How the University of Johannesburg uses AI chatbots to increase student engagement and reduce costs

This is a guest post from the University of Johannesburg, an AWS customer.

The transition from high school to university is a stressful journey for every student and one that is made even harder for applicants in South Africa, with nearly 695,000 applications for less than 11,000 seats.

At the University of Johannesburg (UJ), we’re proud to be a leading tertiary education institution on the African continent and a player on the global stage. It’s exactly this prominence that brings us over 100,000 new student applications each year.

This influx of applications is directed toward the university’s Central Academic Admissions (CAA) department. The CAA team are tasked with assisting students, often aided by their parents or guardians, through the application and registration journey.

This annual peak comes at a cost to the university as we need to expand our temporary staffing to handle the spike in queries by email, telephone calls, and in-person. The extra cost amounts to R800,000 for the month and a half and is an outlay to handle the influx of queries both from students and legal guardians.

Leveraging technology to reduce costs

At the start of 2019, the university embarked on a journey to find ways to handle this situation in a more efficient manner. Specifically, we were looking for ways to harness the power of AI as a means of walking the talk as we had just started incorporating elements of the 4^th Industrial Revolution (4IR) into our teaching curriculum. Leveraging the drive of 4IR, the task was to dovetail to this theme and look at technology to assist the university with solving these challenges.

Of the many benefits that the university was looking for, the following were front and center:

• The solution must help alleviate the volumes on our staff.
• The solution must ensure students are provided with access to basic information without undue delays.
• The solution must present a low barrier to entry for our students (new and existing) and their parents or guardians.

The AI solution must also provide accurate information, as a mistake in communication could cause considerable inconvenience to a student, their family, and their learning journey.

After an extensive selection process, the university finally selected CX/SIM late in 2019 to be our solutions provider for this industry-leading AI-powered service automation chatbot. CX/SIM, an Amazon Web Services (AWS) Partner, then set out to develop a bespoke, cloud-native solution that put the Amazon Lex chatbot at its core.

Using AI-powered chatbots to increase student engagement

UJ introduced MoUJi, a digital assistant (or chatbot) in 2020. MoUJi is designed to help students, prospective students, alumni, and guests with a wide range of inquiries.

MoUJi utilizes AI and natural language processing to simulate human conversation, providing instant answers to questions related to applications, registrations, financial statements, and more. The chatbot is integrated with UJ’s student administration systems, allowing users to access and update personal information, check admission statuses, view class timetables, and even reset their login PINs.

MoUJi offers 24/7 support, which enhances the university’s engagement with its community and streamlines various processes for students.

The MoUJi solution focus on three primary components:

• An interface through which the student can interact with the chatbot or live agent.
• An interface through which the chat agents can interact with students and review chat transcripts.
• A central hub to direct the conversations and manage the flow of information.

The CAA team played a pivotal role in preparing and training MoUJi. Leaning on their years of experience, the team was able to quickly pinpoint the most common and critical student queries—over 100 in total—and ensure that MoUJi was able to accurately detect these queries and respond correctly.

Increasing ROI by expanding the reach of MoUJi

With the common questions baked in, MoUJi’s scope was quickly broadened from being a simple FAQ chatbot to an integrated one. MoUJi is able to securely integrate into the university’s student information systems to retrieve and display specific pieces of a student’s personal profile such as their exam timetable, exam results, certificate reprints, and financial information.

How does a student interact with MoUJi? The landscape in 2019 was different than what it is today. Back then, few sites offered chat as an option and even less offered assistance through a chatbot. By deploying a clean and simple chat widget to our website, the university provided students with a fast and easy way to access MoUJi.

Additionally, Facebook Messenger was integrated shortly after the web launch, giving MoUJi the ability to serve visitors to the university’s Facebook page. As of August 2023, students can access MoUJi on WhatsApp from the comfort of their own phone. Using WhatsApp has broadened the reach of the university because it enables marketing to send links to students and allows for outbound notifications.

Another essential feature that aided in the success of MoUJi is how it uses channel-specific capabilities to deliver an easy and intuitive visual experience for the student, reducing the cognitive overhead and making it quick and easy to locate the necessary information.

Using a “build once, deploy many” approach, the university has been able to achieve greater ROI since the single MoUJi chatbot works across a variety of channels. Leveraging the power of the AWS Cloud, the CX/SIM centralized processing hub meets the peaks, ranging from 500 chats per 30 minutes up to 1,000 when applications are opened. This hub is responsible for directing the incoming chat requests from the various channels. During orchestration, these requests undergo numerous pre-processing steps before being delivered to the Amazon Lex chatbot.

Once a successful Amazon Lex response is received, the response is then further enriched by the processing hub before delivery to the student. Under specific circumstances, MoUJi’s responses are tailored to the student through the injection of student-specific information.

Failures to correctly detect student intent is an expected outcome—one that has been catered for in the design. Using a series of techniques, inputs that result in failures are flagged for review to aid in improving MoUJi’s understanding or expanding its field of knowledge when necessary.

All of these processing operations occur completely within the AWS environment. CX/SIM’s bespoke application framework leverages their depth of development expertise paired with AWS’s wide range of managed services. This combination ensures that the MoUJi platform respects the tenants of high availability, scalability, resilience, reliability, and performance. The security of the platform and its data underpins everything.

By using AWS managed services, CX/SIM are able to provide industry-leading solutions that are extremely cost-effective.

With MoUJi at the front, the university has dealt with increasing volumes of queries without needing to increase the temporary staff complement, and without compromising the student’s ability to get access to information. The reduction of agents, from 120 to 60, has shown that the technology, when deployed correctly, is effective in reducing costs amounting to over half a million rands and bringing in efficiencies for the peak period.

MoUJi’s ability to handle queries and ensure a high First Call Resolution (FCR) means that skilled CAA representatives have more time to deal with complex queries and scenarios where more human empathy is necessary.

MoUJi has been a game changer for the University of Johannesburg and an instrumental piece in our strategy of using AI technologies to continually improve the service the university offers to its stakeholders.

Looking further at how generative AI can improve student engagement

With the emergence of generative AI, the next evolution of MoUJi is surely set to wow the students with conversational AI. Using technologies easily and readily available in the digital age, the university is already looking at incorporating the use of large language models (LLMs) to augment an already successful use case.

The use of generative AI and integration to an LLM allows the university to better understand students, reduce the effort of training, and also allows us to manage integration using functions and custom code nodes to the LLM and the business systems. Amazon Bedrock enables the safe, confidential, and compliant deployment of LLMs and generative AI engines to support the compliance and security aspects of the university’s data.

Deploying generative AI responsibly requires a strong partner. Over and above the servicing of students of the university, CX/SIM are constantly looking to deploy solutions using generative AI on back-end processes, allowing for retrieval of transcripts, summaries and sentiment analysis, and also suggest next best actions for business processes and student chats.

Conclusion

In this post, we discussed how—with the increasing number of student applications and queries leading to increased costs—UJ worked with AWS Partner CX/SIM and AWS to develop an AI-powered chatbot. This chatbot increased student engagement across multiple channels, while decreasing costs for the university.

Colleges, universities, and research institutions around the world use AWS to support research and learning, improve the student experience, and make data-driven decisions that save money and resources. Learn more by visiting the AWS Cloud for Higher Education hub.

Also, discover how AWS has worked with other South African institutions, such as the University of Limpopo, Tshwane University of Technology, the University of the Witwatersrand, and the University of Pretoria.