Tanglin Trust School · CPD 2026 Hard Thinking with GenAI

Session 02 · Infant School

AI in the learningenvironment.

Part one: how children encounter AI, and first-hand experience before AI-mediated experience. Part two: a four-station jigsaw on practical AI for educators.

  • Duration 90 minutes
  • Setup Department tables · devices
  • Focus Practical AI for educators

Part one — opening

Three principles before the tools.

  • First-hand experience first. Children draw their own pictures before AI pictures, see real animals before AI animals. The doing comes first, then the comparison.
  • The environment does the teaching. Continuous provision makes the right choice the easiest one. AI-generated material in the room should provoke questioning, not provide answers.
  • Pattern-matching at scale. AI does not understand, reason, or know. It predicts plausible next words and pixels.

GTT D4 hooks: 4.1 Structuring (sequencing prerequisite experience before AI), 4.3 Questioning (AI-generated images as prompts for observation and comparison).

Part two — the jigsaw

Four stations, four use cases.

01

Deep Research — curriculum planning and scope and sequence

Using Gemini Deep Research (or Copilot Deep Research) for the heavy synthesis you used to do across a dozen open tabs. Useful for researching curriculum frameworks, comparing approaches across jurisdictions, and building scope-and-sequence documentation.

  • Frame the question precisely — Deep Research rewards specificity.
  • Audit the citations. The output is a starting point, not an answer.
  • Treat it as a first-draft synthesis to argue with, not adopt.
02

Voice transcription — capturing faculty conversations

Turn spoken planning conversations into structured documentation. Useful for PLT meetings, planning discussions, and the post-it-note conversations that never make it into a document.

  • Always tell participants what is being recorded and why.
  • Voice in, structured prompt out — ask it to format minutes, not just transcribe.
  • Pair with NotebookLM for queryable team memory across the year.
Sample PLT meeting transcript Use if you don’t have your own · .txt
03

NotebookLM — managing professional learning

Upload your sources — research papers, policies, planning documents, meeting notes — and query them as a single set. Particularly strong for PLCs, action research, and PLT cycles where the same material needs to be revisited many times.

  • NotebookLM only references the documents you upload — citations link straight back to the source.
  • Useful for keeping research in conversation with your context, not in a separate folder.
  • Audio overviews are surprisingly good for a 20-minute commute summary.
04

Gems / Agents — unit planning and admin

A Gem (Gemini) or Agent (Copilot) is a chat with persistent system instructions and uploaded knowledge. Build one for your phase or department, then use it to draft unit plans, admin documents, and templates that already know your context.

  • System instructions = how it behaves. Knowledge files = what it knows.
  • Start narrow (one Gem or Agent, one job). Resist the temptation to build a do-everything assistant.
  • Test with a real task before sharing with colleagues — Gems and Agents compound, so do their flaws.

Part three — debrief

Reflection questions.

  1. Which station felt most immediately useful, and what would your first 20-minute experiment look like?
  2. Where in your week does AI save you time without compromising your or your children's thinking?
  3. Where might AI-generated content already be entering your children's world, and how can you prepare them to be critical observers?

One thing to try this term: use an AI-generated image as a provocation in your continuous provision area, and observe what children notice. Spotting what is wrong is the hard thinking.

Session 03

Junior School →

Session 04

Senior School →