Where educators dream, AI builds, and together they teach.

Five teaching tools built so far — more to come.

Role-Play AI Coach

When students break into groups for a role-play, you wish you could be in every room — watching how each one performs and giving them personal feedback after. With several groups running at once, you can’t. So an AI sits in with each group: it follows the whole role-play and gives every team an individual written critique, as if a coach had been in the room with them.

How it works

  1. Record & transcribe. The group records its live role-play on a laptop or phone — ordinary room audio, nothing to install — and it is transcribed with each speaker identified by role.
  2. Assess against the literature. The transcript is read against the principles in the subject’s standard texts — the benchmark for what good and poor practice look like.
  3. Score objectively. Where the exercise has a measurable outcome, the result is computed in code — measured, not estimated by the AI.
  4. Compare with the class. Each group is benchmarked, anonymously, against every other group that ran the same exercise.
  5. Report. Everyone gets a plain-language report: what happened, what was missed, the turning points, and one concrete change to make next time.

Weekly Recall Game

Keeping a class on top of the material week by week usually means writing quizzes you don’t have time for. Here you don’t: set the course up once and a short game runs itself every week — questions built from your own slides and readings, students kept honest, and feedback that points each one to exactly what to revisit.

How it works

  1. Set up once, at the start. You load the course’s slides and readings up front; from these the game builds every week’s questions and answers automatically, each mapped back to the exact slide or reading it came from.
  2. A short round each week. A quick set opens from that week’s material — multiple-choice and short answer, one attempt, pausable until it closes. Minutes, not an evening.
  3. Marked instantly and fairly. Auto-graded, with a randomised set per student, so it’s their own understanding being tested.
  4. See where you stand. A weekly and term-long class leaderboard — low-stakes each week, but it counts toward the course grade, so it pays to keep up rather than cram.
  5. Targeted review after it closes. When the window ends, students get the answers, the reasoning, the exact slide or reading each question came from, short expert clips where provided, and how the class did.

Group Presentations at Scale

A group presentation is an assessment wrapped in a scheduling problem. With 150 students that is 30-odd groups to form, several topics to keep apart, 30-odd slots to find, and every group marked the same way — usually held together by a spreadsheet, an email chain and a video link half the class loses. Here it is one platform, online, start to finish. The students do the organising; you turn up and watch.

How it works

  1. Groups and topics sort themselves out. One student starts a group, shares the code, the rest join — sizes capped, stragglers pooled. Each group then claims its topic first come, first served. No allocation spreadsheet, ever.
  2. Students book their own slot. You open the times — parallel rooms, evenings, extra sessions — and groups take one each, confirmed by email. Every slot carries a named tutor. You never build a timetable.
  3. They present in the browser. Click Join and you are in — video, screen share, no accounts, no downloads, nothing to install. One switch moves every session to Zoom or Teams instead.
  4. You mark it live, then moderate. You and each tutor score your own groups from a phone, deck to hand. Nothing is released until you have second-marked blind and compared tutors — then it all lands in the gradebook at once.
  5. And the quiet one: who actually did the work. Each student rates their own group, confidentially — you see what each person received, not who said it. It is there for the group in ten where somebody coasted.

Concept Labs

Every subject has an idea or two a slide can’t teach — the kind a student only understands once they build it and watch it move. Showing how a company’s ownership dilutes across funding rounds, or how a discount rate turns cash flows into a value and a return, dies on a static diagram. So instead we hand the student the model itself: an interactive simulation they drive, with the mechanics laid bare, so they reach the exam already fluent in what used to lose them.

How it works

  1. Take the concept that won’t stick. The ones where a single number ripples through everything else — worth playing with, not just re-reading.
  2. The student drives the model, not a slide. They move the inputs and the whole picture redraws live — no setup, nothing to install, all in the browser.
  3. Nothing is hidden. Every figure is explained, with a “how the maths works” panel and a plain-language glossary one click away.
  4. Learn by consequence. Push a term to its extreme and see who gains and who pays — the intuition sticks because the student caused it.
  5. Ready for the graded work. A worked example means no blank page, so they walk into the case or exam confident with the moving parts.

On Air

There are always more readings than a student will get through — and the ones who fall behind rarely climb back in through a dense PDF. So each module’s readings become an AI-generated audio conversation: two hosts walking through the key ideas in plain language, the kind students actually put on for the commute. It doesn’t replace the reading — it pulls them back to it — and it’s earned, not simply handed over: the recap unlocks once the student has put in the week’s work.

How it works

  1. Built from the readings themselves. Each module’s required readings become one AI-generated audio conversation — two hosts, in plain language, walking through the key arguments.
  2. Earned, not handed out. The recap stays locked until the student has put in that module’s work — so the reading and the thinking come first.
  3. Made for dead time. A podcast they play on the walk in or at the gym — at 1×, 1.5× or 2× — turning time they weren’t going to read anyway into revision.
  4. A companion, not a shortcut. It refreshes the ideas and ties them together; the exam and the graded work still test the readings themselves.
  5. One per module, as the term unfolds. A fresh episode lands with each module, right where the student already tracks the week’s work.
In progress · coming soon

Practice Studio

The student generates their own practice questions from the course slides and readings — choose the week, question type, difficulty and length, answer in their own time, and get instant AI feedback. Private, with an optional “share with my lecturer”.

Building now — full detail when it ships.

FINS3623

For enrolled UNSW students.