There are few AI experts with in-the-trenches enterprise experience like Akshay Sabhikhi. Together with his co-founders, the CEO of CognitiveScale, a leading provider of industry-specific artificial intelligence software, was part of the IBM Watson team and has worked to deploy AI solutions at some of the world’s largest organizations since the company began four years ago.
“Shay” and the CognitiveScale leadership believe that humans have far exceeded the threshold for interpreting all of the data surrounding us and that machine-augmented intelligence will drive break-through levels of engagement and productivity. The company was started in order to bring practical and scalable AI to the enterprise using an augmented intelligence approach to attack challenging industry vertical problems in healthcare, financial services, and commerce.
I asked Shay about what he sees as the major AI developments of 2017 and 2018, what types of enterprise-class AI solutions are gaining traction, and how CognitiveScale reconciles the human/machine intelligence trade-off for its clients.
M.R. Rangaswami: Thinking of 2017, what are the major milestones that shaped the AI space? And what developments do you foresee for AI in 2018?
Akshay “Shay: Sabhikhi: We’ve been in this business for about seven years and so we’ve seen the market evolve significantly over that period. In the early days, experts spent a lot of time “pushing” AI and what it could do for companies. In 2017, we really began to see enterprises “pulling” us in and asking to understand how AI can drive their businesses. In the same way that mobile, cloud, and social strategies got going, the drive to adopt AI typically begins with boards telling C-level executives, “We’ve gotta do something about AI.”
This year, we have seen many clients proactively reaching out to us asking for an education on what enterprise AI initiatives make sense. We spent the year carefully choosing clients because, with an emerging technology like AI, you have to be careful not to engage with clients who live in a constant education cycle but never end up doing anything else about it.
Today, AI has reached a level of democratization we didn’t anticipate. With the toolkit libraries available from Google, Microsoft, and IBM, developers at large companies are like kids in a candy store: They’re excitedly looking at all the possibilities but don’t know where to start. This is where a company like ours steps in: We can leverage our enterprise know-how, apply the right AI technology, make it work at scale and deliver value to the company.
Looking forward to 2018, we expect AI to continue its progress into mainstream enterprise IT. We plan to see more industries and digital commerce companies infusing AI into core business processes.
In the past, AI experimentation and development was often relegated to an “innovation lab.” IT leaders threw a few experiments into this AI bucket and waited to see how each one panned out. These tests show how AI works but didn’t have any impact on the business at all. Now we’re seeing concrete deployments of AI being pulled into mission-critical business processes and delivering success for some of the world’s largest companies – and we expect to see many more of these in 2018.
Enterprises are also getting a more clear idea of who should be leading AI initiatives. In the past, most enterprises equated AI with data science. They hired really smart data science experts to develop enterprise AI applications but few were successful. We now recognize data science and enterprise developers as two separate worlds that function completely differently. Delivering a successful enterprise AI app involves collaboration between both spheres of expertise.
At CognitiveScale, we’ve always looked at the market through a vertical lens: We understand it is more useful to provide AI solutions that go narrow and deep, and build domain expertise rather than build a general purpose application that relates somewhat well across the entire organization.
We’ve found this approach is a must-have for compliance-driven industries. Many of our clients have to worry about the auditability and explainability of an AI solution – it cannot be just a black box driving some outcome. Auditors work with true evidence so as AI moves into the mainstream, clients must provide explainability of insight for all types of audit, compliance, and liability issues. Today, humans provide copious documentation of these decisions and machines must do the same.
M.R.: How has CognitiveScale worked with clients to realize the best AI has to offer? What challenges have you overcome as you have grown the company?
Shay: We’re proud that CognitiveScale has nearly 30 customers who have used our two AI solutions to solve specific business problems on both the back end and the front end.
“Amplify,” our back office product line, uses AI to improve the efficiency and cost-effectiveness of complex back office processes. Take the voluminous nature of claims sent to insurance companies by a major hospital. When claims are submitted, approximately 75 percent are likely to be denied for any number of reasons and must be resubmitted – sometimes several times before the insurance company will pay the claim. This can keep $500 million to $1 billion in revenue tied up in collections for an average of 83 days.
Applying AI to the process enables inspection of all claims before submission. Using prior history and recent activity, the AI engine identifies those claims likely to be denied and suggests how to revise them before submission. This machine-augmented intelligence keeps the claims – and the revenue – from getting stuck in the submission-resubmission cycle. The hospital realized a 20 percent improvement in first-time claim acceptance rate.
An AI application of this type is not limited to healthcare – it has positive implications for process improvement in a variety of industries. Applying augmented intelligence to an accounts receivable operation improved the accuracy of the team from 64 percent to 94 percent and resulted in a $12 million savings for the company within 90 days of deployment.
Looking at AI’s impact for one of our financial services clients reveals AI’s ability to improve the performance of mundane back office processes: data retention. Banks have to retain data in certain locations for specific amounts of times. Banks have to maintain audit teams who go through and look at many thousands of systems to constantly determine which are in compliance and which aren’t. AI enabled the bank to quickly go through many data sources and locate violations that need to be corrected. Again, this ability to augment human power with machine intelligence delivered increased productivity and a reduced likelihood of fines for the bank.
Looking at the “front end,” our “Engage” product has helped clients use AI-powered personalization to personally interact with their customers. About 90 percent of all site visitors don’t end up buying anything or authenticating themselves in any way. Even sites that use the latest and greatest in engagement solutions only end up with massive buckets of thousands of similar unauthenticated visitors – for example, upper income, middle-aged women in the Northwest.
AI enables sites to drive better engagement and conversion – the ultimate online retail metric. Even when a visit isn’t authenticated, AI can deliver a personal profile of the visitor based on what other sites were visited, what products are hovered over, what fills other sites’ shopping baskets, and the “DNA” of those products – not just what they were but their characteristics, reviews, images, and more.
The visitor profile enables digital commerce sites to present the products most likely to be appealing based on personal preferences, navigation profiles, and many more profiling measures. The goal is to increase time on the site and have it translate to more purchases. In just a few months, our clients have witnessed a substantial jump in engagement and conversion. This application is not limited to digital retailers – it applies to banking clients, healthcare patients, and many other online consumers.
M.R.: How do you work with clients to best understand the relative strengths of human intelligence and machine intelligence?
Shay: If you approach a client with an AI solution that replaces human workers, it isn’t going to fly. Not only is the technology too young and untrusted, but you will find resistance at all levels of the enterprise – from the C-level to the rank-and-file workers who feel they could be replaced by machines.
The human worker is still extremely important when it comes to AI solutions – that’s why we talk about “AI” standing for “Augmented Intelligence.” We wrote an eBook to demonstrate how augmented intelligence is AI app with the biggest potential. Enhancing the productivity and insight of human workers and providing credibility, auditability and documentation for the decisions has taken off in a big way – and we predict it will become even more important in 2018.