Digital labor adoption has become the priority initiative in most organizations. Robotic Process Automation (RPA) uses non-invasive BOTs in a big way to remove operational routine activities and are adopted rapidly across the industry. To make better decisions, business processes need to be user-aware and enriched with contextual insights. Business Process Management (BPM) and other integration platforms are evolving with cognitive capabilities and processes are being reimagined with AI-infusion. Cognitive Process Automation (CPA) focuses on delivering stakeholder value while traditional RPA tools, which are originally designed for replacing human routine and repeated activities with BOTs, are moving up the value chain to incorporate AI capabilities. There are 5 major pitfalls to be avoided while designing CPA and other AI-infused platforms.
Absence of digital value metrics in automation strategy
Automation with the intention of removing routine activities will take away the focus of maximizing stakeholder value. Process reimagination with customer value (and associated digital value metrics) paves the way to fine-tune the digital objective and help in monitoring the value delivered by the process. This pushes the automation from establishing digital labor to self-thinking platforms using CPA.
For instance, the back-office team in a bank receives lending fulfillment request along with instructions and checklist to create accounts manually. Reimagination and transformation of the end-to-end business process will lift and shift operational activities, automate manual tasks and simplify customer communications. This transformation delivers measurable value, namely faster turnaround time, transparency in compliance checks, customer experience at touchpoints and less fatigue in executing transactions. On the other hand, traditional RPA ends up in simple automation of reading email, checking and updating at the backend.
In another example, a large European bank has 50+ people physically validating the mail address before dispatching to their courier service as the total penalty for bouncing of letters for incorrect address has become more than the value gained through bulk mail discounts. The transformation through CPA focuses on reimagined digital communication and delivers value; namely consistency, accuracy, relevancy and latest information through preferred channel. On the other hand, traditional RPA will remove the human labor of checking addresses and help in saving cost. But the digital values delivered, will influence greatly the top-line and the bottom-line of the bank.
Assuming a shorter automation timeframe
Automation in processing free form documents, namely sale deed, company annual reports, contracts, etc. uses “LEARN and ADAPT” method, wherein it must resemble the judgements applied by humans. Depending on the complexity, it takes time to bring more accuracy to the model as the system LEARNS with more and more input data. Automation is a continuous process and cannot be addressed by a single, large transformation program. An automated sub-process can lead to the next level of automation opportunity and can follow a path to higher orders of cognitive computation. An automation excellence center, which continuously evaluates the end-to-end process, will be effective to find automation opportunities and cascade change requests during the process life cycle management.
Ignoring operational empowerment
The automation strategy should help in the eradication of non-productive activities from the operations team and allow them to think on adding value to the top-line and bottom-line of the company. An operation team member should get up-skilled from an analyst role to a business executive role with specialization in automation.
He gets trained to extract financial ratios and takes care to provide automation input to a data scientist for continuous automation.
The standard claim processing checks can be executed through multidisciplinary, cognitive BOTs to achieve lighting speed in execution. Business executives can dedicate more time to think of further enhancing financial ratios, namely underwriting profit / loss. By doing this, the insurer can bring new innovative offerings as operations get better control of the underlying finance. Also, to enable continuous automation, the operation team should be empowered with real-time insights and data visualization through automation.
No governance as BOTs proliferate
A BOT is one among the many components of a wired business process. It has dependency on IT applications. It can introduce unseen points of failure at production if its design and behavior are not managed properly or not aligned with IT strategy.
Siloed BOT creation, deployment and management will introduce more complexity when BOTs proliferate. It can introduce issues with data integrity, end-to-end SLA violations and inefficiency in operations. It is also very important that the business team should refrain from creating BOTs without IT involvement for internal applications like HR and Finance.
An integrated approach to BOT creation, management and governance of its life-cycle is a must. BOTs should be treated as an enterprise asset by maintaining a registry and with a well-defined governance process. The governance should check compliance in onboarding BOTs and propagate re-usability. A center of excellence will help in centralizing best practices and reusable components.
Absence of intelligent oversight and accountability
Unmanned decisions can sometimes result in legal battles with parties involved, if any terms and conditions of contracts are violated. Similarly, to change a system record by unmanned BOTs needs the right credentials, treatment, audit trails and accountability to a person in the organization. A RACI matrix must be defined and accepted at the organization level and must be adhered to as part of a bigger, intelligent governance model.
Cognitive models contain hypothesis, features, algorithms and learned parameters. The models should be hypothesis tested with parallel run and compared with actual output obtained. There should be flexibility to rewind the intelligence for decision-making failures. The governance model should help in better intelligent oversight and control.