At U.S. Bancorp, David Berglund sorts hype, promise of artificial intelligence

Almost nothing gets as much buzz in executive suites these days as artificial intelligence, or AI. David Berglund began studying it as a technologist at UnitedHealth Group and developed expertise in a particular area, natural language processing. Last year, he joined U.S. Bancorp as senior vice president of innovation, a job with multiple roles. He keeps on top of what is developing in the field to consider what ideas and innovations in AI may solve a problem or need at the nation's fifth-largest bank. He also keeps up with AI projects in the company's various business units. Amid all the hype, Berglund often tells his colleagues, "AI does not break the laws of business." It's just another tool, Berglund says, though one that is still not well understood and that seems spooky to some.

Q: What do you mean when you say "AI doesn't break the laws of business?"

A: There are still core problems and opportunities we have at the bank.

If we can be more effective and more efficient in personalizing experiences, for instance, that's beneficial to customers and to shareholders.

Q: What is one way AI can help with that basic function of the bank?

A: We're storing more data about our customers. We have far more advanced computing power and better models, which means we can find patterns in the data that before were lost on us. We can do a better job with machine learning and other AI methods.

Q: You recently gave a speech to an industry group in which you described AI as being more than Siri but less than HAL and Skynet [the scary computers of sci-fi movie fame]. What did you mean by that?

People often jump to "Terminator" and the fear of Skynet or HAL refusing to open the pod bay door [in "2001: A Space Odyssey."] The problem then is we forget about the near-term challenges we have with AI, such as creating echo chambers or bias in our data and reinforcing the mistrust people have in information. Those are the things we have to fix today. It can be distracting to think too much about the science-fiction stories.

Q: Do the science-fiction stories overpromise what's possible with AI?

A: We get promised this future of the human race being impacted in positive ways with robotics and super-intelligence.

But I often joke about that we see flying skateboards in the movies and we're stuck with Lime and Bird instead. Which is great. I appreciate the chance to go between buildings quickly, but it's not the fantasy future that many of us thought we were going to have.

We have all been frustrated by auto-correct.

A: There's two sides to the work I do. One is driving the use of AI so it can be maximally beneficial to our customers, our organization and our shareholders.

To do that, we've had to create a strategy and a risk framework and make sure we're aligned with technology and business partners. We needed alignment with colleagues in privacy and ethics, some core folks across the bank. And by bringing in folks from the business lines who know the problems they have, I can connect the dots to a number of different solutions, whether it's from the startup community, large tech corporations, consulting partners or, of course, we have our own muscle.

The other side of the work is to build products and capabilities.

You can't have one without the other. If all I had was the strategy, I wouldn't know if it was working as well as I do by being involved in the process.

Q: You developed expertise in natural language processing. Where is U.S. Bank on that?

Bringing customers in to interact with some prototypes put U.S. Bank in a good position to act once things like Google Assistant, the advanced version of Siri and Amazon Alexa became available. We were the first big bank to be on those three major voice platforms. That wouldn't have been possible if they hadn't followed the instinct to learn early.

Q: Where else will voice and language processing pay off for the company?

A: We've got a tremendous amount of information that is unstructured and that we historically haven't been able to derive a lot of insight from, whether that's phone calls or chat logs or other interactions. If we can unlock that information through natural language processing, that will save us a lot of time so folks can work on other, more impactful things.

Q: For business executives and owners, is AI hype or something to invest in?