Becoming an even more trusted advisor

Auditors, bankers and underwriters spend a considerable chunk of their time wading through appraisals, applications, client histories, evaluations, scoring data and so on. Miss an important detail and they can put the organization at risk. Within the life insurance sector, for instance, underwriters have to pour through every page of a customer’s application, which can be several hundred pages long. If the underwriter misses a relevant fact or issue on page 352, for example, the company could be exposed.

Auditors face many of the same challenges. Accounting firms want to demonstrate they’ve given the audit their deepest, most considered approach. But there is a limit to the number of documents an auditor can check, so they assess a sample set and render an opinion. As with any manual process, errors can occur. Although an auditor has combed through data thoroughly, a regulatory agency may still come back with a finding, and that could jeopardize the audit. Cognitive technology allows auditors to greatly improve due diligence; instead of testing a sampling of data, they can test 100 percent and release opinions with far greater degrees of accuracy.

As a result, client teams reduce margins of error and achieve greater consistency. They process more information in much less time, and hit their quality targets repeatedly. As Konstantin Konstaninovsky, Director and Project Manager of Audit Cognitive Technologies at KPMG noted, “At the end of the day, we could have kept the process about automation and efficiency. But cognitive is more than an efficiency play. It’s about improving service and being able to enable our staff to do their job better.”

Connecting the dots faster

Similarly, cognitive technology allows financial services professionals to connect the dots faster, so they can spend their time doing the job they’re trained to do. Underwriters, for instance, often spend 50 percent or more of their time gathering information. They need to review customer applications, medical questionnaires, insurance questionnaires, attending physician’s statements and other background material. They must then flip through the industry’s “bible,” an underwriting manual that is roughly 8,000 pages long, to study the guidance for any findings or impairments. Only then can they determine whether to decline the policy, accept the risk at a standard rate or accept the policy at a premium rate.

Cognitive platforms accelerate the research, evaluation and approval process. A cognitive initiative underway at one European life insurer, for example, will provide underwriters with a dashboard that aggregates key patterns and findings from all pertinent documentation. “The whole synopsis is right at the underwriter’s fingertips,” said Julian Selz, an IBM Client Innovation Executive with Watson Health, “so underwriters can start right in, look at what is important, see if there are any concerns, and then take a position on how much risk the institution might be prepared to take.”

Creating sustainable differentiation

Leading financial services firms want to do more than improve productivity and tamp down risk, they want to create new markets. Cognitive technology allows firms to combine public and private content, such as benchmarks and financial reports, with private knowledge, such as their own subject matter expertise, to generate breakthrough approaches to assess, manage and price risk.

“Consider a firm that specializes in transfer pricing,” said Marc Teerlink, Chief Business Strategist at IBM. “With cognitive, it’s not that you get 60 hours of transfer pricing down to five minutes. It’s that you get many multiples of 60 hours down to single minutes.”

When that happens at scale, it opens enormous opportunities. Rather than having a junior staff member spend their days chasing down information, they can now use their professional training to model a range of scenarios that creates new value and services for their clients. For the transfer pricing team, noted Teerlink, that means instead of three months’ throughput time, they can bring the client into a new market three months faster. They can do so by anticipating all the variables needed to create a transfer pricing contract that protects the business and satisfies the needs of the local tax authorities.

With cognitive technology, the financial services industry has the opportunity to shift away from transactional engagements and take on more reflective and consultative work. Enabling that kind of innovation, however, requires them to shake off some of their traditional conservatism and define business scenarios that are predicated on driving top-line results and not just bottom-line savings. One executive told us that their goal was to have a transformational impact. They wanted to lead with a cognitive capability that was bold in ambition and that delivered significant and measurable returns in order to build interest and buy-in across the company.

Take the life insurance industry for example, where the ability to price risk accurately has always been a cornerstone. Whereas current methods allow underwriters to gauge the relative risk for individual health factors in the aggregate, such as smoking, it becomes challenging to price policies for customers who may have a combination of health issues. The customer may have great cholesterol and lifestyle habits, but they’re also asthmatic and have a family history of heart disease. Cognitive technologies can absorb reams of health histories, medical data, actuarial data and other information to make underwriting precisely tailored so customers aren’t being dinged unnecessarily, and life insurers aren’t being unwittingly exposed.

The ability to do this repeatedly and accurately at scale has the potential to revolutionize the life insurance sector, giving firms the ability to price competitively, sell directly to customers and deflect the advances of hedge funds and other new entrants. Martin Lippert, EVP and head of global technology and operations at MetLife said, “We see cognitive computing as a way to bring more humanity in the way we interact with our customers. As our ability improves with respect to being able to price to an individual based specifically on that individual’s profile, it begins to move us away from risk pools. So we see cognitive computing as potentially having a very significant impact on MetLife as well as the industry.”

Getting the most from cognitive technology

While cognitive platforms carry extraordinary promise, cognitive initiatives need careful planning. They often require significant investment and resources, a high tolerance for trial and error and enormous amounts of vetted, high-quality data.

Many of the biggest challenges concern data quality. Flip through any mortgage statement, credit application, financial report, appraisal or insurance policy binder, and you’ll see page after page of tables, charts and text. Very little of these are standardized, either within a business line or institution, or among ratings agencies, banks and other bodies. Scanned documents such as faxes and PDFs present special complications. In addition, tables may have nested rows and columns that need to be broken out, translated and reformatted before they can be fed into the cognitive system. PDFs and faxed materials can stretch the limits of current optical character recognition (OCR) technologies. Some may have been scanned two or three times before they enter a file, making them grainy and hard for cognitive systems to understand. These issues add to the ramp-up time needed for cognitive implementations.

“Tables within documents present a significant challenge” said Holt Adams, an Executive IT Architect at IBM. “The semantics of textual content at the intersection of a table’s columns and rows are not always obvious, especially when nested and hierarchical headings are used inconsistently.” The industry doesn’t have a common approach to address these issues currently. However, IBM is working with a number of banks and other financial institutions to capture and refine the set of requirements to allow programmatic extraction and annotation of key concepts and their relationships.

Normalization is another big technical challenge. Within the financial services sector, risk rating scales vary widely. Some scales use alphabetic labels, others numeric ones. A bank may have a Triple A rating for their lowest risk assessment, but another a risk rating of 1. How do you compare the two? Likewise, dates and currencies are frequently expressed in different forms. Putting these scales and references on a common footing is critical. That requires building data models that can work with data in different formats and structures and reuse the information in a meaningful way. That’s doable—but can take significant time and expertise depending on the scope of the solution.

Finally, given the heightened regulatory environment, effective data curation is critical. To ensure compliance, financial institutions need to be clear on what data sources they’re using, how data is being created, and how it’s being managed. Once the data is collected, it often requires cleansing and processing before the cognitive system can ingest it for analysis. Many organizations are aware of the need to create policies with specific criteria for business controls, but when it comes to cognitive initiatives, the management and curation of data is often one of the biggest things cognitive customers underestimate. An organization needs relevant, high quality data, and enough of it to allow the cognitive platform to generate statistically valid hypotheses. Companies that have documented decision-making processes and have strong knowledge management systems will be able to train cognitive platforms faster.