star-bg.jpg

CognitiveScale Blog

Getting Started in AI: Five Entry Points

By Akshay Sabhikhi Feb 8, 2018 9:53:57 AM

Artificial intelligence has quickly gone from science fiction to mainstream headline news. While some enterprises are reluctant to embrace change, others are overwhelmed by the possibilities. Where and how can AI improve established and next-generation business practices?

5-entry-points.png

Based on over 50 customer engagements, I have found that there are five powerful ways to introduce AI within an enterprise that are particularly ripe for success. These five entry points have been successfully tried and tested in industries as diverse as financial services, healthcare, and digital commerce to deliver measurable business outcomes. As you read about each of these five entry points, consider your own organization and think of areas that would most benefit from AI.

People:

If it is done right, AI provides an amazing chance to personalize human engagement. Think for a moment about the times you have been pandered to as a member of a voting bloc, a market segment, a persona, or even worse, just an account number. Most of us recognize that those interactions are merely arms-length transactions. Stated differently, automation without personalization leads to alienation. They don’t really know you as an individual. They know how to contact you and to ask for something from you. What if, instead, you received a text or phone call from your financial advisor who had informed you that your favorite industry was likely to be positively impacted by a market event that had occurred only seconds before? And what if that same communication recommended that you consider purchasing a particular stock that was poised to benefit from that bit of news? In addition to this, what if you were provided the rationale behind the recommendation, tying together your stated interests, goals, investing style, and preferences? Lastly, what if that financial planner regularly received these timely and insightful recommendations for each client based on continuously improving analysis of declared, observed, and inferred information? Automation alone cannot provide this level of personalization, but AI can. And AI-driven personalization can drive new levels of client, customer, and patient satisfaction and loyalty at scale.

Process:

In addition to novel ways to improve human engagement, AI also drives terrific improvements in process efficiencies. Consider an invoice processing operation at a large healthcare provider. Staggering volumes of invoices are processed every day, and all of them need to be reviewed for accuracy. Each invoice can have multiple numerical codes as well as lengthy, jargon-filled descriptions of procured goods and services. But how can you efficiently analyze each of these documents to uncover arcane errors? Once again, AI provides the solution. A successful approach to this problem is to use natural language processing (NLP) to confirm that the text-based descriptions match the numerical codes, and to use machine learning (ML) to process ongoing feedback from employees to the machine to assess the machine’s ability to find errors and to avoid false positives. One client used this system to drive process intelligence into its invoice exception handling operations to save millions of dollars annually.

Product:

Nearly every company wishes to make its products and services more compelling, more differentiated. Ideally that differentiation would help justify charging a premium in the market. One clever example of this is a television broadcaster which hosted the largest sporting event of the year. For legal purposes, let’s call it “the big game.” Ads for the big game are among the most expensive in the world. But what elements actually make an ad successful? If you are paying $5 million for a 30 second ad, you want to know what works and what doesn’t. The broadcaster sought our help. We used deep learning computer vision modeling for attribute generation and compared the results with audience engagement. So, what drives audience engagement? Is it flashy graphics? Smiling faces? Puppies? AI provided the answer and helped the broadcaster achieve significantly deeper insights into ad efficacy as well as a much-expanded feature set.

Data:

We are drowning in data and yet the information we need - information that helps us make better decisions – is often just out of reach. It’s a frustrating situation for workers and managers alike. Recently, a global financial services firm grappled with maintaining compliance while controlling costs in a dynamic environment. It relied on manual and expensive legacy processes to determine which external and internal rules and regulations applied to which of its disparate databases and other IT assets. These error-prone processes created multiple blind spots and exposed the bank to unnecessary risks including audit gaps, violations of corporate compliance, legal issues, and possible punitive fines. The bank had the information it needed, but it did not have a map to get to it.

In response, the bank deployed AI software that used IT asset metadata to map enterprise concepts and applicable enterprise controls to those concepts, building a bridge between the massive volume of IT assets and policies that correspond to those assets. The bank’s efforts not only increased compliance and decreased costs, it also unlocked the value of its previous IT and big data investments.

Knowledge:

Unlocking the value of your data residing in your systems is one thing. Capturing and systematizing your enterprise knowledge is another. I’ll explain. We recently helped a large cancer hospital with a project to use AI to analyze its bad debts (bills that go unpaid for 120 days or more). Given that there are tens of millions of dollars at stake, knowing this number is an important element in Revenue Cycle Management (RCM). The hospital traditionally relied on the combined subject matter expertise of its team to estimate their expected bad debt. Ingesting millions of patient account records from their EHR (electronic health record) system, our AI software used machine learning to filter records and apply the domain knowledge of the hospital’s experts. Now, more than 90% of bad debt dollars are correctly identified. As important, the hospital has new options including intervening with offering payment plans, discounts, and enhanced communication with the patients to prevent accounts ending up as bad debt.

In previous years, executives rightly wondered if AI was ready for prime time. Those days are thankfully gone. There are so many success stories throughout the enterprise that the real question is no longer if AI will work for the business, but rather, where in the enterprise to get started to make the biggest impact right away. To get your AI journey started on the right foot, I recommend a structured approach that puts a premium on speedy results and choosing an AI software partner with a proven track record.

About the Author:

Akshay Sabhikhi is CEO at CognitiveScale. Follow him @sabhikhi on Twitter.

About CognitiveScale

CognitiveScale sells augmented intelligence software that solves complex business problems at scale for financial services, healthcare, and digital commerce markets. Built on its Cortex augmented intelligence platform, the company’s software help enterprises apply AI and blockchain technology to increase user engagement, improve decision-making, and deploy self-learning and self-assuring business processes.

Headquartered in Austin, Texas, CognitiveScale has offices in New York, London, and Hyderabad, India, and is funded by Norwest Venture Partners, Intel Capital, IBM Watson, and Microsoft Ventures.