Video produced by Steve Nathans-Kelly
At Data Summit Connect 2020, Elliott Ning, cloud advisor, Google, discussed how AI has replaced business intelligence as the key driver of strategic decision-making.
Full videos of Data Summit Connect 2020 presentations are available at www.dbta.com/DBTA-Downloads/WhitePapers.
There are four categories of analytics use case, Ning explained. First is descriptive analytics, like aggregating business data during a period of time. For example, getting information about total revenue from last month, last quarter, or last year. Next is diagnostic which entails examining data with content to troubleshoot a problem, like to discover a bottleneck in the system at scale. Predictive is making predictions about future events, like, what are my predicted sales numbers for the next quarter? The last one is prescriptive. Prescriptive is finding out the best course of action on a given situation. For example, autonomous driving. The cars make tons of calculations and decide when to speed up, when to slow down, turn left or turn right, Ning said.
"So you can see: To get a summary or conclusions in the past took business intelligence. To take actions now or make predictions in the future is AI. To me, AI is not a new thing. It's an extension of BI. If you have a solid BI foundation, you will have a much higher chance to win with AI. If you don't, I would highly suggest you to build one," Ning said.
Organizations that have a good BI systems can start looking for a problem in the right analytics category for a good UI AI use case.
"Then what about data? You should have data. And if you don't, start collecting today. Think about the problem you're trying to solve and what might be relevant. Sometimes it's easier, but sometimes not," Ning said. "If you have the right data for training a model before you're actually doing it, if you have different kinds of data and are trying to decide, where to start, I will recommend you start with structural."
According to Google's own research, data in this basic form, structured, is likely to drive the most AI impact.
"Why? Because you should have structured data sitting in some sort of database or data warehouse, like sales data, user data, product data, it's processed. So you don't have to have a separate pipeline to transform it," Ning said. "There may be some minimal activities like changing formats, comforting types. It's organized and clean. So you can quickly run quarterlies to determine the basic patterns and trends."