The inaugural Cogniti Connect brought together more than 150 educators from across Australia, New Zealand, and further afield. The purpose was not simply to pulse check what others are using Cogniti for, but to connect: to see who is working in this space and to hear directly from colleagues about what they have built.
A great tool is good, but a great tool plus a community is even better, and the aim of Cogniti Connect sessions is to help educators learn from one another, find inspiration and potential collaborators.
Three contributions anchored the hour: two case studies and a live demonstration.
Three virtual teammates, one clear favourite
Nuddy Pillay from the University of Auckland began with a challenge familiar to most people who have experienced a team assessment: teamwork is difficult. It can be challenging to scaffold effectively, particularly when students are learning across different campuses and modes of study. Nuddy wanted students in his foundation business course to have equitable access to meaningful support as they navigated the academic and interpersonal demands of teamwork.
Working with Cogniti, Nuddy involved students in identifying the forms of support they needed most. Their feedback informed the co-creation of three AI-supported virtual team members, each designed to address a different dimension of effective teamwork.
Terri, the task-focused AI teammate, helped students organise their work, clarify responsibilities, and maintain progress. Val, the value-adding AI teammate, encouraged students to contribute meaningfully and strengthen the quality of the team’s work. Ray, the relationship-focused AI teammate, supported students with communication, conflict, participation, and the interpersonal challenges that often arise in group assignments. Care to guess which agent got used the most?
Of the three agents, Ray emerged as the clear favourite and was used far more frequently than Terri or Val. This finding highlighted the importance of the human dimension of teamwork. Students appeared to value accessible, just-in-time guidance for managing difficult conversations, unequal participation, and team relationships.
Ray was designed not simply to provide students with an answer. The agent offered possible approaches and then used reflective questions, such as asking students what they might do next, to help them consider the situation, evaluate their options, and take responsibility for their response.
Accessibility was also central to the design. The agents were embedded directly within Canvas, allowing students to access Terri, Val, or Ray from their course site without moving to a separate platform. This made the support visible, convenient, and closely connected to the team assessment.
Nuddy has submitted this work to ASCILITE later in 2026 and welcomes opportunities to collaborate with others interested in AI-supported teamwork, inclusive learning design, and student-centred educational innovation. Feel free to contact him at nuddy.pillay@auckland.ac.nz if you’d like a copy of his prompt(s).
Rethinking maths support
Imagine senior leadership asks you to trial an AI-first approach to numeracy support, and you’re the one who has to make it work. You might be sceptical. Luis Camacho and Chris Bridge from Federation University understand that feeling, but they also know what it looks like on the other side, once you’ve seen AI numeracy support work, and work well.
Luis was responsible for implementing maths support across four campuses, serving a cohort with a high proportion of first-in-family, regional, remote, and online students. His Cogniti agents, NiNA (for nursing numeracy) and EdNA (for the numeracy component of LANTITE, the national Literacy and Numeracy Test for Initial Teacher Education students), were built around transition pedagogy and self-regulated learning, guided by a simple design principle: just in time, just for me, and just enough.
The data told its own story. Students returned for scaffolded, follow-up help rather than one-off answer-seeking. Most support was sought after hours, with clear surges around assessment periods, evidence that students were using the agents to build understanding, not shortcut this.
Good ideas rarely start from nothing. Luis adapted and built on prompts colleagues had already developed, exactly the kind of reuse and iteration that a community like Cogniti Connect is designed to promote.
The work has been published in the Student Success Journal (STARS). See the article here: https://studentsuccessjournal.org/index.php/studentsuccess/article/view/4019
Luis also generously shared the agent prompts.
Mini-apps: what used to be out of reach now takes 5 minutes
To close, Danny Liu from the University of Sydney demonstrated Cogniti’s mini-apps, interactive learning activities that educators create simply by talking to an AI builder (vibe coding for educators).
Mini apps are particularly useful for multimodal activities, visualisations or providing branching scenarios.
The examples Danny shared included Japanese listening practice, an interactive roof rafter activity for carpentry students, a business simulation in which students make choices and experience changing market conditions, a medical-terminology quiz, and a Year 12 maths tool pairing a simulation with an embedded AI tutor.
Want to see the demo? Watch this 9-minute video.
A call for contributions
The community meeting was a success. Emails were exchanged in the chat, collaborations began to form in real time, and presenters offered to share prompts.
The vision for Cogniti Connect is to keep sharing, and to keep building on one another’s ideas so the whole community can achieve more.
Cogniti Connect will run monthly. The organisers are actively seeking presenters for future sessions and warmly invite educators to bring a colleague, an idea, or a problem worth solving together. Please reach out to olga.kozar@cogniti.ai if you would like to share. And if there’s someone you think has a great story to tell, feel free to nominate a colleague too!