Back to Resources

Big data and AI: 3 real-world use cases

By Kevin Casey | October 25, 2019

The relationship between AI and big data is a two-way street, to be sure: Artificial intelligence success depends largely on high-quality data, and lots of it. Managing massive amounts of data and deriving value from it, meanwhile, increasingly depends upon technologies such as machine learning (ML) or natural language processing (NLP) to solve problems that would be too burdensome for humans to contend with on their own.

When big data meets AI: Use cases across industries

Let’s take a closer at one piece of that broader cycle: Examples of how AI can be used as a powerful lever with big data, whether that’s for analytics, improved customer experiences, new efficiencies, or other purposes. Consider these three significant possibilities for how AI and big data can become a productive pair.

1. Gleaning structured data from non-standardized sources

70 percent or more of enterprise data is unstructured, analysts estimate.

The challenges of big data can be plenty – storing it in a usable, cost-effective manner, for example. The “usable” piece can be particularly tricky when it comes to unstructured data, which accounts for the lion’s share – 70 percent or more, according to some estimates – of enterprise data. (When people talk about how big data will inevitably continue to get bigger, unstructured data is the big driver of that growth.)

Turning unstructured information into usable formats can be a beast of a chore for humans, especially in repetitive (but entirely necessary) back-office operations.

Mathias Golombek, CTO at Exasol, points to invoice processing as a specific example that illuminates the broader possibilities of using AI to automatically extract structured data from unstructured (or non-standard) formats.

“An example [of how AI can be applied to big data] would be training a model that learns from scanned invoices and the historical data of extracted structured data: invoice ID, due date, recipient, etc,” Golombek says. “This information normally had to be interpreted by human beings since every invoice looks a bit different, has different names or languages. But if you use the historical data of thousands of invoices, it’s possible to create a model that could give you automatically the structured data by just scanning new invoices.”

This same principle of using AI to automatically extract structured data from unstructured sources could apply widely, not just to operational areas like finance or HR, but to the broad (and often unglamorous) category of enterprise content management. This is a potential boon for data analytics, robotic process automation (RPA) and other forms of automation, and other purposes.

“Organizations are using AI to change their most valuable asset – their content. Up to 90 percent of enterprise content is unstructured and growing [at a rate of] up to 65 percent per year,” says Anthony Macciola, chief innovation officer at ABBYY. “Most unstructured data goes unanalyzed, leading to valuable information being lost and unusable. With AI, organizations are transforming unstructured data into actionable information that can be used within intelligent automation systems. This enables business leaders to make better business decisions faster.”

2. Streamlining complex and bureaucratic processes

Where there’s big data, there’s often complexity and bureaucracy. Think sectors such as healthcare, insurance, and financial services, which have plenty of both. As a result, these industries are increasingly experimenting (if not moving full-speed ahead) with the potential ways AI technologies can be used to cut red tape and generally improve processes and outcomes amid complex requirements around compliance and other issues.

Let’s focus on the financial sector for a deeper example:

Fintech, or financial technology, illustrates perfectly how AI/ML is shifting how banking institutions provide financial services to consumers today,” says Sameer Dixit, general manager of data, analytics, and AI/ML at Persistent Systems. “Back-office operations at banks involve large and complex data sets that are labor-intensive. When handled by robotic process automation [combined with] AI/ML, there’s significant savings on time and costs when performing tasks such as ‘know your customer’ checks, where the identity and address of the customer is verified. The loan process itself is also labor-intensive. With AI/ML, the ability to reduce costs and offer loans at more attractive rates to those with limited credit history is widening a previously underserved market.”

Arvind Jagannath, director of product management at AI Foundry, points to the mortgage lending industry as a specific subset of the financial industry that is currently experimenting with AI.

"Artificial intelligence is improving data analysis in the mortgage industry right now in a number of ways,” Jagannath says, pointing to three areas as examples of where it can deliver benefits to lenders and customers alike:

  • Throughput: “The industry average to close a mortgage right now is about three to four weeks. Using AI to automate ‘critical path processes,’ you can get mortgage processing down to just a matter of days,” Jagannath says. “This increase in throughput makes the home-buying experience faster and less stressful for the home buyer and helps banks and other lenders process more loans faster.”
  • Speed of analysis: Loan processing is, in a sense, another way of saying information processing. And with mortgages, there’s a whole lot of it. AI can speed this up, to the point of real-time processing. “AI is increasingly being used at the point of sale to provide more borrower self-service,” Jagannath says.
  • Accuracy in processing and outcomes: “Using AI and automation, you can process mortgages with high accuracy rates,” Jagannath says. “Humans can get tired, and that fatigue can lead to mistakes, while AI can work on a 24-7 basis with no fatigue and high accuracy."

Of course, financial, healthcare, and other companies are going to have to fight AI bias as they cut that red tape.

3. Making better use of video and voice assets

When you think of media formats that can produce inherently “big” data in various organizations, voice and video typically come to mind. Both provide examples of how AI can be applied to improve how companies manage and derive value from existing media assets, or to improve how they use these and other formats going forward.

Brian Atkiss, director of advanced analytics at Anexinet, notes that AI disciplines like NLP create considerable new improvements in how organizations use their voice data, from speech analytics to voice-to-text transcription.

Every time you’ve ever heard “this call may be recorded for quality assurance and training purposes,” big data is getting bigger.

Moreover, AI can solve the challenges associated with the underlying data, such as huge storage impacts. Every time you’ve ever heard “this call may be recorded for quality assurance and training purposes” is, in effect, making big data bigger.

“Previously companies would store call recording data for manual review and compliance reasons, in some cases for seven years or even longer. This data was recorded in mono (single-channel) and heavily compressed to reduce file sizes and storage costs,” Atkiss explains. “With the advances of speech-to-text algorithms, this call recording data has suddenly become a treasure trove of useful data that companies can utilize for measuring customer experience and improving operational performance.”

The AI-driven opportunity for new analytics also reinvented the storage challenge associated with call recordings and other voice data. Those mono recordings won’t cut it.

“Higher-quality audio files produce much better accuracy from speech-to-text algorithms, so companies need to use uncompressed audio, which can be more expensive to store,” Atkiss explains. Enter AI again, this time for its ability to transcribe voice recordings automatically.

“These recordings can now be transcribed in real-time or near real-time, and the resulting transcripts provide a record of the call and can be used for advanced analytics,” Atkiss says. “These text transcripts can be stored while the high-quality uncompressed audio files are able to now be deleted and don’t need to be stored. The ability of companies to provide real-time access to this data has also required advances in how the data is stored and processed.”

Video can present similar opportunities and challenges. AI is now enabling businesses to better manage and find value in their enterprise video assets.

“AI technology empowers businesses to understand and optimize video content libraries with advanced metadata enrichment and previously untapped insights,” said Chris Zaloumis, senior director of enterprise video offerings, IBM Watson Media. “From heightening engagement and increasing discoverability to automating closed captions and furthering inclusivity, AI arms businesses with the tools necessary to operate in a truly global, always-on environment.”

Speech-to-text technologies can be a huge help in terms of improving the accessibility and inclusivity of video applications, including in live feeds: “Practical applications like AI-powered live and on-demand automated captioning help bridge the communication gap for ESL employees and members of the deaf and hard of hearing community,” Zaloumis says.

view article online+