LordCrawford
One New Zealand
Visit One NZ →From July to August 2026, I lived in Auckland, New Zealand and interned on the AI & Data team at One New Zealand through a partnership with the Stanford Doerr School of Sustainability. I worked across four projects: agentic interfaces, a digital inclusion website, a digital twin of work, and responsible AI testing. I now serve as an unofficial liaison between Stanford and One NZ.
Students build real-world skills in summer sustainability internships - Stanford Doerr School of Sustainability | September 29, 2026
Specifications
Dates: July - August 2026
Team: AI & Data
Tools Used: Replit, Claude Code, Google CX Agent Studio, ElevenLabs Agents
Role: Human-Centered AI Intern
Contributions
Agentic Interfaces
I prototyped how customers could reach One NZ through assistants they already use, like Siri and ChatGPT, to ask about upgrades or a bill. I also designed an in-store flow from both sides of the counter: the customer's app knows why they're visiting, while an employee assistant prepares their account details before the conversation starts. Alongside this, I compared third-party agent platforms and recommended where each fits in the team's experimentation.
*Visuals withheld under NDA.
Digital Inclusion Website
I designed the website and landing page for research on how rangatahi (young people) experience being online. The goal was to turn a long report into a page people can scan or explore in depth, with young people's voices up front. Read the full report →
Digital inclusion website
*Design mockup. The quotes and contributor details shown are unofficial placeholders, not taken from the published research.
Digital Twin of Work
I explored how AI could map the people, tasks and knowledge behind a team's work, then built a working prototype on my own projects. It pulls blockers, next actions and risks across projects, hands runnable tasks to Claude Code, and walks through scenarios I describe to suggest options and what to watch for.
Digital twin scenarios
Responsible AI
I reviewed the team's testing guidance for employees building assistants with low/no-code tools, pushing for concrete behavioural tests and clearer points for when to escalate. I also shared findings with the AI team on measuring adoption, layering safeguards for agents, and evaluating multi-agent systems.