Where Does AI Fit in Business Decision Making? (VIDEO)

Video produced by Steve Nathans-Kelly

In his closing keynote at Data Summit 2019, John O’Brien, principal advisor and chief researcher, Radiant Advisors, shared a clear path forward for data and analytics leaders based on the evolving concepts in data and analytics within the context of an enterprise-scale data strategy. Data managers must understand how to make the biggest impacts with advances in AI and machine learning for analytics, enable self-service with governance, and support BI and data engineering processes with the data lake.

“Where does AI fit?” asked O’Brien. “The goal is business decision making, and one of things that we have done research on over the years—and that we keep coming back to and I hear today—is augmented intelligence. AI is augmented, it's not artificial; it's assistive intelligence, recommendations.”

DBTA’s next Data Summit conference will be held May 19-20, 2020, in Boston, with pre-conference workshops on Monday, May 18.

Often, when people look at data in their organizations, they will say, "That's not right" or "Oh, that's wrong and that should be cleaned up" or "Oh, wait, that's from when we did that change a year ago,” and “That's kind of the old business rules, but, you know, we just overloaded that data column."

So, said O’Brien, there's a lot of the human side, which is still involved with the idea that people simply know it when they see it. “Empowering those people, broadly, is one strategic move you should make. As you move into having the human do self-service work, the original hypothesis—‘Hey, I have a hunch, I want to go see some data’—that is humans.”

AI doesn't do that, but organizations can now have AI come in and do recommendations, said O’Brien. Many tools are self-service data prep tools or data visualization tools that enable users to bring in datasets and they then make recommendations, for example, with suggestions such as “These two data columns are likely to be joined, is that what you want? Inner join? Outer join?" or “This is probably the best graph or chart to use in representing your data and what you're trying to communicate.”  In order to make that development or that self-service to move faster, AI is coming in and assisting people and that helps them become faster, said O’Brien.

“And then, of course, on the far right-hand side, you can get to a point that, after you've figured something out, built the model, got the training data, and you're comfortable with what it's doing,” said O’Brien, where you turn it loose and get into that in-the-moment, hey this just happened, immediate reaction because a lot of times, engagement with customers or within systems happens within seconds, and that is the window of opportunity to act.

“If I went back to that first generation and second generation, what that line really represents for me—and I think we saw this in a couple of presentations this afternoon—is scale,” said O’Brien. “Today, we have companies that ingest one petabyte per day in volume of data. You can't have enough analysts to go through that data, you can't have enough people working on it, analyzing it, making decisions with it.” The only approach for scaling to keep up is to deploy the new architectures that work with AI. AI can then be deployed thousands of times all over the place—working and learning, and improving on the spot because we can't scale.

“So, we've moved into this next generation of data, now the scalability aspect is two-fold,” said O’Brien. “One is the volume that we have to tackle. Two is the complexity. AI is better at complexity; it will find patterns based on things. And, if you think about changing worlds, retraining the data is easier than trying to code.”

Today, what is going is a move to probabilistic reasoning, which does not say, here is the answer, but instead, there is an 83% probability that this is the right answer with this confidence level, and that good enough to move forward, said O’Brien. Having that kind of accuracy is a vast improvement over what we used to have before, he noted. “So, what you find is this scale trying to tackle complexity, increased complexity, increased volumes, and that's where AI is allowing us to take over—because we can't.”

Many presenters have made their slide decks available on the Data Summit 2019 website at

To access John O’Brien’s full Data Summit 2019 keynote titled, “Bring it Home: How to Advance Your Analytic Strategies, go to