Uday Birajdar, CEO and co-founder at AutomationEdge, a leading AI-powered IT automation and robotic process automation company.

getty

We may be aspiring to a paperless world, but we’re still a long way away. Automation technology in almost every industry — banking, financial services, manufacturing, business process outsourcing — still runs on the back of system-generated, digitized forms and invoices.

In these automation processes, workflows start with inputs in terms of different templates, style, formatting and sometimes language. Many technologies have to be used before the data ever sees its target destination, be it an ERP, CRM or legacy system. To understand this, let’s take a look at the key steps in the journey of data processing:

1. Collection of heterogeneous data from multifunctional devices-scanners, computers, handheld devices or hard copies of documents.

2. Separation of images. Banks receive thousands of images, which requires resolution and separation of images into distinct files. Image processing includes cropping, alignment, tilt and orientation correction.

3. Digitization using multiple engine options, such as APIs or optical character recognition (OCR).

4. Classification involves a bot recognizing the document to be a purchase order or a sales order, an invoice or a KYC, a contract or an app form. This step uses machine learning (ML) and natural language processing (NLP).

5. Extraction of data using a bouquet of ML algorithms, key values, coordinates, tabular data, QR codes as well as data validation and standardization using rules-based approaches, formats and constraints.

As you can see, there are many steps in the run-up to the processing of a document, steps that seem trivial but are varied and necessary, and that need help while being iterative. While OCR seemed to meet the earlier needs of the market, it is woefully inadequate to meet the challenges of digitization today because we increasingly expect our data to be processed and understood just as a human would understand it.

What is RPA with hyperautomation? 

Hyperautomation was coined by Gartner in 2019 as a way to “rapidly identify, vet and automate as many business and IT processes as possible.” In comparison to traditional robotic process automation (RPA), hyperautomation is focused on expanding RPA’s boundaries to encompass decision-making and analytics, among many other advanced technologies, enabling it to perform more complex business and IT processes and forge the digital transformation journey more toward unattended automation. 

Your RPA platform is nothing but a deserted island if it is not in touch with the critical systems of the enterprise. All the data processing — separations, digitizations, classifications and validations — are done simply to make data palatable to resident ERP, CRM, payroll or HRMS systems.

Hyperautomation platforms, on the other hand, can greatly improve accuracy rates, according to Deloitte. For example, they can help users manage the five steps listed above, help when working with PDFs that are searchable, working with compliance regulations that require documents to be redacted or providing role-based access to documents. 

Still, even with these developments, many challenges await enterprises in adoption of hyperautomation platforms, especially in the areas of modeling AI and in changing the status quo at an enterprise. With modeling AI, for example, businesses should expect to face the following challenges when adopting a hyperautomation solution: 

• The sensitivity of training data (e.g., PII, financials) could have a cybersecurity risk.

• Training test data might be slow and cumbersome.

• For some edge cases, last-mile manual effort might be a better alternative.

Likewise, businesses might face additional challenges trying to change the status quo of the organization:

• Lack of existing process documentation at a granular level could lead to bias toward the information for which documentation exists.

• The specific, extreme and frequent variations in customer demands with tight timelines might rush results.

In order to mitigate these challenges, buyers should ask hyperautomation platform vendors about specific features, such as:

Service orchestration. Does the platform empower both business and IT users to implement business services through a combination of technologies, enabling them to automate their individual environments?

Chatbots. Some platforms enable users to talk with the system via email, chatbots, interactive voice response (IVR), etc. Can users speak to the system in a conversational manner to trigger automations or get automated fulfillment?

Native AI/ML. Is the platform capable of delivering the required AI/ML capabilities without sending any data over the internet?

Data processing speed. Can the platform process structured data at speed irrespective of the source of data or type of source?

Reporting capabilities. Does the platform have a native ability to create custom reports from custom data sources with complete flexibility around the nature of data and reports to be generated?

In their quest for end-to-end automation nirvana, businesses struggle with an ever-growing portfolio of best-of-breed tools that do not work well together. Typically, developers spend time custom coding or stitching these tools to make the entire automation quilt work. Hyperautomation platforms, though, are intended to alleviate the integration issues that arise from using a mosaic of tools.


Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?