Working document · May 2026

AI Pilot Programme.

Discrete pilots, each a working artefact. Pick the ones that point toward the partnership you actually want.

For Liz Gillies, Rohan Martyres
From Chris Marmo, Paper Giant
Ahead of Wednesday session

§ 01 — What we heard

Our understanding.

You're asking AI to be a participant, not just a tool — a synthesiser between MLF, SVA and the Ethics Centre, an active contributor in your innovation containers, a different kind of engagement with young people. You're already piloting one space with ANU CLL, and you want to extend the experiment.

You've named four threads: AI agents in innovation containers, a public-facing engagement platform, content drafting from your SharePoint, an animated video pipeline. The window is around five weeks. The public-facing deadline sits in term 3. The budget is flexible.

What follows is a menu of discrete pilots that respond to what you've described. Each is a working artefact running on Paper Giant infrastructure. Each points toward a different long-term shape for the partnership. Build the package that's right for you below.

§ 02 — The pilots

Four discrete
pilots.

Each runs on Paper Giant infrastructure. Each delivers a working artefact, light user testing where relevant, and handover documentation. Tap a card to add it to your package.

01
Methodology partner

Innovation Container Agents

Investment$25,000

What does AI-as-active-participant in a social innovation container actually feel like? What do specialised agents — each holding a specific perspective — add that human-only containers can't?

Full scope

What we'll do

  • Design the agent framework — hats, parameters, role alongside humans
  • Build a working multi-agent system (Claude API, possibly one alternative model for voice variety) with a facilitation interface
  • Internal testing with synthetic scenarios
  • Run it live in one of your actual innovation containers as a participant
  • Synthesis output — what the agents added, what surprised participants
  • Handover documentation

Delivered

A working multi-agent system, a recorded session showing it in use with real human participants, and a methodological note on what worked and what to refine.

What it doesn't do

Doesn't address the public-facing platform or animated video. Doesn't deliver content drafts (though could in a future iteration).

Best for

Aligned with the methodology-and-learning direction. Strongest natural relationship to ANU CLL's parallel pilot — clear differentiator if theirs is research-focused and ours is methodological-practical.

02
Production partner

Conversational Module Prototype

Investment$25,000

What does an AI-native conversational learning module actually feel like for a young person? Is it credible at the pedagogical bar we need for term 3?

Full scope

What we'll do

  • Module design — one topic (e.g. polarisation, civility-as-practice), journey arc, conversation scaffolding
  • Build the conversational interface (Next.js + Claude API on our stack)
  • Author underlying content + young-person contribution capture
  • Light user testing with a small group (school proxy or directly if available)
  • Polish + public URL hosting
  • Documentation

Delivered

One fully working conversational module at a public URL that a young person can actually use, plus user testing notes.

What it doesn't do

Doesn't deliver the full LearnWorlds off-ramp or a complete platform. Doesn't include animated video (see Pilot 04). One module is small evidence — you may want more before committing to a scaled commission.

Best for

Aligned with the public-facing delivery direction. Delivers a visible artefact you can show stakeholders by term 3.

03
Tool / utility partner

Partner Corpus Synthesis

Investment$25,000

What does a working synthesis tool feel like when it draws on MLF, SVA and the Ethics Centre's combined material? Can the team use it to surface tensions, generate content drafts, find shared ground?

Full scope

What we'll do

  • Initial corpus assembly — manual export from SharePoint plus partner uploads (bypasses MS365 integration for the pilot)
  • Stand up the synthesis interface — chat-with-corpus, voice input, contradiction detection
  • Add content-drafting workflows — "draft a module script from these materials," "compare how SVA and Ethics Centre frame agency"
  • Testing with each partner, iterating
  • Recorded session showing the tool in use
  • Documentation, including a clear path for future SharePoint integration

Delivered

A working synthesis tool on our infrastructure, accessible to partners via web, with documented session outputs and a path for future SharePoint integration.

What it doesn't do

Doesn't integrate with SharePoint in this pilot — corpus is loaded manually. Partners need to actually use the tool to give signal. If engagement is light, the evidence is thin.

Best for

Aligned with the "mycelial connector" framing from our conversation. Foundational — if this works, it's the substrate other pilots can build on.

04
Video production partner

Animated Explainer Pipeline

Investment$25,000

Can we produce credible animated educational content from your existing scripts using current AI video tooling, and at what quality bar?

Full scope

What we'll do

  • Script adaptation and storyboarding for AI video production
  • Production workflow setup — Veo 3.1, Runway and voice synthesis stack assembled into a repeatable pipeline
  • Production of 2–3 short animated explainers from supplied scripts
  • Review and iteration cycle with you
  • Documentation, including production workflow handover

Delivered

2–3 short animated explainers produced from your existing scripts, plus a documented production workflow you can use to commission more.

What it doesn't do

Doesn't deliver a self-service tool — production still requires a producer. Doesn't address the public-facing platform around the videos.

Best for

The "convert text and scripts to animated videos" ask from the brief, sized realistically.

§ 03 — Build a package

Pick what's useful.

Tap the pilots that interest you, or start from a preset. The numbers update as you go. There's nothing committal here — it's a tool for thinking.

Start from a preset

Or build your own

§ 04 — Sequencing

Across
the weeks.

The recommended order: Pilot 03 first (foundation) → Pilot 02 (uses the corpus) → Pilot 01 or 04. Light overlap between consecutive pilots. Updates based on your selection above.

Select pilots above to see the sequencing.

§ 05 — Out of scope

What this
doesn't address.

  • MS365 / SharePoint integration

    Deferred on purpose. Each pilot runs on our infrastructure. Once a pilot proves out, integrating with your stack becomes a discrete Phase 2 conversation with clearer scope.

  • WordPress replacement

    Not in scope for these pilots. Could be addressed in a future engagement if useful.

  • Long-term hosting / SaaS commitment

    Pilots are pilots — they sit on our infrastructure as proofs of concept. Production hosting is a separate conversation that becomes much easier to scope once we know which pilot has the strongest signal.

  • ANU CLL coordination

    Worth a separate conversation about what's transferring between their parallel pilot and ours, and whether there's a useful interface between the two.