AI Workshop with AI Sandbox

November 10, 2025
Workshops

🧠 Rātā Workshop #4 – AI Education and Empowerment

Supported by the Rātā Foundation
Delivered by Nelson AI Sandbox

A foundational session introducing non-profit and community organisations to Artificial Intelligence (AI) — focusing on understanding the technology, using it safely, and planning for responsible adoption in Aotearoa New Zealand.

👥 Speakers

Sloane Bayley – Programs Manager, Nelson AI Sandbox
AI enthusiast, with a focus on supporting the non-profit sector.

J Norness – Fractional Chief Marketing & AI Officer
Director and advisor helping organisations use AI strategically and safely.

🎯 Workshop Overview

The session covered:

Introduction to Generative AI
Privacy, Safety, and Security
Practical Tool Demonstrations
Sustainability and AI Impact
Example Case: SOLE (a fictional NFP)
Creating and Implementing an AI Policy
Understanding AI Agents and Tools
Planning for Organisational AI Use

🌏 AI Context in Aotearoa

New Zealand currently lacks strong AI regulation, creating risks around data scraping and privacy (e.g., Meta using NZ/AU user data for AI training).
The non-profit sector is adopting AI faster but often without governance frameworks in place.

Key Stats (Infoxchange AU & NZ, 2024):

76% of organisations use generative AI tools (up 52% in a year).
Only 33% are actively investing across the organisation.
20% hesitate due to lack of understanding; 13% are concerned about privacy or data sovereignty.
89% have no AI policy in place.
In Te Tauihu, 75% of NFPs have no AI policy.

🧩 Understanding Generative AI

Artificial Intelligence = machines performing human-like tasks.
Machine Learning (ML) = systems that learn from data.
Deep Learning = advanced ML that identifies complex patterns.
Generative AI = creates new content (text, images, video, music, code).
Large Language Models (LLMs) like ChatGPT are trained to predict language patterns — they don’t “know”, they predict.

💡 Four Ways to Think About Generative AI

Superpower: Extend and enhance human ability.
Many Interns: Multiple assistants for different tasks.
Thought Partner: Helps test, clarify, and refine ideas — not replace thinking.
Autonomous Interns: AI agents that can plan, decide, and act toward a goal.

🔐 Privacy, Safety & Security

AI tools come with risks:

Bias (how models were trained)
Hallucination (incorrect or made-up info)
Privacy leaks (data exposure or misuse)
Cyber risks (e.g., deepfakes, malicious code injection)

Frameworks introduced:

Lethal Trifecta: Risk arises when systems have (1) untrusted content, (2) access to valuable data, and (3) external communications. Remove one to reduce risk.
Three Safety Tiers:
1. Open (low risk) – marketing, public info
2. Controlled (medium) – internal reports
3. Locked (high) – payroll, IP, HR data

Golden rule:
➡️ Human > AI > Human — every output should be checked by a person.

🌿 AI Sustainability

AI requires large energy and water use (Amazon used 105 billion gallons in 2021).
Challenges: computing power, resource extraction, and rebound effects.
Solutions:

Measure and manage energy impact.
Use smaller, specialised models.
Support clean energy and efficient infrastructure.

🧦 Case Study: SOLE (Society of Left Enthusiasts – Nelson & Tasman)

A fictional non-profit created to demonstrate AI in practice.
SOLE’s mission: “Making Lefts Right” – investigating the disappearance of left socks.
Used to show:

How to generate branding, jingles, and visuals.
How to build custom GPTs to handle policies, minutes, and agendas.
How NFPs can safely test AI tools.

💬 Prompting Basics

Prompt = what you tell AI to do.
Clearer prompts → better results.

CARE Framework:

Context – what’s happening?
Action – what task to do?
Role – who should AI act as?
Expectation – what output do you want?

Examples:
“Summarise this policy for a board meeting.”
“Act as a grants officer and rewrite this in plain English.”

⚙️ Custom GPTs & Agents

Custom GPTs:

Personalised versions of ChatGPT tailored to an organisation’s tone, data, and needs.
Example: SOLE Virtual Director — stores minutes, creates agendas, answers policy questions.

Agents:

AI systems that act autonomously.
Can learn, plan, or collaborate with other agents.
Used for chatbots, scheduling, website optimisation, data analysis, etc.

🛠️ Useful AI Tools

Fireflies / Otter AI: meeting notes and transcription
SmythOS / N8N: build agents without code
Vapi / Bland AI: voice and phone-based AI
Synthesia / Leonardo / Sora: create avatars, images, and video
Infoxchange.org: guidance for NFP AI adoption

🧭 Strategic AI Planning

Use frameworks to plan and prioritise:

Envision where AI adds value
Prioritise high-impact, low-complexity projects
Identify barriers and risks
Mitigate before rollout

Next Steps for NFPs:

Identify AI Champions.
Develop an AI Policy and Tool Register.
Run AI training sessions.
Improve data quality and governance.
Plan integration and maintenance of tools.

📍 Key Takeaways

AI is transforming all sectors — including non-profits.
Ethical use, governance, and Māori data sovereignty are essential.
Start small: experiment safely, train your team, and document your approach.
Policies and human oversight protect people, data, and trust