For Higher Ed, Chatbots Need to Be Smart

With the onset of the COVID-19 pandemic, chatbots quickly became a critical tool for a variety of operations, especially in colleges and universities. The strictly virtual interactions with students dramatically increased the need to be digitally nimble enough to produce optimal online experiences. Chatbots facilitate a very useful function as part of that experience, but only if they can adequately and reliably serve peoples' needs.

Many institutions that purchased templated chatbots in the past have come to regret that decision. Complaints about the bot's inability to sufficiently answer questions (or in some cases, causing more confusion) are commonplace. Anyone who has found themselves yelling "agent!" into their phones when speaking with a company's voicebot can likely understand the frustrations associated with an unhelpful chatbot.

Historical Issues

The primary challenge leading to this level of dissatisfaction is a chatbot's inability to truly understand what is being asked. The degree to which a bot can understand the intent of a user's question, as opposed to strictly relying on the "correct" combination of terms, is what determines whether the chatbot will be able to successfully resolve the user's issue.

Meanwhile, most stakeholders do not have the time, training, or inclination to develop an intelligent chatbot to address this problem. Their hope is that once deployed, the bot will have the ability to absorb the bulk of routine calls or emails to free up time for more meaningful aspects of their jobs. Unfortunately, this is seldom the case when the bot's responses are templated, requiring a great deal of time to manage the quality of its responses to real-world scenarios.

In higher education, IT help desks, along with admissions and financial aid offices, were among the earliest adopters of chatbot technology. Staff in these areas recognized the potential for chatbots to resolve a significant volume of questions before those questions became support tickets. However, they also learned that to work effectively, bots require ongoing development. Essentially, the problem comes down to resource constraints around building a sophisticated bot.^Footnote1

A Response to the Problem

A truly artificially intelligent chatbot that meets the standard of prestige required by higher education institutions needs to possess several characteristics.

Customizable

When an institution purchases a chatbot, the bot's knowledge must be customized to the institution and be distinctive from the knowledge of any other bot. The information it provides to users must be specific and complete with as few clicks as possible. The bot's knowledge should be curated straight from a variety of sources developed and managed by each institution.

Multi-Channel

Students are accustomed to multi-channel access, so chatbots must be accessible on a variety of channels. Students should be able to take the conversation beyond websites to SMS, social media, email, and voice assistants such as Amazon's Alexa and Apple's Siri.

Multilingual and Resistant to Implicit Bias

Chatbots must be inclusive and able to support diverse student populations. A chatbot that speaks more than one hundred languages can broaden the range of students it can help. In addition, a chatbot based on a range of user interactions can support a wide variety of users and eliminate implicit bias.

Secure and Scalable

Given the structure of higher education institutions, chatbots must be deployable in multiple departments, each with its own distinct knowledge and objectives. Administrators need a bot that does not require extensive development in order to provide high-quality, customized interactions in each area. A bot with technology that builds itself will put AI within reach across the institution and deliver anonymous, aggregate data that administrators can use to keep pushing the needle forward.

Self-Learning and Increasingly Intuitive

Finally, a chatbot must not be easily stumped or require an inordinate degree of effort to stay up-to-date. To that end, chatbots should not be exclusively rules-based or rely on keywords to understand meaning. Instead, a bot must learn from each conversation it has, perpetually improving its ability to answer questions at an increasing level of complexity and, ideally, to update its own knowledge when possible. As is the mission with any AI instrument, the chatbot should mimic a human resource and apply knowledge from its gained experience.

Today's Outcomes

Chatbots that incorporate all of these elements at once can have a significant impact on higher education. The bot implemented at Broward College, an institution with a student population of more than 65,000, reduced call volume by 9.6%—a total of 30,041 calls. Broward leaders estimate that this led to a cost savings of $210,287. Their bot also deflected live chat conversations by handling 60,842 inquiries, leading to an additional cost savings of $304,210. The University of Portsmouth reduced live chats by 50%, while Temple University reduced calls by the same margin. Meanwhile, with the help of "SoonerBot," the University of Oklahoma enrolled its largest class of first-year students in Fall 2019.^Footnote2

These achievements were possible only because their chatbots kept learning, increasing the scope of support they could provide and delivering complete and accurate answers that fully satisfied the needs of students.

Next Steps

Those in higher education who have been seeking a chatbot solution to produce real, tangible results for their institution should take the following steps:

1. Get others at your institution onboard when beginning the process, to ensure that goals are aligned throughout deployment.
2. Evaluate the internal resources you want to commit to the ongoing success of the project (e.g., SMEs for each department, staff for live chat escalations).
3. Seek out a chatbot that fits your needs as well as those of your end users (e.g., breaking down silos, providing direct answers, rating chats).

As AI advances in the coming years, chatbots will get increasingly smarter, more intuitive, and as a result, more valuable for higher education.

Notes

1. For help selecting among the various types of chatbots on the market today, see How to Select a Chatbot in 2022, [https://go.ivy.ai/chatbots2022] Ivy.ai (website), accessed April 25, 2022. Jump back to footnote 1 in the text.↩
2. See "Case Studies," Ivy.ai (website), accessed April 25, 2022; "OU Uses Artificial Intelligence in Recruitment," Inside OU, September 18, 2019. Jump back to footnote 2 in the text.↩

Mary Frances Coryell is Chief Revenue Officer at Ivy.ai.