At Intuit, our mission is powering prosperity around the world. One of our strategies for achieving this is around delivering personalized experiences. Our educational programs are one of the ways we can deliver on that strategy.
Our Intuit ProConnect™ Lacerte® and ProSeries® customers can enjoy our online training portal within “My Account,” where they can access nearly 100 different pieces of content, from white papers to bite-sized video content less than 10 minutes long on average. We have had more than 30,000 users access these portals and take advantage of the valuable training contained within.
Machine learning technology allows personalization within this experience. The portal currently will suggest relevant content to users based on a module that they take. While users engage in their suggested content about 15 percent of the time, we feel like the level of personalization to an individual user could be much stronger. The way we can get there is through the continued implementation and evolution of AI and machine learning models.
The interesting thing about using AI to drive personalized experiences is that it is not a single algorithm or set of code; it leverages a series of models that create a holistic set of criteria by which to feed an engine that suggests the right content to the right person in the right way. It can also continue to be enhanced (or enhance itself) as time goes on, without the need for additional implementation. Simply put, it just gets better over time as it continues to get more data that feeds the precision of the recommendation engine.
In our case, we will use four models to drive the recommendation engine. Here is an overview of how they will work:
- Training Engagement Model – Analyzes the content leveraged by the individual user and aggregate user data to “score” individual modules that feed the recommendation engine.
- Multi-Class Customer Profile Model (Neural Network) – Looks at customer product use profile to assess things like their tenure with the product (how long they have used a given product), which products they own and other factors.
- User/Content Similarity Model – Assesses similar (and non-similar) user behavior and suggests content based on similarities with other users.
- Multi-Dimensional Categorical Model (Neural Network) – Assesses and buckets content in a network of categories (vs. a singular category) that adds multiple dimensions of content analysis and suggestion to the user based on what we know about them. Further, this model will allow for new content to be automatically assessed by the entire AI model instantly. Today, a user must engage in content at least once for the AI model to apply the algorithms to it. This is because the current models only assess existing interactions with the content and apply to behaviors and profile of a user.
To access this exciting training portal, you must be a current Lacerte or ProSeries user with an active Fast Path License. Simply log in to your My Account online portal and browse the training section from the left-hand navigation bar.
Editor’s note: Check out another one of Jasen Stine’s articles on artificial intelligence. This article was originally published Sept. 17, 2018, and updated on Jan. 6, 2020.