From RPA to Hyperautomation to HyperApps: Level Up Automation Deployments in 2021

In the early days of automation, robotic process automation or RPA brought the promise of radical transformation and improvement. Organizations could automate mundane, repetitive tasks, potentially giving back thousands of work hours to the business and reducing FTE efforts. 

However, these automations were not integrated or even necessarily connected to automate end-to-end tasks or processes – leading to fragmentation. A decentralized approach and a focus on “a bot per user” has increased technical debt for enterprises, putting true digital transformation out of reach. Over time, enterprises cobbled together disparate automation technologies to protect their original investments in RPA and were forced to assume the risks involved in integrating them. Gartner coined the term “hyperautomation” to define this integration of technologies, encompassing RPA, machine learning, artificial intelligence, and these technologies’ growing sophistication. Despite RPA’s massive market share, it was fast becoming apparent that RPA alone could not keep pace with today’s digital transformation requirements, necessitating hyperautomation – but this had its own share of issues. 

Organizations choosing to automate via RPA as well as those venturing into hyperautomation report a significant trade-off in terms of growing complexity, mounting technical debt, and a snowballing total cost of ownership (TCO) – which does not make sense in the long-term. 

In 2021, as we enter a new era in digital transformation, it is time to revisit our automation approaches and level-up. 

Traditional RPA is more a white elephant for enterprise automation

During COVID-19, we saw several years’ worth of digital transformation (3-7 years, according to McKinsey) take place in a matter of months. As we enter the next phase marked by consolidation, maturity, and long-term sustainability, organizations should rethink one of the core tenets of digital transformation – automating business processes. 

Robotic process automation (RPA) is entirely task-based, where you define precise rules to guide workflows in business process automation. Let’s say you are setting up an RPA software for invoice automation. At the invoice registration step, you can configure RPA to read from a file/folder, but every new source has to be manually configured. As you receive invoice submissions from multiple sources like cloud-drives, email, etc., the RPA script has to be updated and managed accordingly. 

Over time, this leads to RPA becoming more of a white elephant than a genuine value generator, as you will be spending outsized efforts on updating, cleaning, and maintaining your automation scripts as your enterprise grows into diverse functions/areas. 

A survey found that over 4 in 10 enterprises are having to spend more time and resources to maintain RPA than originally expected. 

Another issue is deployment timelines. Enterprise leaders start with the best of intentions but adapting RPA to a typical enterprise’s scale, and process complexity takes time – often up to three years. More than two-thirds of deployments take anywhere between 1 and 3 years, delaying your time-to-value. And once RPA is in place, just 4% are able to scale, mainly due to the complexity of projects (57%).

This leaves you with mounting technical debt and sunk costs, further increasing your TCO. 

Improving on this approach, Gartner introduced hyperautomation as the next phase of maturity, which would take advantage of AI/ML to cut down some of the inefficiencies of traditional RPA. 

The rise of hyperautomation, the no. 1 strategic technology trend from 2020

Gartner calls hyperautomation “the application of advanced technologies, including artificial intelligence (AI) and machine learning (ML), to increasingly automate processes and augment humans,” with the ultimate goal of enabling AI-driven decision making. 

It was the no.1 technology trend from 2020, poised to simplify several of the complex scenarios that would stymie traditional RPA.

Here’s a simple AP automation example: If you are using automation to extract invoices, RPA would require you to pre-train the engine and create separate templates for each supplier. Hyperautomation improves this through ML so that the data extraction isn’t template dependent. Similarly, when it comes to validating invoices, hyperautomation can crosscheck via intelligent OCR, in contrast to RPA, which only reads specific ERP fields or structured information.

But even hyperautomation does not match up to the promise of true digital transformation. Breaking down the above scenario, you will find frequent human involvement (often at preventable intervention points). For example, hyperautomation-based invoice extraction still lacks continuous learning capabilities. ML models are mostly a “black box” that cannot be adapted to business user behavior. For invoice validation, you still have to write complex scripts – only now, it is compatible with both structured and unstructured information. 

For this reason, hyperautomation remains confined to the “promising trend” segment, with limited real-world usability. Research names only Amazon and Google as key players, owing to their rich AI/ML capabilities. 

Does this mean enterprises who need immediate and effective outcomes from automation are left in the lurch unless they are willing to spend on a 5-year-long ROI generation roadmap? 

This is where HyperApps come in. 

Progressing to HyperApps – a pragmatic model with human-in-the-loop

HyperApps combine the functional principles of RPA, the intelligence/cognitive capabilities of hyperautomation, and the self-service convenience of SaaS apps to enable automations that show value in months and last for decades. 

Continuing with the scenario of invoice automation, here is how a HyperApp would do it: 

  • Invoice registration – Business users can integrate their preferred invoice source through a simple, point-and-click UI.
  • Invoice extraction – Any exception not covered by existing formats is routed to the business user. The user’s behavior is taken as a learning point, and the ML will adapt its future actions accordingly. 
  • Invoice validation – All validation rules are pre-configured; business users can toggle a rule on/off for a specific supplier when validating. 
  • True cloud native – Pushing new configurations to existing automation implementations is easy, allowing for constant upgrades of the HyperApp’s business process automation capability. 

HyperApps introduce a few important changes to the RPA-to-hyperautomation maturity curve. 

First, HyperApps rely on self-service, empowering business users to set up automated workflows and configurable business rules. What the HyperApp eliminates is the dependency on technical resources to make business configuration enhancements and changes. HyperApp designers can also add new functionality to the app and business users can turn them on based on their needs. 

Second, HyperApps are modular, with their components reusable as you grow, by applying the same components to multiple scenarios. This brings down the total cost of ownership and generates cost savings, while also shrinking time to value because of its turnkey nature. 

Finally, the human-in-the-loop user interface can replace the bulk training ML approach in cases where it is not possible to create a pre-trained ML model. This business user-led approach allows enterprises to build or enhance ML capabilities with their own business data.

As you can see, HyperApps address the key impediments to traditional RPA and hyperautomation. They ensure fast deployment and low maintenance, adapting to complex processes during business growth. They also keep a human in the loop to power continuous learning, reducing your efforts for manually configuring AI/ML models. Importantly, HyperApps are already in action at several enterprises, enabling long-term digital transformation without having to wait for technology or infrastructure maturity. 

Learn from the frontlines and level up today

Demonstrating a remarkable improvement over RPA alone, one of the world’s largest automobile manufacturers was able to achieve 85% straight-through processing (STP) for invoicing processes in just a 12-week period. The company first tried RPA in their AP automation journey to replace manual execution. But it was too rigid and rules-based, unable to handle frequent changes in invoice templates as the manufacturer added new vendors, new invoice formats, new types of suppliers, etc., as part of its growth journey. 

RPA solutions couldn’t keep pace with the company’s 5000-strong supplier network, processing 150,000 invoices per month. 

An Invoice Processing HyperApp successfully addressed this by learning from 12 months’ worth of historical invoices and continually updating itself whenever it encountered an exception. Using a HyperApp, the manufacturer can process one invoice in three minutes vs. the pre-automation 24-hour turnaround. And unlike most implementations, it saw measurable ROI in six months. 

At JIFFY.ai, we help organizations around the world with their digital transformation roadmaps by making it possible to level up their automation projects. This pragmatic progress from RPA to hyperautomation and finally, to HyperApps has proven to bring about battle-tested outcomes. 

To learn more or discuss any automation bottlenecks you might be facing, please email us at AcceleratingAutomation@jiffy.ai.