Human–in–the–loop AI across training, communications and digital capability
The Digital Transformation Hub (DTH), delivered by Infoxchange, uses artificial intelligence as a supporting tool across training design, communications, webinars and AI readiness work. In line with the National AI Centre’s principles for safe and responsible AI, our approach is grounded in transparency, human oversight, privacy safeguards and a clear commitment to social justice. AI is used to augment human expertise, not replace it, and responsibility for every output remains with named staff.
This case study outlines how AI is embedded in practice across the DTH team and how those uses are governed.
AI in training design: building flexible learning at scale
Traditional eLearning platforms can be expensive, rigid and dependent on proprietary formats. Matt, Training Program Designer and Manager, uses Google Gemini, ChatGPT and Claude to support the development of custom Gems for webinar planning and the design of self-paced learning modules.
“We use Gemini “gems”, ChatGPT and Claude to develop webinar planning documents, and to help design and develop custom self-paced modules,” shared Matt.
The decision to introduce AI into learning design was driven by the need for flexibility and long-term sustainability. AI-assisted development enables rapid prototyping, modular design and full visibility of the underlying architecture. Rather than being locked into subscription-based systems, the team can build and maintain its own structures.
In practice, AI supports the drafting of webinar planning documents, the structuring of modules and the prototyping of layouts. It also assists with coding tasks in JavaScript, HTML and CSS. Importantly, AI does not determine curriculum, learning outcomes or final content. It supports the technical and structural build, while pedagogical design and quality assurance remain human-led.
A key example is the development of Responsible AI: Foundations for NFPs; a self-paced module endorsed by the National AI Centre and now live for participants to complete and earn a Responsible AI badge.
“I used Gemini to help build the recently released Responsible AI Essentials. This involved building a modular engine/template and designing custom layouts using JS, HTML and CSS.”
Rather than relying on off-the-shelf learning software, the module was built from the ground up using AI-assisted development. This approach allows the DTH team to update content as AI guidance evolves, refine structure based on participant feedback and maintain full ownership of the source code. AI accelerated the technical build and enabled faster iteration. Human expertise ensured alignment with responsible AI principles, verified factual accuracy and provided final approval before publication.
AI in communications and community delivery across the Asia-Pacific
AI also plays a role in enabling the DTH’s AI Community and webinars to remain accessible and responsive across the Asia-Pacific region. Richard, Digital Community Lead at Infoxchange, uses Gemini to streamline graphics production, support website code organisation and assist with translation for non-English speakers.
“I mainly use Gemini for a few tasks to help streamline my work. This includes generating our webinar graphics, helping organise code for our community website, and helping with some translation work for non-English speakers.”
In a field where tools and policy settings shift quickly, timeliness is critical.
“As content shifts so fast in AI, it's important that everything is made available to our audiences as soon as possible, before it loses relevance.”
Translation is a strong example of how AI supports inclusion without replacing human review. Previously, preparing multilingual content required significant manual work. AI now generates an initial translation draft, allowing the team and regional partners to move straight to a review stage rather than starting from scratch.
“While we still work with our partners in other countries, we can now use AI to provide an initial translation and commence straight to a review stage.”
Richard has also developed a custom Gemini Gem that merges translated text directly back into webpage source code without disrupting styling or links.
“I have created a Gemini Gem that I can provide the source code of a page to, and the updated translated text, and it will return with the two meshed together, retaining all of the important styling, links and more, while just updating the language. It's made a 40-minute task into a 5 minute one.”
This significantly reduces production time while maintaining structural integrity and accessibility. AI improves efficiency, but no content is published without human review and sign-off.
AI in webinars, articles and readiness work
Across webinars, articles and readiness assessments, AI may assist with drafting, summarisation, pattern identification and structural refinement. It can help prepare background research, summarise qualitative feedback or identify emerging themes in engagement data. In some cases, it supports meeting transcription and captioning to improve accessibility.
Infoxchange has also developed an AI-driven AI Readiness Assessment tool, applying our deep knowledge of the not-for-profit sector and the current AI environment. The tool analyses information provided by organisations to assess their technology environment — including data maturity, cybersecurity posture and foundational technology readiness — alongside broader indicators of AI readiness such as executive buy-in, governance structures and staff capability.
Using these inputs, the system generates a structured narrative around an organisation’s overall readiness score and produces a report designed for organisational leadership and staff. The report outlines practical next steps to strengthen readiness, identifies opportunities for safe and responsible AI exploration, and recommends relevant guides, research and training resources to support capability building.
However, AI is not used to make decisions about individuals, training priorities or service access. It does not independently publish content or provide final advice. Every AI-assisted output undergoes human review for accuracy, tone, context and alignment with Infoxchange’s values. Factual claims are verified using trusted primary sources. Outputs are reviewed for potential bias, misrepresentation or unintended harm, particularly where content relates to vulnerable communities or policy guidance.
The DTH team is mindful of data governance and privacy. Sensitive or confidential data is not entered into AI tools. Responsibility for final content always sits with a named staff member.
Alignment with safe and responsible AI principles
The DTH approach reflects core principles of safe and responsible AI. Transparency is maintained through clear disclosure of where AI is used. Human accountability is preserved through explicit review and sign-off processes. Privacy and data governance are embedded in operational practice. Fairness and harm minimisation are addressed through contextual and bias checks before publication. AI is adopted not for automation alone, but to strengthen digital capability and public benefit in the not-for-profit sector.
In practice, this means AI may generate drafts, code prototypes or structural suggestions, but people refine, verify and approve them. AI may accelerate translation, but human partners review and contextualise the final version. AI may help build a modular learning engine, but human educators design the curriculum and ensure alignment with responsible AI frameworks.
Infoxchange’s Digital Transformation Manager Sophie Souchon shared that AI will never be a replacement for critical judgement.
“AI at the Digital Transformation Hub functions as an augmentation tool, accelerating learning design, improving multilingual accessibility, streamlining communications and supporting readiness work” Sophie clarifies. “Yet it does not replace professional judgement, ethical oversight or accountability.”
From the translation to delivery of community content across the Asia-Pacific, AI is introduced deliberately, disclosed clearly and governed carefully. Innovation and responsibility operate together. AI supports the work – people remain responsible for its purpose, integrity and impact.