By Tom Haunert
From web store recommendation engines to autonomous cars and databases, artificial intelligence (AI) is already part of your personal life and business processes. But AI is also part of today’s business applications, and it’s assuming an ever-larger role in time-saving, money-saving, and money-generating business processes.
Oracle Magazine sat down with Clive Swan, senior vice president of Oracle Adaptive Intelligent Apps at Oracle, to talk about evaluating when to use AI, measuring the value of AI in the enterprise, Oracle’s AI for apps strategy, and more.
Oracle Magazine: How do businesses evaluate or consider various emerging technology application and platform choices?
Swan: The first thing is, as with any other technology, they’ve got to make choices based on business needs and not just choose a new technology for technology’s sake. Businesses need to identify target use cases with a target end state in mind. They need to identify where that emerging technology can apply, and then they need to plan to adopt that technology, based on the skill sets they have available, their attitude about risk, and how quickly they want to achieve a win.
Businesses also often need to make a decision between build and buy. Do they have the resources in-house and access to the specialist skill sets required to build it themselves? For example, data scientists—an expensive and rare commodity—are needed to build AI solutions. So even if a business does have ready access to data scientists, the company still needs to decide whether a given use case is sufficiently differentiating to the business to warrant using those data scientists to develop a solution in-house instead of buying one off the shelf.
We’re shifting the user experience from something user-initiated and user-driven to AI push-driven activity.”
Whether businesses build or buy, there are three components that are core to how AI solutions work. There’s the data from the applications, there’s the real-time context in the applications at that point in time, and then there’s measuring the outcome in the applications. For AI solutions to adapt and learn, they have to assess the efficacy of their recommendations very rapidly and frequently.
Businesses can buy third-party AI point solutions, and those solutions can be loosely integrated with ERP and human capital management (HCM) suites, for example. But in many cases, those point solutions will be suboptimal, because they can’t be integrated deeply enough within the base applications to have full and timely access to the three core AI data components I mentioned. Alternatively, AI solutions can be a part of the core ERP and HCM suites, as they are with Oracle’s offerings, with tightly connected AI component processes, leading to optimal AI performance and requiring no additional integration work.
There are cases where it makes sense to build a new custom AI application based on the type of data and how it will be used. For example, car insurers can train a neural net with photographs of thousands of cars that have been in crashes to determine which cars are most likely a write-off. In such cases, the data and context are lightly coupled with the base application, so a custom standalone AI application makes sense.
Oracle Magazine: How do companies look at or measure the value of AI in back- office applications?
Swan: The preliminary benefits can often be measured quite readily, but the related benefits, which are often more substantial, are much harder to measure and probably aren’t even recognized in many cases.
First, looking at AI today, the majority of solutions are not doing something that a human being couldn’t do. But AI solutions do a better job, because they do those tasks very quickly; never take a tea break; and execute predictably, the same way, time after time.
So I would argue that in many cases, when people start measuring the benefit of an AI solution, that benefit is measured in terms of productivity gains. But there are additional positive side effects—employees’ spending less time on mundane tasks results in improved employee retention and frees the employees up to make more-strategic contributions to the business.
For example, consider a recruiter who needs to find 10 candidates for further phone screening from 100 résumés submitted. That recruiter might take two hours to review 50 résumés and then pick the best 10 from that 50. An AI solution doing just as good a job as the recruiter in ranking those résumés will scan all 100 résumés and statistically find a better set of 10 candidates. And it will do it in a few minutes, plus the recruiter will have gained two hours to spend on more-valuable tasks, such as selling the company to potential candidates.
Oracle Magazine: What is Oracle’s apps AI strategy?
Swan: We term Oracle’s apps AI strategy “pervasive AI,” and there are three components to it.
First, we’ve got Oracle Adaptive Intelligent Apps, which are ready-to-go, out-of-the-box, fully integrated, functional application extensions that are typically involved in making recommendations and decisions for professional users, consumers, and end users in our applications. We’re delivering these solutions across the enterprise application suite, including B2C, customer experience, B2B CX, ERP, HCM, and supply chain.
To give a few examples: In Accounts Payable, the app recommends to the accounts payable professional which suppliers will generate the most savings if paid early. In Sales, the app can predict the likelihood of a sales rep closing a given deal and recommend the next actions to take to increase the likelihood of winning that deal. Commerce apps recommend products for consumers, and marketing apps recommend the time of day for sending a communication to a consumer and over what channel.
In many cases, when people start measuring the benefit of an AI solution, that benefit is measured in terms of productivity gains.”
Then, we’ve got intelligent UX [user experience]. We’re shifting the user experience from something user-initiated and user-driven to an AI push-driven activity in which AI nudges tell users what they need to know and what they need to do.
And third, we’ve got virtual assistants. We’re bringing the conversational interaction models from consumer apps into enterprise apps. These solutions use AI, of course, not only to map voice to text but also to interpret what the request is and then serve up the best response, taking into account the context of the requester.
And all of these solutions are underpinned by Oracle’s smart-data strategy, which enriches application data with third-party, web-scale, trusted data, enabling our AI solutions to make even better decisions.
LEARN more about AI and machine learning.
LEARN more about Oracle Adaptive Intelligent Apps.
Photography by John Blythe