information in a structured format, as well as a logic-based (if-then) rules structure, typically present a reasonable case to be considered for RPA.
Automation through RPA can offer a return on investment (ROI) often ranging between 30% and 50%. However, in situations where structured, digital information becomes difficult to capture on the front-end, or the application of business process rules requires judgement and context, how do we deploy RPA? Artificial intelligence (AI) is here! With the expansion of AI (or cognitive computing) technologies in the areas of data capture, pattern recognition, and decision support, compelling techniques exist to address current deployment challenges and expand the reach and range of traditional RPA products. The outcome will be greater enterprise value and the rise of cognitive automation.
This article provides a point of view on how cognitive computing can address the challenges of RPA and speaks to specific examples of complex process automation and intelligent automation in business process services.
What does RPA do today?
According to the Institute of Robotic Process Automation and Artificial Intelligence (IRPAAI), “Robotic process automation is the application of technology that allows computer software or a ‘robot’ to capture and interpret existing applications for processing a transaction, manipulating data, triggering responses and communicating with other digital systems.”
RPA enlists software robots to perform complex, nested routines that cut across multiple applications and interact with these systems without the need to build complex and rigid system-to-system interfaces. Although robots can be configured to interface directly with applications through pre-existing application programming interfaces (APIs), these platforms are equally able to work through systems’ graphical interfaces, including web browsers and Citrix terminal sessions. This means that robotic automation can be deployed in a relatively short period of time measured in days and weeks, instead of more heavy, traditional IT programming efforts that can take months and require a much larger upfront investment. This makes software robots an attractive way to drive operational cost reduction, particularly within repetitive, rules-based processes that failed to attract the attention of past system-to-system automation efforts from IT and business stakeholders. However, despite the attractive potential, there are noteworthy technical limitations within today’s RPA products that prevent achieving the full value of RPA projects. These challenges include:
- Conversion of non-digital process inputs.
- Ability to identify target data fields within unstructured document formats.
- Ability to accommodate for changing rules or business logic with relative ease.
- Generation of insights from transactional performance data within a robotised process.
- Ability to interpret contextual understanding of instruction sets upstream of robotic activity.
The ability to address these technical challenges will unlock significant additional value within RPA-enabled business operations. IBM leverages the powerful capabilities introduced through IBM Watson services for assistance. Let us explore how these technical capabilities are being used in RPA operations.
On-ramp to non-digital process inputs
Cognitive computing provides a new way of interacting with non-digital, upstream processes to collect and transcribe information in a manner that can then be leveraged downstream in the process by robots. For example, chatbots or digital assistants are deployed within helpdesk functions to interact with an end-user, replacing the human agent in a traditional call centre.
These chatbots can engage in natural language dialog with the caller and discern, through a series of interactions, the nature of the call and then capture information required to further handle the request. The chatbot will then trigger a robot, passing the required information to the robot to complete the request.
Chatbots can be used for helpdesks, where they replace a human agent to interface with suppliers who want to understand the status of their payments. Using chatbots, the call centre agent is replaced and a virtual agent is used to interact directly with the supplier. Through several scripted questions, the virtual agent can surmise that the supplier is asking for status on their payment and, thus, trigger an RPA robot to access the appropriate enterprise resource planning (ERP) back-end system, validate the caller’s details, retrieve necessary information, and provide status back in real-time to the supplier. The chatbot then closes the call and updates the ticket management system used to track call centre activities.
Ability to identify target data fields within unstructured document formats
Approximately 80% of data in organisations is unstructured, the majority being in documents, emails and email attachments. Robots are unable to initiate an automated process if the source data is in a format that prevents the robot from ingesting this data into the automated flow of activities. For instance, in today’s financial operations, it is not uncommon to find human engagement required to map, survey and extract key data fields from within these documents and then pass this information to a robot for further processing. Cognitive capabilities can be used to ingest these unstructured documents and extract the relevant fields, thus removing additional manual work from the end-to-end process.
Ability to accommodate for changing rules or business logic with relative ease
Business processes typically contain rules and policies that must be followed during transaction processing. One can hard-code these rules into the RPA products today; however, the maintenance of rule logic can become labour-intensive for the developer if the process rules change frequently.
Furthermore, in some cases where nested if-then logic routines exist, programming these rules become more problematic. In light of this situation, IBM has introduced Watson Policy Management Library (WPML), which allows business process owners to draft, create and amend rules and policy changes outside of the RPA product using normal conversational text. The robot then interacts with WPML to look up and apply rules related to automated process.
Fig. 1: IBM’s intelligent automation continuum.
Draws insights from transactional information
A wealth of process diagnostical information is captured by RPA tools during the normal course of transactional execution. Information such as transaction volumes, transaction cycle time, exception or defect rates, and reasons for exceptions can all be extracted from the robotic tools and used to drive increased throughput and yields. To do so, statistical techniques and methods must be applied. IBM has leveraged cognitive analytic features within Watson Analytics to perform these sorts of statistical analyses that discover relationships, test correlations and search for insight—thereby enabling more effective root-cause analysis and identifying sustainable improvement opportunities.
Ability to trigger robots based upon understanding of upstream situational awareness
Intelligent automation using AI technologies replaces human involvement in work dispatching. They can monitor and survey system and business processes and initiate robotic work dispatching to perform actions based on observations and conditional logic. IBM leverages a combination of tools to accomplish this goal. IBM utilises AI application performance monitoring to bring the surveillance function into play, and combines with another product called Golden Bridge, which completes the work dispatch function, resulting in the assignment of tasks to robots for execution.
The benefits of combining RPA and AI
AI can be combined with RPA to enable new and compelling use cases and unlock new levels of value in two primary new ways:
- Extending RPA to areas that were previously unfit for automation.
- Increasing the yield of robotics within a currently enabled process.
Deployment of these cognitive tools promises new levels of productivity and innovation – freeing human workers to focus on higher-value activities and bringing industrial-level quality, consistency and auditability to business processes. Furthermore, human workers can use these techniques to interface with robots in more human-like ways, providing the opportunity for humans and machines to interact and cooperate seamlessly, driving new levels of efficiency and productivity.
How can you get started?
Organisations are attracted to the potential of cognitive automation but recognise the need for expert guidance in developing an overall technical roadmap, implementing a robust governance model, selecting and integrating appropriate technologies, and achieving financial objectives. One approach is to launch an enterprise-level automation competency chartered to design and champion a cognitive automation strategy and provide the necessary support to operationalise and execute this strategy.
Download the full IBM white paper on AI and RPA here.
Contact Jennifer McDonough, IBM, jennifer.mcdonough@us.ibm.com
Related Articles
BROWSER-BASED E-ZINES
BACK ISSUES
SERVICES
DAILY CARTOON
-
