Our approach to AI

AI is now part of how we work at Luminary. Rather than leave you to guess what that means for your project and your data, this page spells it out. It is the plain-English companion to the commitments in our client agreements, and we keep it current as our tools and the rules around them change.

Where we stand

Our position is simple: AI makes our people faster and sharper (and that benefit should show up in your timelines and budgets), but it does not replace their judgement. Every piece of work that reaches you has been shaped, checked and stood behind by a human at Luminary. That has been our mission since 1999 – 'to make digital bright, and the human experience brighter' – and AI has not changed it. If anything, it has raised the bar for what the human layer needs to add. 

What we use AI for

We use AI as a working tool across most of what we do, in the same way we use an IDE or a design platform. In practice that means:

• code assistance, review and testing on CMS and DXP builds
• drafting and editing content, which is always reviewed and refined by our people
• research, analysis and summarisation
• interpreting analytics and experimentation data
• project documentation and delivery support.

AI supports this work. It does not make decisions on your behalf, and it does not publish, deploy or send anything without a human approving it first.

These uses map to the Australian Government Digital Transformation Agency's AI use classification, the same framework government agencies report against in their own transparency statements.

What happens to your data

This is the section most clients want first, so here it is, in full:

• We use enterprise-grade AI tools only for client work. Free and consumer AI accounts are not used with client information, full stop.
• Your data is never used to train AI models. We only use tools on commercial tiers where the provider is contractually barred from training on our inputs and outputs.
• We know where your data goes, and whose laws govern it. We only use AI services hosted in jurisdictions with strong privacy law and independent judicial oversight of government access to data – currently the USA, the EU, the UK and Australia. We do not use services hosted in, or subject to the laws of, jurisdictions where authorities can compel access to data without due process or independent review.
• We minimise before we share. We give an AI tool the smallest slice of data the task needs, and we de-identify personal information wherever practical before it goes anywhere near one.
• Where we work with your analytics data (GA4, for example) or customer data, we agree the approach with you first, strip or mask anything that could identify individuals, and handle it in line with the Privacy Act and the platform's own terms.
• For sensitive engagements, we can go further, including zero data retention arrangements where the provider stores nothing beyond the immediate response.
• The same security discipline applies as everywhere else at Luminary. Our AI use sits inside our ISO 27001-aligned information security practices, including access controls and staff training.

Who checks the work

A human, every time. AI-assisted output goes through the same review and quality assurance as anything else we produce: peer review on code, editorial review on content, and sign-off by the accountable lead on your engagement. If AI has materially contributed to a deliverable, we will tell you. Accountability for the quality of our work sits with us, not with a model.

How we choose our tools

Before any AI tool touches client work, we check its data handling terms, confirm the no-training commitment in writing, and confirm it meets our security requirements. Tools that cannot give us those assurances do not get used with client data. The list of tools we use evolves quickly, so we maintain it as a living register rather than printing it here – your statement of work sets out the specific tools on your engagement, and you can ask us for the current list at any time.

We work at the enterprise end of this market ourselves. In November 2025 Luminary became the first digital agency in APAC accredited on Optimizely Opal, and we apply the same scrutiny to Opal that we apply to every other tool (for the record: it does not train on customer data either).

AI in the things we build for you

This page covers how we use AI to deliver our work. Sometimes, though, AI is part of the work itself – a chatbot, a search experience, an agentic feature we build into your product. Anything like that is scoped, governed and disclosed in your statement of work, with its own data handling and human oversight arrangements agreed with you before we build it.

The standards we work to

We align our AI practice to the Australian Government's AI Ethics Principles and Australia's Voluntary AI Safety Standard, and we are extending our ISO 27001-aligned security practices toward ISO/IEC 42001, the emerging international standard for AI management. As a Certified B Corp, we hold ourselves to account for the impact of our decisions on people and the planet, as well as profit, and that includes how we adopt AI.

Your choices

You are in control of how AI is used on your engagement:

• Ask us anything, at any time. We will tell you what tools we are using and how.
• Set restrictions. If there are categories of data or work you want kept away from AI tools, we will write that into your statement of work.
• Opt out. If you would prefer we not use AI on your engagement at all, tell us and we will agree what that means for scope and timeline.

Questions about any of this can go to your Luminary contact, or reach out to us here.

AI and the environment

AI has a real energy footprint, and we do not pretend otherwise. Luminary has been certified carbon neutral, with emissions offset back to our founding in 1999, and we factor the footprint of our AI use into that commitment. We are also taking steps to offset the water footprint of our AI usage.  In addition, we favour using the right-sized model for the job rather than the biggest one available. (See our annual Impact Report). 

Keeping this page honest

AI tooling, provider terms and the law are all moving quickly. We review this page quarterly and update it whenever something material changes.

Last updated: 22 July 2026

This page explains our approach in plain English. The binding commitments sit in our master services agreements and statements of work. If anything here appears to conflict with your agreement, the agreement prevails.