AI IRL Podcast Episode 11: How Intuit’s AI handles money smartly
If AI is where the smart money is going, how does AI handle the smart money?
We’ve all read about how Wall Street firms have invested billions in AI and machine learning, but what about the people on Main Street?
To find out, I spoke with Bharath Kadaba, Chief Innovation Officer at Intuit. Intuit’s brands include personal finance products like Quicken and Mint, business accounting software like Quickbooks, and TurboTax the #1 best-selling tax preparation software to file taxes online.
Kadaba runs a group within Intuit called Technology Futures.
“I have a team of people looking across our businesses, understanding our customer needs, but focused not just on what can we do today,” says Kadaba. “We are also focusing on what can we do for our customers tomorrow as well.”
Where AI excels
Intuit’s space is really around compliance, like doing your taxes as well as managing your finances, whether you’re a small business or a consumer.
Because AI is good about following rules, compliance is a natural fit, from paying payroll taxes to making sure all procedures are followed.
When AI meets large datasets, it gets really powerful. “We do have an incredible number of customers for whom we are managing their financial daily life,” says Kadaba. “And that means we know a lot about their finances, which are numbers, and this lends itself to artificial intelligence and machine learning.”
AI can help users not only track their past spending, but help them find their financial future as well.
The future of taxes
“When we think about taxes, we think about the fact that very few people in the world actually enjoy the task of doing the taxes.” — Bharath Kadaba
Intuit is using AI to create a process they call “Taxes are done.”
“Wouldn’t it be nice for a customer to simply say it’s tax time and you hit a button and it says, ‘I took care of the whole thing and I just need one or two pieces of data,’” says Kadaba. “And then you do that and you’re done.”
Intuit is using the classic rules-based AI to quantify all the IRS regulations so that the software can prefill all of the data fields that a user needs. Saving time on data entry alone is a mental health lifesaver.
The software’s AI will ask you a few questions that determine which forms it needs. Using datasets from similar customers, the software streamlines the process with the forms pre-filled based on your profile and answers.
“It takes about 108 minutes to do your taxes on average. Year after year we measure how much of that we can take away using knowledge engineering,” says Kadaba. “Looking at people like you who filed taxes with us, we can, with a great amount of confidence, say you should be doing standard deductions versus itemized deductions.”
This machine learning can shave at least 30 minutes off your tax preparation time.
Other time savers include auto-importing forms like W2s and 1099s. If you don’t import your tax forms, you can import the PDFs or scan the paper copies and the software is smart enough to take that data and pre-fill your tax forms.
Your personal virtual CFO
Starting in 2011, when Apple’s Siri debuted, Intuit began to create a Quickbooks assistant that would respond to queries about your finances. They launched the service in 2017 for Quickbooks Self-Employed customers.
“The idea was to get this assistant to answer the questions that the customers have in a way that they want to ask those questions,” says Kadaba. “Today you can ask a question like ‘how much money did I make this month?’ or ‘how much money did I make in August?’”
The assistant will show you your estimated taxes with graphs and charts. It is able to handle program queries like “how do I send an invoice?” which saves the customer time and also saves the company expensive call center time.
The dawn of the conversational, empathic AI
The goal is to get the AI to a level where it can not only answer your questions, but anticipate your next questions.
The idea is to make the dialogue more conversational so that users can understand their finances by engaging in what-if discussions that the software can answer using the customer’s real data.
The next big challenge is to bring empathy to that conversation. In order to do that, the AI has to understand the emotional import of the data it delivers.
“For example, the tax refund is for lots of people the biggest paycheck of the year. And when we do a great job of finding you the maximum refund, that’s a delightful experience,” says Kadaba. “We want the AI to show some emotion: ‘Wow, look at this refund! This is brilliant!’”
But not all financial news is happy. What if the AI tells you that you can’t afford to go out tonight? The AI needs to deliver that message sympathetically.
Many people fear their finances and the possible judgment of a human financial advisor. So instead they ostrich whenever money is mentioned.
An empathic AI could have a nonjudgmental heart-to-artificial heart intervention with the user to make a plan so they can meet their financial and recreational spending goals.
AI’s greatest trick: Making itself disappear
As more people use voice to interact with their technology, Intuit’s Future Group sees a time when the AI as part of a program disappears.
“In 2025, we envision a world where going to a machine and logging on and interacting with a two-dimensional visual based paradigm is not the way to do it,” says Kadaba.
“You want to be able to get to this concept of ambient computing or ambient interactions where intelligence is surrounding you everywhere and it follows you and you’re able to simply interact with it without having to go through that human/machine barrier.” — Bharath Kadaba