By Chris Murphy
When Oracle President of Product Development Thomas Kurian mapped out the technology path ahead for Oracle during his Oracle OpenWorld keynote, he cast the 10-year effort to deliver all of the company’s technologies via the cloud as just a starting point. The next wave of innovation, Kurian said, will include weaving new technologies such as artificial intelligence (AI), the Internet of Things (IoT), and new forms of human interface into those cloud foundations.
The result, said Kurian, is “a canvas on which you can paint your vision and your ambitions and dreams; to use information technology in a fundamentally new way; to transform your organization, your companies, and the world.”
Kurian’s keynote brought these grand ideas to life with demos. In preparation, members of his development team created the scenario of a basketball team, the Huskies, using cloud-based technology to quickly launch new platforms for fans to engage via social media. Kurian and his team members then took the stage to show how the Huskies organization could use cloud-based AI, analytics, marketing, and IoT systems to turn that fan engagement into revenue. Here are snapshots of the five technology demos supporting that vision.
Demo #1: Cloud Infrastructure for AI-Powered Apps
The first demo of the presentation began with an app idea—one that lets a fan snap a photo of a player and upload it to get stats on that player and vote for him as the league’s most valuable player. What this simple concept revealed, however, was the cloud infrastructure necessary to support that kind of AI-powered app. Such an app requires machine learning, because it needs to be trained on a bank of images to recognize players. That requires an infrastructure using high-performance GPU microprocessors—plus the ability to handle the streaming data of images and to use open source machine learning libraries such as TensorFlow.
The following three steps are needed to set that up in Oracle Cloud Infrastructure using only a browser: 1) set up a multiple-GPU infrastructure cluster by clicking the preferred options for a virtual cloud network and for the cluster itself, such as the operating system, bare metal servers or virtual machines, the number of nodes, and the like; 2) set up a streaming data service to ingest the images; and 3) train and run an AI application using that image data.
Once the streaming data service pulls in tens of thousands of player images, the app needs a deep learning model to recognize a player and pull stats on him. Oracle’s infrastructure supports leading open source machine learning libraries such as TensorFlow, which this demo used. The high-speed GPUs speed up the work, which means paying for less computing power than a slower option would require.
Demo #2: Let Fans Buy Tickets Using Facebook Messenger
Chatbots let customers use their favorite text platform or voice-powered assistant to ask questions such as, “Are there tickets for Sunday’s game?”—and the second demo showed how to quickly build such a chatbot for Facebook Messenger using Oracle Mobile Cloud.
Say a fan asks the bot the date of the next Huskies game. The bot responds with ticket options at various price ranges and then lets a fan go through to purchasing, without ever leaving the Facebook Messenger platform. Developers and non-developers alike can use the new bot-building capabilities of Oracle Mobile Cloud to build a bot and map how these conversations should flow. For example, creating dialogues of possible customer conversations starts with typing in a few different ways to say, “I’d like to buy tickets,” and the built-in natural language capability understands other variations. If a fan uses new phrasing, the platform’s “intent” capability shows the developer that the bot has an 85 percent confidence that it falls under the “buy tickets” option.
With machine learning, the bot also continually learns the different ways that fans ask similar questions. And tools for mapping the flow of conversation let a business manager insert options such as offering a discount for prepurchasing beer or wine before checkout. “This is a very visual, drag-and-drop way of building the dialogue flow,” said Oracle Vice President of Product Management Diby Malakar.
But fans don’t use only Facebook Messenger. One of the most valuable bot-builder features in Oracle Mobile Cloud is the connection to a range of messaging platforms such as Slack, WeChat, and voice-powered assistants. This lets developers build once and add a new channel by clicking these new options—rather than having to recode for all the variations of how message platforms display data and handle errors.
Demo #3: Analyze Fans’ Facebook, Twitter, and Instagram Activity
The Huskies business team wanted to better understand and analyze fans’ social media activity. So while the first demo showed how the team could ingest images from social media, the third demo showed new ways to analyze that data using image recognition and AI and to share it among team members via mobile channels. Kurian described Oracle’s analytics vision as allowing “anyone in the world, not just professional analysts, to be able to analyze any type of data from any datasource—not just numbers, but images, audio, video, textual data, sensor data.”
Oracle Analytics Cloud uses AI to make recommendations about how to cleanse data and parse it. Before analyzing image data gathered from social media, for example, it makes recommendations ranging from the simple (filling in missing zip codes) to the more complex (filtering out bot traffic based on typical bot behavior). The analytics platform also suggests ways to tag and sort image data, such as identifying brands, detecting certain objects such as shoes or basketball jerseys, and even identifying emotions.
We’re doing this to give you . . . a canvas on which you can paint your vision and your ambitions and dreams; to use information technology in a fundamentally new way; to transform your organization, your companies, and the world.”–Thomas Kurian, President of Product Development, Oracle
With the data cleansed, the system can make recommendations for analysis, such as identifying which players are most popular with different age groups of fans. Analysts can also run their own analysis by comparing different datasets. And, they can share that data in new ways, such as via a mobile app that offers a smart feed to automatically deliver analysis the system thinks is most important. Using natural language capability in the app, analysts can set up alerts—when social mentions of a star player exceed the 4 million mark, for example.
“We’ve moved from a world where analytics is a passive system, where you have to go ask questions, to a system that’s constantly looking out for the best interest of the company,” said Oracle Senior Director of Product Strategy Jacques Vigeant.
Demo #4: Recommend Products to Web Shoppers Using AI-Powered Apps
Tracking and understanding social engagement doesn’t help the Huskies if it doesn’t lead to higher ticket and merchandise sales. The fourth demo of the keynote showed how, by combining anonymized, third-party browsing history data from Oracle Data Cloud with first-party data such as an individual’s purchase history, the Huskies can provide personalized ads and offers in real time on web platforms, email, or chat. Across its cloud applications, such as Oracle Commerce Cloud and Oracle Marketing Cloud, Oracle is applying AI, through what it calls adaptive intelligence, to use data to make recommendations such as what product to offer when and what channel to use.
That kind of data-driven personalization is increasingly common, but doing it right takes two big things, said Oracle Vice President of CRM Product Management and Product Strategy Melissa Boxer. The first is having the volume and variety of high-quality data to power the system. And the second is being able to understand and react as the data changes, moment to moment.
“We’re able to capture changes in customer interests, or searches, or past orders, and process and rescore for every user, click by click,” as would-be customers buy or reveal new browsing interests, Boxer said. A data-as-a-service user can’t see a shopper’s browsing history directly but can see anonymized profiles based on people with interests and tendencies similar to a shopper.
Demo #5: IoT Plus Supply Chain Cloud and Chatbots
The final demo of the presentation envisioned an online shopper taking a picture of a Huskies player’s shoes, uploading it via a chatbot, and asking the bot to order those shoes in a particular size with gold lettering instead of black. Connecting such a bot to a supply chain means integrating with Oracle’s cloud applications for supply chain, order management, and manufacturing.
Once the shoes are in production, the process gets oversight from Oracle Internet of Things Production Monitoring Cloud, which can track data from thousands of sensor readings from machines on the factory floor. If it spots a serious problem, the IoT application can trigger a maintenance request, a technician can use Oracle Internet of Things Asset Monitoring Cloud Service to pinpoint the problem, and the supply chain app can flag orders that would be delayed by a production slowdown. Oracle Internet of Things Fleet Monitoring Cloud can track the goods to their destination.
The demos as a whole showed how companies can use emerging technologies in highly practical ways, and how Oracle has honed its cloud offerings to support and integrate those tools. As Kurian put it, “We’re infusing the new technologies of autonomous computing, artificial intelligence, IoT, blockchain, and new forms of human interface into our cloud offerings.”
Illustration by Wes Rowell