Built to resemble digital assistants for employees, robotic process automation is known to be useful for streamlining business operations without increasing costs, while reducing human error. However, RPA software alone has its pitfalls when it comes to content processing, due to incompatible intelligence.
However, there are ways to overcome these drawbacks, as revealed by five experts in the field.
One way to overcome content processing drawbacks is by combining other intelligent technologies and integrating it into the system.
“RPA technology is mainly used for automating rule-based processes and mimicking human actions, such as processing an invoice and entering data into SAP or Oracle systems from a Microsoft Excel spreadsheet,” explained Gopal Ramasubramanian, senior director, intelligent automation & technology at Cognizant.
“However, when it comes to processing content in documents, there is a need for additional smart intake technologies that combine optical character recognition (OCR), natural language processing (NLP) and machine learning (ML) to be able to extract metadata from the documents and automate the processing.
Why metadata is crucial in implementing a solid data strategy
“The content can be of different types, such as structured/printed, structured/handwritten, unstructured/printed and unstructured/handwritten. It is quite easy to extract structured content using standard OCR technologies. However, extracting unstructured content poses a challenge and, increasingly, we are seeing adoption of NLP and machine learning technologies to address it.”
Arpit Oberoi, RPA specialist at delaware, added: “The biggest challenge facing RPA technology today is the fact that it often still struggles to process unstructured content and data. To overcome this ongoing issue, organisations can try to harmonise their data into more structured datasets and also, where possible, combine AI and RPA to optimise or automate content processing.”
Third party involvement
“Historically the RPA technology has been able to integrate with third party applications to assist content processing,” said Rayner. “For example, many of the OCR vendors (Abbyy, IBM, etc.) have direct integration, allowing semi-structured or structured documents to be classified and recognised.
“At UiPath, we have heavily invested in document understanding to provide an ‘out of the box’ solution for customers, with the flexibility to apply different techniques such as pattern matching, templating and machine learning to deal with unstructured and semi-structured document types.
“As we think broader about content processing this plays well into hyperautomation, we now have long-running workflows with human in the loop, allowing both robots and humans to work seamlessly on a transaction.
“There have been huge advances in terms of application connectivity to process content through the user interface or API’s, and with the introduction of RPA and ML, robots can now classify, understand sentiment and suggest next best actions for content that is unstructured.”
Reducing the threats posed by third party contractors
Invest in tools, with care
While the need to enlist assistance from additional software is valid, organisations must be careful about overspending, and ensure that the tools they invest in are for a clear, specific purpose.
“Businesses have a lot of unstructured data in many different formats throughout their organisation, whether that be documents, emails or even system-based data that’s not structured, such as payment data for reconciliations,” said Chris Porter, CEO of NexBotix. “That causes an issue for RPA, which can only handle structured, rule-based digital processes.
“There’s a couple of different ways for customers to overcome these shortcomings. One is to buy a tailored point solution like an OCR tool, which can extract data from documents, or they could invest in a workflow tool to help them orchestrate robots and humans, or perhaps buy some machine learning from Google to try and extract insights from their complex documents. These tools are designed to solve a very narrow set of problems, within tight parameters.
“However, each of these has its own technical challenges; when embarking on one of these projects, you face significant cost, plus you need the right skills and tech to support each initiative. Each use case needs to be treated as an individual project, because you’re effectively buying for that particular need, and if you have lots of different types of data in your organisation, lots of different processes that have this level of unstructured data, you need to start again each time and buy the right solution to fix each individual problem.
“The key is applying the right technology to solve the right problems but doing so in a in a scalable way that focuses on business value. For example, we have off-the-shelf invoice processing which we can implement in any company leveraging reusable components and automating the end to end business process in the accounts payable. We have already done the hard work to build that and make it work for the customer.”
How to maximise your cloud computing investment
A final way to overcome RPA’s drawbacks around content processing is to implement extra capabilities.
Neil Murphy, global vice-president at ABBYY, explained: “The biggest challenge with RPA is that it is unable to process unstructured content such as invoices, emails, forms, receipts, or correspondence. However, companies can – and do – overcome this.
“All it takes is content intelligence ‘skills’ that make RPA bots smarter by adding cognitive capabilities, such as analysing, understanding, and processing unstructured content. Organisations can deploy these content intelligence skills with easy to use no code or low code solutions, which enable their staff to build RPA bots that can handle a vast array of documents.
“Already, we are seeing adoption across all sizes of business where the technological barrier to entry is removed by such an approach. This in turn is driving innovation – some companies are now combining these skills to offer an advanced cognitive understanding of complex use cases. Customer onboarding is a good example, where there is a multitude of documents that need to be processed, from identity documents and onboarding forms, to bank statements and proof of address.”