Rising to the data extraction challenge

Editorial Type: Feature Date: 2020-05-01 Views: 822 Tags: Document, AI, Workflow, BPM, RPA, Recognition, Hyperscience PDF Version:
Charlie Newark-French, COO of Hyperscience, outlines some best practice approaches for achieving 'document processing excellence' even during high pressure situations such as the current pandemic

Fast and easy document processing, which is critical to smooth-running operations in the best of times, becomes absolutely essential during a crisis. That's when organisations are confronted with new challenges and must find ways to do more with the same or less. Redundancies or weaknesses in operations often become more apparent during challenging times and create more problems if left unchecked. Organisations often have to let go of the old ways of doing things and rethink operations.

We're seeing this issue play out now on the world stage. Financial services firms, such as banks and insurance companies, as well as government agencies, have been hit with a deluge of forms, as people rush to take advantage of economic stimulus packages and interest rates.

In the UK, lenders haven't been able to issue money quickly enough due to the overwhelming number of applications, and similarly, bounce-back loans for small businesses are expected to be in high demand. Governments in the U.S. are feeling the pain as well. Operations at the state level are being stretched thin in response to record unemployment claims, and federal refunds are being delayed as the IRS struggles to process the huge volume of paper-based returns. Many financial institutions are hard-pressed to process the huge volume of small business loans and mortgage refinancing applications as well.

The difficulties in legacy document processing workflows are magnified during and following a crisis, when organisations return to usual operations and are faced with an influx of forms, on top of growing backlogs. Organisations need to be able to handle variable document volumes, even outside normal peak times, in an efficient and effective manner.

In the midst of all these challenges, many organisations are faced with a stark reality: their manual processes and outdated systems are failing them at a time when they need them most. Despite the desire to be "digital first," many organisations are still relying on manual processes, using data entry staff to index and extract data from handwritten or computer-generated forms and key it into their systems for processing. This is an inefficient, costly and error-prone process that is costing businesses nearly US$60 billion each year.

Other organisations are relying on offshore manual operations to handle data entry because of the reduced cost; however, due to quarantines, staff in locations such as India have been forced to stay home, and these outsourcers lack the IT infrastructure or contingency plans to keep operations running.

Even during normal times, these processes are inefficient - a stop-gap or band-aid approach that worked in part, but never delivered the quick and accurate processing of applications, claims and other forms at the scale that organisations need to stay agile and responsive in today's world. As organisations grapple with uncertainty, these pains are all the more salient, and relieving the administrative burden, re-allocating employee time, obtaining accurate data, and having the ability to quickly respond to customer needs have become more important than ever.

So, why has data extraction remained such a challenge? To start, documents are messy and vary in layout, quality and complexity. In addition, when it comes to high-value transactional processing, reliable and accurate data is critical, and one incorrect digit can be the difference between a mortgage application or claim being approved or rejected.

Following are some of the factors that have made document processing so difficult for organisations:

  • Many of the forms are handwritten, which make them hard to read. In addition to the challenges of handwritten text, a smudge or low-quality image can contribute to readability issues, for example making a "c" look like an "o." There could be scratched-out text, writing outside of the boxes and other issues that can further complicate a computer's ability to recognise and extract characters correctly. In the absence of reliable technology solutions for processing data, organisations have relied on humans to decipher hard-to-read words based on context and comprehension. Fortunately, advances in technology have bridged the gap between human understanding and machine processing, since handwriting, just like paper-based forms, will not be going away any time soon.
  • The lack of standardisation impairs machine-to-machine communication. It's hard for computers to read machine-generated forms largely due to the lack of global standardised formats. For example, there are a number of ways an invoice number can be represented by different vendors, such as "Inv. No.," "Invoice #" or "INV-#." This makes it hard for traditional rules-based technology systems to identify the page, locate the right field for extraction, and accurately extract the data for downstream processing.
  • Forms often have imperfections. Client-generated documents can often have fax marks, stamps, or other imperfections. If a scanner is used to digitise a paper document, it can result in faded print or uneven scan, all of which make it difficult for a computer to process reliably.

The good news is that many companies are turning to technology to help tackle these challenges and remove the inefficiencies and errors of manual processes and legacy systems.

To date, rudimentary automation technologies have been favoured because they're quick to implement and fit into existing processes, but they fail to unlock and lift unstructured document data. This data, which can be found in PDFs, images and other document types, doesn't have a pre-defined data model and is not in a format that can be used by a company's processing system - yet it makes up the bulk of an organisation's data stack. The ability to unlock and process these documents is critical to achieving meaningful efficiencies and driving end-to-end automation. The advent of artificial intelligence technologies, however, has been a game-changer, enabling companies to take a more sophisticated, robust approach to document processing. Machine learning (ML), for example, can train and continue to learn based on the data it's exposed to. With the right approach and training, ML-enabled solutions can capture, extract and classify unstructured data from a variety of documents and structure it into a format that is usable by the internal processing system.

This new category, Intelligent Document Processing (IDP), automates the ability to read pages with context in order to capture and process data with a high degree of accuracy, efficiency and speed. It not only enables enterprises to achieve cost savings, but also furthers workforce productivity and overall customer experience.

During crises and other challenging times when document processing inefficiencies are more glaring and painful, companies are particularly focused on ways to increase effectiveness. Here are some best practices organisations can implement that will enable them to scale when document volumes increase and also prime them for long-term success:

  • Take a good, hard look at your processes to identify where technology can have the greatest impact. Evaluate processes end-to-end to measure and quantify current bottlenecks or pain points. Given current critical needs and the evolving global economic situation, it's important to invest resources and budget in solutions that solve a key problem, are easy-to-deploy and add tangible value. Look for practical applications of AI and other key technologies rather than experimental initiatives.
  • Make the best use of human and computing resources. Now is not the time to waste people resources on low value, repetitive tasks, when advanced document processing technology can handle that with a high level of accuracy. Humans will continue to play a role by handling exceptions and providing the knowledge and context to process those hardest-to-read documents. However, by freeing up staff time previously needed for manual processing, technology enables employees to focus on higher-level activities like providing better customer service or uncovering new opportunities - at a time when it's particularly important.
  • Create a plan and framework for successful implementation. Map out processes needed to design, develop and implement your automation strategy. C-level buy-in and organisational adoption will be key, so make sure to include key stakeholders at the beginning of the process and every step of the way. Ensure the sustainability of the initiative by setting aside the ongoing resources and investment that will be required after deployment.

By implementing best practices and relying on the intelligence of advanced technology, organisations can meet critical need as market conditions evolve, elevate the customer experience, and position themselves for success in all economic climates.
More info: www.hyperscience.com