Intuit's CDO talks complex AI project to improve finances

Intuit Inc. may be a 35-year-old company, but it's still looking to disrupt how personal and small-business finances are managed. One vision? A digital financial assistant powered by cutting-edge AI tech that would transform how the company operates and how its customers save, spend and invest.

"In the United States today, we're living in a world where, I've heard, between 40% and 50% of Americans don't have $400 dollars to deal with a small emergency," said Ashok Srivastava, senior vice president and chief data officer at Intuit, based in Mountain View, Calif. "So, if they have a flat tire, if the transmission goes out in their car or something else happens, it can be a very significant event in their financial life and potentially in their overall life."

Srivastava, who is Intuit's first chief data officer, said he's in charge of developing a vision and a strategy for the company's "entire data pipeline," from the way data is captured to the way data is ingested by the algorithms that power the company's data products.

Here, Srivastava talks about the company's vision for a digital financial assistant -- an AI project that he described as being both expansive and innovative.

Editor's note: This interview has been edited for brevity and clarity.

What's the most exciting AI project at Intuit right now?

Ashok Srivastava

Ashok Srivastava: We're excited about this idea of creating a digital financial assistant -- we call it a digital persona. This would be a service that enables consumers and small businesses to get better information that empowers them to make better financial decisions.

How would customers interact with this digital persona?

Srivastava: It would allow for different types of applications to be tied to it. The most obvious is a chat bot, but it could also provide augmented intelligence for our customer care representative; it could provide Augmented intelligence for our accountants who are working on Intuit's behalf or private accountants who are using Intuit software. It could be deployed in internal processes where product teams learn how people interact with our software through a set of focus groups. So, it's one technology that could be instantiated across many different platforms and touchpoints.

That's one of the exciting aspects from a technology perspective. If you think about how a human works, there are so many things that are amazing about humans, but one is that they have the ability to rapidly change contexts and rapidly deal with a changing environment. The touchpoint doesn't matter. It doesn't matter if you're talking on video, on the phone or in person. Generally speaking, people can deal with these channels of communication very easily.

But it's hard for technology to do that. Technology tends to be built for a specific channel and optimized for that channel. And what we're trying to do is build a single technology that would then interact with people through multiple channels.

Talk about the underlying technology in this AI project. I assume natural language processing will be a major component. But what else?

Srivastava: It's going to draw from multiple technology sources. So, certainly, natural language processing and natural language understanding are fundamental. Closely related to that is what in the AI field is called 'reasoning under uncertainty,' so making decisions in situations where there isn't complete and perfect information. This is quite a difficult task. Closely related to that is information fusion. So, the idea that you have multiple data sources -- some of them might be owned by Intuit, some of them might be third-party information sources -- and bringing them together through a technology base that allows for information fusion. That's another important capability that needs to be built.

There are the conversational aspects. So, if the channel is a conversational channel, that has an entire technology stack associated to it -- whether it's speech-to-text or text-to-speech capabilities, those types of things become quite important.

What's the status of the AI project?

Srivastava: We have parts of this fielded already. When I say fielded, I'm talking about examples of the capabilities we're trying to build. We're working on platform development and scaling the technologies that would allow us to meet the ultimate goal that I described. But we have capabilities already fielded in customer care, as well as in QuickBooks, which is a chat bot Intuit customers can interact with today to get insights about how their small business is operating and other aspects like that. But it's in its infancy with respect to the overall vision that I'm describing.