Please Try a Different Browser

You are using an outdated browser that is not compatible with our website content. For an optimal viewing experience, please upgrade to Microsoft Edge or view our site on a different browser.

If you choose to continue using this browser, content and functionality will be limited.

Robotic Process Automation
Blog

The Connection Between Information Capture and Artificial Intelligence

Congratulations, you’ve made it to the final installment of our six-part blog series on how information capture can revolutionize your document and data processing. This series wouldn’t be complete without some discussion of where information capture is headed. I can sum up in two words: artificial intelligence (AI). Thanks to the rise of graphics processing units (GPUs), data analytics, cloud computing, and decades of research, AI has now hit the mainstream, and everyone is jumping on board.

But what exactly is AI? AI is used as a broad term for any “intelligent” thing a machine can do, but I will use AI in this blog to refer to algorithms that help computers understand, decide, learn and predict to better automate business processes that require information capture.

AI in this context consists of:

  • Natural Language Processing (NLP): algorithms that help a computer understand natural human language contained in text-rich documents and data sources (contracts, letters, emails, websites, social), and make that language actionable.
  • Machine Learning and Cognitive Automation: machine learning algorithms enable computers to “learn” and apply knowledge to new situations. Cognitive automation includes algorithms that learn, estimate, plan and decide, just like humans would.

HfS Research projects AI automation spending to increase from $1.6B in 2018 to $2.7B in 2021, and AI-based information capture will play a big role in this growth.

Natural Language Processing (NLP)

Information capture extracts raw data from documents and electronic data sources, but has historically struggled to deliver contextual understanding of that extracted data for better decision-making. For example, if a customer email contains the words “account opened,” “change,” “address on file” and “bad,” what is the intent of the document? Is it a complaint about bad service for a recently opened account that needed an address changed? Or a change to a beneficiary name on a recently opened policy at a particular address due to a marriage gone bad? Is it a report of a poor account opening experience due to a bad address on file, and a notification the customer will change providers?

In this instance, none of the above. The email actually stated, “My account opened recently needs a change due to a bad accident to vehicle X. My address on file is still the same.” Raw data extraction wasn’t enough. What was needed was the context of that data. As you can see in this example, without automation or a clear highlighting of key information, humans are needed to manually read the entire email and determine next actions. This is costly and slows the processing of the request.

NLP can automatically make sense of:

  • Classification (type of content)
  • Topic/summary (what it’s about)
  • Keywords/phrases/entities (key data points)
  • Sentiment (positive or negative content)
  • Intent/purpose/request (action being requested)

NLP makes sense of the context of the document/data to help the user take appropriate action. Organizations can leverage NLP technology in information capture use cases such as:

  • Assisting operators with processing and responding to incoming emails (open/close account, add/cancel policy, change beneficiary, change address, change vehicle)
  • Comparing extracted entities with black lists/white lists to support fraud or compliance detection
  • Processing contracts and agreements – assisting operators with understanding key players, locations, key dates and amounts, and the context of what these entities represent
  • Processing website pages and social media streams to gauge customer sentiment and competitive position
  • Improving internal document and database search capabilities, both for documents/data that already exist in a system of record, and new incoming documents/data

Machine Learning and Cognitive Automation

Machine learning and cognitive automation enable computers to learn and then act on that knowledge, just like humans would. Take the machine learning of documents for more automated processing. The system learns from document samples provided initially, and then gets smarter over time as it learns from the actions of humans manually processing new documents and variations of known ones.

Machine learning can also be applied to robotic process automation (RPA), as the system learns the repetitive, manual steps taken by a human computer user, and then generates a robot that mimics those steps, freeing up the human for more value-added work.

Machine learning and cognitive automation also find use in predictive analytics. Business owners not only want to know what happened historically, but also want to use that data to automatically improve business operations and profitability. Cognitive automation can use document, data and process analytics to build models that learn, estimate, plan and automatically decide on the best action to take depending on the situation – just as a human would. For example, if a customer onboarding system encounters a new applicant profile that closely matches prior applicants who were manually rejected, the system can dynamically adjust settings to automatically reject the new applicant without human intervention.

Machine learning also helps make information capture as-a-service a reality. Cloud-based information capture lends itself to document and data learning at a massive scale, growing its knowledge base over time as more and more users submit their information and get back the results from the cloud. With cloud-based learning and knowledge bases, you can now submit an invoice to a cloud capture service, and your invoice data will be returned to you automatically or sent to your ERP system for processing, with little to no human involvement. Expect to see vertically-oriented knowledge bases in the cloud in the future; for example, your mortgage package will be automatically classified and extracted, and encrypted data and images will be returned to you or sent to your loan origination system.

Where Does AI Go From Here?

As we’ve seen in this blog series, information capture increases visibility, boosts operational efficiency, ensures compliance and improves the customer experience through digitally transforming business processes.

Going forward, 451 Research sees AI playing a key role in enhancing customer engagement through applications such as:

  • Virtual Assistant
  • Forecast Management
  • Content Recommendations
  • Sales Conversation
  • Engagement Insight
  • Graph and Influence
  • Response Recommendation
  • Lead Prioritization
  • Product Recommendations
  • Data Hygiene

AI Expertise You Can Use Today

We will be at the forefront of AI for years to come, but you may not realize we have been developing and selling software based on AI algorithms for over a decade. Information Capture software employs AI technologies such as machine learning, classification with machine learning and natural language processing to perform document clustering and classification, OCR, data extraction, document detection, facial recognition, signature verification and signer identification

To learn more about how best to start your digital transformation journey with information capture, download this free AIIM report.