Around this time last year, the chatter around artificial intelligence (AI) was escalating as Salesforce and IBM unveiled splashy programs to help advertisers make sense of massive and complex reams of data, in order to enable more personalized marketing.
Oracle, too, announced the creation of Adaptive Intelligence Apps at the time, and at its annual OpenWorld conference this week, it offered a more comprehensive view of the AI-based capabilities it’s been beta-testing across four of its SaaS clouds, including Enterprise Resource Management, Supply Chain Management, Human Capital Management and Customer Experience, which houses the Marketing Cloud.
The Adaptive Intelligence Apps, which are prebuilt and baked into the cloud applications, will be released as an upgrade over the next year.
“Eventually, all of Oracle's applications, all their business applications and capabilities, will have this Adaptive Intelligence built into them,” said Jack Berkowitz, VP of products and data science for the Oracle Adaptive Intelligence group.
When the company began designing the Adaptive Intelligence Apps, the goal was to immediately deliver value to users while the apps reacted and learned to improve accuracy.
“What people might do is, say, run a bunch of customer data though it in the first day or two,” said Des Cahill, vice president and head CX evangelist for Oracle. “Then its accuracy is going to keep growing and growing, especially as it starts interacting in the wild, meaning in real ecommerce and digital marketing situations. But we've pre-tuned it based on our experience with existing digital marketing and ecommerce companies that are beta-testing it.”
Berkowitz and Cahill declined to discuss initial accuracy rates. Berkowitz noted that the data that fuels the Adaptive Intelligence Apps – users’ first-party and third-party data from the Oracle Data Cloud – first provides much of the context.
“The second step to the accuracy is these models we've built using our partners to train them,” Berkowitz said. “The third step is: We don't share your data with other customers. You get a generic model, but then the model actually adapts to the shape of your data, your customers and your products automatically.”
Finally, the active learning system is continuously changing, click by click, and constantly optimizing.
For marketing, Oracle is looking at use cases in three areas. The first is ensuring that the content, offer or product being given by the marketer is the right one. The second area is in acquisition, via better lead or channel management, for instance. The third area is optimizing engagement patterns so that marketing messages are delivered in the best, most relevant way without annoying customers.
In one scenario, a retailer may use Adaptive Intelligence Offers to launch a dynamic ad campaign to sell umbrellas. While one recipient who opens the email the same day will see an offer for the end-of-season umbrellas, another recipient who opens the email on a warm day a week later may see a more relevant ad for sundresses.
Another capability called Optimized Marketing Orchestrations allows marketers to use Responsys to determine the best channel to reach a consumer with relevant messaging.
The company’s AI program has been several years in the making, Berkowitz said. The company recognized years ago that the business applications space was dividing up with three essential elements.
One is data; Oracle believed that applications would become increasingly data-driven. So around 2014, Oracle began beefing up its Data Cloud with several notable acquisitions “specifically to attack programmatic advertising but with a secondary impact that we could actually use that data in other ways,” Berkowitz said.
The second element is artificial intelligence and machine learning. Oracle built the Adaptive Intelligence Apps both in-house and using best-of-breed open-source technologies, with a team that included folks from some of the companies it acquired and key hires from outside Oracle, including from major retailers and Google.
And as Oracle built out its suite of business applications, it focused on the ability to combine the three layers.
“So, with the domain knowledge of the applications, the data and then the machine learning and artificial intelligence on top of it,” Berkowitz said, “we actually had a nugget that we could go to market with.”