Point of View

How artificial intelligence helps financial services companies make sense of their data

  • Facebook
  • Twitter
  • Linkedin
  • Email

Business applications have made exponential progress in a remarkably short time. The one exception: gathering and automating high-quality, reliable data from diverse and constantly changing sources. This is the case despite the explosion in exchange protocols, better optical character recognition technologies, and a range of markup languages.

Not surprisingly, this hinders the end-to-end performance of well-tuned business processes. Manually extracting data, reworking it, handling errors, and managing exceptions can easily represent 80% of the operating costs of an automated process. Such inefficiency doesn’t make headlines, however, because businesses see the cost as necessary, even as they try to save every dollar.

The financial services world faces these challenges, too, although it has also made progress, such as XBRL (eXtensible Business Reporting Language). XBRL is a global standard for exchanging business information, including creating, extracting, and exchanging financial and business information, and a taxonomy for US GAAP reporting. In 2011, the Securities and Exchange Commission (SEC) mandated XBRL as the filing standard for all US public companies. In 2017, the SEC stated that foreign private issuers that prepare their financial statements in accordance with International Financial Reporting Standards (IFRS) as issued by the International Accounting Standards Board (IASB) must submit their financial statements in XBRL for fiscal periods ending on or after December 15, 2017.

Take a copy for yourself

Download PDF

A key challenge with XBRL is that it has little to no standard structure. The reason is that its many constituents have unique reporting needs and it must accommodate them all. This underscores the continuing challenge of source data extraction in the financial world. It’s still hard to normalize and interpret the data. Contending with paper documents and image files, too, only complicates matters.

Although there have been big technology improvements, data extraction solutions are still inadequate. The software development cycle simply isn’t flexible or iterative enough to handle documents that vary widely in form, format, and quality. Clearly, business needs a new paradigm.

Manual approaches still dominate the market for data extraction and normalization. Typically, companies outsource and offshore this work in an effort to drive scale, keep up with business needs, and advance downstream applications.

At Genpact, we’re delivering an artificial intelligence (AI) based alternative for our clients.

Visit our Cora LiveSpread page

Learn More

Genpact Cora LiveSpread - Our breakthrough AI product transforms financial spreading

  • Purpose-built for rapid implementation and scalability. Cora LiveSpread can be quickly deployed and is easily scalable to handle increasing amounts of financial information
  • Flexibility, agility — and service at scale. The product has a strong, patented, and rigorously stress-tested set of extraction components that have handled millions of transactions, every day, for over a decade

With Genpact’s patented technology, Cora LiveSpread extracts financial information from any source, including feeds, uploads, and paper. It lets you and your customers upload documents, whatever their format, through a custom web portal. It then imports the right numbers from the right places into the right format, and normalizes these numbers to get them ready for analysts. What’s more, Cora LiveSpread integrates with your firm’s existing internal and external systems, and can be modified rapidly on the fly.

Cora LiveSpread addresses all the fundamental automation challenges of source data variability, non-standard taxonomies, integration with existing systems, speed of implementation, and scalability.

Case Study: Scaling the implementation

Who we worked with
A leading financial services company with global reach

What the company needed
A simpler way to gather and interpret critical data arriving in a wide range of format

How we helped
With our Cora LiveSpread product and wealth of experience, we automated what was formerly a tedious, error-prone approach to data management.

What the company got

  • It automated the spreading of more than 150,000 financial statements annually
  • It had a working spreading prototype up and running for a set of public companies within 24 hours
  • It can handle documents for domestic, international, public, and private companies
  • It reduced process costs by 70%
  • It became future-ready with a product that is fully automated, scalable, and XBRLcompatible
  • It got a solution that handles images, electronic feeds, or paper documents — and does so securely


Bringing order to data chaos

Like other companies in its field, the firm faced conflicting priorities. Lowering costs while managing growing volumes was becoming more difficult. Then, there was the quality of its spreading process. And it saw no immediate or future possibility of automating.

The firm’s existing technology couldn’t rapidly extract, interpret, normalize, and spread multi-format, multisource financial data – and replacing or upgrading the system would not be cheap. Plus, doing so would be disruptive across the company – and the improvements weren’t guaranteed to work.

So, the company — again, like many of its peers — defaulted to a manual approach. And since managing credit risk is mission-critical for the firm, it kept adding people to address scale, improve data quality, better manage error levels, and so on. It tried to reorganize for process effectiveness and to fine-tune manual processes. But these efforts didn’t yield results and the firm wasn’t future-ready.

This frustrated the company. Sluggish manual extraction and spreading exposed the enterprise to greater risks. It was clear that XBRL wasn’t a silver bullet after all. The firm’s leadership recognized that it urgently needed a long-term solution. But if it was going to automate these processes, it faced the following challenges:

  • Source document variability: Financial statements are the raw material in the spreading procedure, so accurate and timely data extraction and processing are critical. The problem is financial statements can come in via Word, Intuit QuickBooks, and IRS tax returns. They might arrive as Excel files, PDFs, JPGs, or other image files, as well as faxed images or hard copy delivered via physical mail. This wide range of source documents was too much for the firm’s conventional extraction approaches and early attempts at automating had failed.
  • Statement variability: Thousands of statements were arriving in different formats and using non-standard taxonomies. The result was inconsistent labels for statement transactions. Analysts often had to “club” data into the next most appropriate field, often without fully understanding the impact of doing this. The situation was even worse for international companies.
  • Ensuring quality: Institutionalizing credit policy throughout the process and the organization was an ongoing challenge, as the onshore and offshore spreading teams kept expanding and turning over. Skilled financial analysts, who were up to date on current policy, had to handle the inevitable exceptions.
  • Internal/external systems: Access to the internal IT organization and the lead time for IT to build, test, and deploy an interface to internal and external systems was unreasonably long, even for a single interface. So, it was impossible to develop multiple interfaces and keep them updated Understanding and manually extracting files from new companies was a slow and error-prone process that took several cycles to stabilize.


Cora LiveSpread combined with our deep financial services domain expertise, and a great partnership

After evaluating the marketplace for solutions, the firm chose Cora LiveSpread. Within 24 hours, the firm had a working prototype up and running. The prototype handled spreads of financial statements from several companies that could be configured and updated in real time. This was well beyond the firm’s expectations.

Scaling up for success

Today, the firm processes 150,000 spreads for 45,000 public and private companies in North America, South America, Europe, and Asia Pacific. And it does so in multiple languages. It uploads financial statements in all formats, including XBRL. The system also takes images of hard copy, tags them, and makes them available for review and audit purposes.

Differentiating feature: Aggregation across multiple sources, including XBRL feeds, reporting mediums [web scraping, mail, feeds, excel, images] built on patented technology to extract data from financial statements.

The system extracts and uploads as-reported financials to a repository for easy review and analysis. The process incorporates automated, rules-based quality assurance steps that make certain, for example, that balance sheets balance after extraction. This drives accuracy at every turn and provides traceability for troubleshooting where required. Highly skilled Genpact personnel and client analysts handle exceptions through the application portal as necessary.

Secure, self-service document upload capability
Portal for comprehensive visibility of source and spread data for effective review, audit trail, and exception

Client-specified rules, such as IFRS and XBRL, govern normalization. And the system normalizes and interprets foreign financial statements. In addition, it normalizes non-US GAAP statements to US GAAP.

The company seamlessly and rapidly scaled up this solution across other parts of the enterprise, despite different business systems architecture throughout.

Visionary leadership, a best-in-class process technology platform, and a strong partnership with Genpact all worked together to give this global financial services firm what it wanted: a futureready solution that meets all of its current needs.

Delivering far more than expected

Our patented technology created an extensible, scalable, and highly flexible platform with potential well beyond the current scope. For example, over time the company can add features, such as scoring, rating, bureau access, rule-based generation of analytical reports, and risk-based pricing. Cora LiveSpread has all these features, configured and ready to go.

Financial statement spreading at a global financial services firm

Figure 1: High level process flow