In response to COVID-19, financial software giant Intuit launched free tools to help businesses navigate government aid and relief.
The package of tools, together called Intuit Aid Assist, help small businesses and self-employed people in the U.S. understand if they qualify for financial aid under the $2 trillion federal aid package dubbed the CARES Act. With Aid Assist, which is powered by Intuit AI technology, users can determine if they qualify for a loan through the Paycheck Protection Program (PPP) or Employee Retention Credit, a refundable tax credit designed to keep employees on their company payrolls.
The vendor released the Intuit Aid Assist product April 12, just before the traditional tax filing deadline of April 15, which was extended to July 15 this year because of the coronavirus pandemic.
Yi Ng, principal product manager at Intuit Futures and software developer at Intuit, leads knowledge engineering at the company and helped create the new Aid Assist tools. She started at Intuit nine years ago, and has spent most of that time working on Quickbooks. Ng helped build the first early version of QuickBooks Self-Employed, a product that simplifies taxes for independent contractors and freelancers.
Knowledge engineering is a field of AI that attempts to emulate how a domain expert would approach a domain problem. It's the process of representing human judgement and behaviors within an AI system.
What are the new Aid Assist tools?
Yi Ng: Small business, they're hurting because they're not able to serve their customers, not able to open. The U.S. government is starting to look at some of the government reliefs that are available.
One of the things that we as a product company are starting to look at is what are the capabilities that we have to serve our customers and how can we help them. So, when the CARES Act came out … I went through the first legislation and the bills and start reading through it, asking how this helps our small businesses. When you think about government relief, and the work that we do at Intuit, and how we can help our customers, the first thing that came to mind was, well, we have a lot of technology, such as knowledge engineering.
So, we started starting to put together some ideas, including calculation tools and ways to get businesses information about the PPP. We interviewed a lot of small businesses trying to figure out how well they understand the PPP, asking questions like if they know how to get access to the loans. They're calling banks or talking to their accountants, but they don't really have any way of figuring it out.
We decided that we needed to create something on their behalf and make it so that it's available to everybody. That's how Intuit Aid Assist was born, which is really about creating a centralized site that gives everybody -- small businesses or self-employed -- the ability to access information, so they can walk through a personalized set of steps to help them understand what their specific situation is, what they have access to and how the relief program can help them.
What AI technologies help power Aid Assist?
Ng: At Intuit, we basically put our AI capabilities onto three pillars: machine learning, which many people are familiar with; natural language processing; and knowledge engineering, which is more nuanced.
Yi NgPrincipal product manager, Intuit
Knowledge engineering is a way of approaching leveraging data first and a way of thinking about how to leverage the relationship between data. To give an example -- in Aid Assist, we ask questions like, are you self-employed? Or, are you a small business with many employees?
Depending on their answers, it takes them down a different path of product experience. If self-employed, there are certain things that they qualify for and there are certain things they do not qualify for. If they're not self-employed, we ask questions about how many employees they have. If it's more than 500, that takes them down one path, while less than 500 takes them down a different path.
So, this knowledge engineering capability is about capturing the relationship between data in the knowledge graph, and really codifying the rules that we already know in the human mind.
While machine learning is about taking data and generating insights, basically leveraging the machine capability and deep learning to generate insights that we may or may not know, knowledge engineering is about leveraging what we already know. For Aid Assist, that includes the information in the CARES Act. We take that information and codify it in such a way that it becomes a personalized experience for customers by relating data to each other and understand what data generates the next set of screens, what kind of information generates the next set of strings, so they can go through the process and feel that the product was designed for them.
As you mentioned, we hear a lot more about machine learning than knowledge engineering. Do not a lot of companies do that?
Ng: Right, I think that's true. As you know, Intuit is a financial services company. Financial services has a lot of details, a lot of compliance requirements behind the scenes. We have a trust that we have built with our customers that we are the go-to for their financial needs. That level of trust, that level of detail, that level of understanding comes from helping them and having their back and helping them with compliance.
From that perspective, Intuit is very much in need of a technology like knowledge engineering to be able to have the rules-based capabilities, in addition to machine learning and NLP.
The need for it, I think, depends on the domain. We need this because, from a tax perspective, a lot of things are already captured in the tax forms.
From keeping your books, keeping your taxes, doing your finances, a lot of those rules are already here for us to help guide customers. There are, of course, a lot of insights that we may not necessarily know out of the box, out of the form, and that's where like machine learning and natural language processing comes into play.
I believe we still have ways to go in combining machine learning and knowledge engineering to start creating an even smarter AI platform going forward.
Editor's note: This interview has been edited for clarity and conciseness.