Education leaders are always searching for innovative ways to enhance their students’ learning experiences and prepare them for the challenges and opportunities of the future. With the rapid advancement of generative AI, a new set of possibilities have emerged for instructional institutions.
The University of Sydney (UofS) recognized the potential of generative AI to enhance teaching practices and assessments, enrich the learning experience, and prepare students for an evolving job market. However, they quickly realized that an out-of-the-box solution wouldn’t suffice, given their unique requirements for security, customization, and contextualization. UofS has demonstrated a successful path forward using Microsoft’s Azure OpenAI Service to build a custom AI solution to meet their needs. By showcasing responsible and effective integration of generative AI, they provide a model for other institutions. Their experience offers valuable insights to tailor AI tools to your unique needs and goals for successful implementation.
Read the full UofS customer story to learn more about their exciting custom AI solutions.
Building a custom AI assistant
Enter Cogniti, a self-serve web app authenticated through Microsoft Entra ID, developed by Danny Liu, Professor of Educational Technologies at UofS. Cogniti is built on the secure and private Azure OpenAI Service platform and empowers faculty to create their own AI agents that they can customize and contextualize to their unique needs, helping students learn in pedagogically meaningful ways. “Faculty love the fact that they can control and therefore trust the AI and that it’s so easy to build their own agents using plain language and embed it into our learning management system for students to use,” said Liu.
With the help of Microsoft’s GitHub Copilot, Professor Liu handled the bulk of Cogniti’s development, from coding to faculty integration. “It was amazing to have Microsoft Copilot working alongside me, troubleshooting, and debugging,” he says. “I wouldn’t have been able to build Cogniti in the time that I did if it weren’t for Copilot.”
By leveraging Azure’s robust infrastructure and GitHub Copilot, Professor Liu developed Cogniti efficiently while maintaining alignment with the university’s objectives. Cogniti addressed privacy concerns by ensuring the confidentiality of prompts and responses, which are not used for further training by OpenAI. This powerful combination provides a secure environment for sensitive educational data and accelerates the development process.
Practical pedagogical applications
Cogniti has been successfully integrated into various pedagogical scenarios at UofS, enhancing student learning and faculty support.
- Bridging theory and practice: Occupational therapy students gained practical experience by role-playing with “Mrs. S,” an AI-simulated client. This allowed them to apply classroom knowledge to a realistic scenario, effectively bridging the gap between theory and practice.
- Consistent student feedback: A coordinator managing large first-year classes with 1,500 students used a Cogniti AI agent to improve grading consistency, quality, and efficiency. The agent incorporated rubrics, assignment expectations, and feedback examples, resulting in faster marking and more consistent feedback for students.
- On-demand biochemistry support: UofS teaches biochemistry to over 800 students who have thousands of questions each semester. With Cogniti, these students can engage in dialogue anytime, discussing course material, explaining their understanding, and receiving valuable feedback from the AI tutor.
- Empowering tutors in complex conversations: Following legislative changes in Australia, the University of Sydney implemented a new student support policy requiring tutors and faculty to advise students on academic issues. These meetings often lead to discussions about well-being, health, or finances. To help tutors navigate unexpected topics, the university fed scenario-specific prompts to Cogniti, offering instant guidance for questions beyond the scope of academic issues.
The benefits of Cogniti are far-reaching. Educators can now allocate more time to deeper engagement and personalized feedback, while students enjoy enhanced learning experiences through interactive AI assistance. Cogniti also offers students valuable opportunities to develop familiarity with AI within a safe, controlled environment, which will be increasingly crucial in their future careers.
The university’s commitment to equitable access to AI is clear in its decision to offer Cogniti for free to all students and staff. As Joanne Wright, Deputy Vice Chancellor of Education, states, “One of the values of Cogniti is that it can be used in a number of different learning settings. Faculty can even personalize it to meet their students’ preferences, weaknesses, or strengths.” Looking ahead, the university hopes to expand Cogniti’s use to more institutions worldwide, believing that its controllability, stability, safety, and accuracy can help dispel fears surrounding generative AI in education. Learn more about the University of Sydney’s inspiring journey with custom generative AI.
Build your own custom AI solution
Kickstart your school’s AI journey
Explore the Microsoft Education AI ToolkitIf you’re interested in building your own custom generative AI solution for your institution, in a similar way to what the UofS accomplished with Cogniti, consider these first steps for getting started.
- Purchase the necessary IT infrastructure, licenses, and software to create a custom AI application using the Azure OpenAI Service.
- Review the Overview of Responsible AI practices for Azure OpenAI models and Azure OpenAI’s Transparency Notes to ensure responsible use of the service and understand any system limitations.
- Assemble a group of school and instructional leaders who are enthusiastic about piloting your AI program to help design, test, and provide feedback.
- As you begin to develop and document your implementation plan, consider addressing the concerns of all stakeholders, including community members. For guidance, consult the “Engaging your community” section within the Microsoft Education AI Toolkit.
- Before you begin building, familiarize yourself with step-by-step technical overviews in the implementation roadmap in the “AI Navigators” section of the Microsoft Education AI Toolkit to understand how educational institutions like the Department for Education in South Australia successfully implemented their AI programs.
The UofS is also keen to partner with institutions who may want to adapt Cogniti in their own contexts. This is an inspiring example of how educational institutions can safely and responsibly build and integrate generative AI tools into teaching and learning practices. By embracing this transformative technology, you can also improve the quality of students’ education and equip them with the skills they need to thrive in an increasingly digital world.
Explore more about Azure OpenAI services and how to build custom chatbots with these resources:
