This video training reviews the basics for adding content or responses to your VirtualAdvisor. Over time, maintaining the content and performing frequent training of the VirtualAdvisor will drastically improve the responses received by students.
Step 1: Review and Override Global Content
As part of the product, CampusLogic creates and constantly maintains a set of global financial aid content to jump start the use of your VirtualAdvisor bot. You can override the content to personalize for institution specific needs and voice if you so choose.
Global Versus Institutional Content
Override Content You Want to Personalize
Step 2: Create New Institutional Content
Train your VirtualAdvisor to answer frequently asked questions from you students for financial aid and/or other departments by creating your own content and training the AI to use your content.
See the following best practice articles to help you develop consistent and understandable responses:
Add New Reponses
Edit or Add Alternate Phrasing for a Response
Alternate phrasing is used to teach the VirtualAdvisors multiple ways a student can ask the same question.
Step 3: Consider Multi-Turn Question
In some cases, a simple response is challenging because there are either multiple options ("It depends") or the VirtualAdvisor needs more context. A multi-turn response allows you to design a conversation with the student using a tree structure. While these responses can take more effort up front, they deliver significant value by improving the guidance a student receives.
Step 4: Test and Publish
VirtualAdvisor allows for new content and training updates to be tested prior to being published. Only published content is seen by students.
Step 5: Monitor Training Workflow and Train VirtualAdvisor Bot
As students use the VirtualAdvisor, training is continually required to improve the responses through adding more content and training, testing, and publishing more responses and improved alternative phrasing. To facilitate the is process, the VirtualAdvisor includes a Training Workflow that identifies the following training opportunities:
- No Response: A question is asked and the AI cannot identify the correct response. In this case the VA answers the student that it does not know the answer.
- Low Confidence: An answer is provided to the student but the AI confidence score is low enough that the response warrants review.