AI is part of how we work. It helps us create, build and explore.

We use AI where it can make our work more useful, efficient or effective. This can include research, writing, images, software development, testing, documentation, analysis and prototyping.

AI can help us get from an idea to something tangible much faster.

It does not mean that everything we produce is generated by AI. We decide where AI is useful and where human judgement, experience and conventional tools are the better choice.

AI is not only something we use behind the scenes. We can also build AI into the solutions we create for clients.

The goal isn't to add AI simply because we can. We look at the problem first and then determine whether AI is actually the right tool.

We don't believe that every piece of information should automatically be given to an AI service. Before using client information with an AI tool or service, we consider the context, sensitivity and requirements of the project.

• What information is being processed
• Why the information is needed
• How sensitive that information is
• How that service handles information
• Which AI service is being used
• Whether the information can be minimised or anonymised
• What security and privacy requirements apply to the project

Where appropriate, we may use APIs, business services, private environments or other configurations designed for the particular requirements of a project. The exact approach depends on the project, the information involved and the services being used.

This page explains how we use AI. Our Privacy Policy explains our broader approach to personal information and privacy.

AI can produce impressive results, but it can also be wrong. AI-generated text can contain inaccuracies, generated code can contain errors, images can contain unexpected details, and research can be incomplete.

We review, test, refine and validate AI-assisted work according to what the project requires. For software and systems, this can include testing functionality, reviewing generated code and checking how the result behaves in the real application. For content and research, it can include checking facts, sources, context and suitability.

Human judgement remains part of the process.

When AI is used to create content or assist with a project, there may be several different elements involved.

These are not necessarily treated in the same way. Where ownership or licensing matters to a project, the applicable client agreement and the terms of the relevant third-party services determine the rights and responsibilities involved.

We don't represent AI-generated material as exclusively human-created when it was not.

Similar requests can sometimes produce similar results. AI-generated content should therefore not automatically be treated as something that could never be produced for another person.

Where originality, exclusivity or intellectual-property protection is particularly important, we take that into account when deciding how AI should be used.

When AI is used to create or modify digital content, some AI systems may add information that records how the content was created or changed. This is generally referred to as provenance.

Provenance can take different forms depending on the AI tool being used. It may include metadata, Content Credentials, machine-readable information, or, in some cases, an invisible watermark or other technical signal.

The purpose is transparency — helping people and systems understand that AI was involved and, where supported, providing information about the content's origin or history.

What provenance does — and does not — mean

Provenance does not automatically transfer ownership of your content to an AI provider, to IdeaOffset, or to anyone else.

It is also not the same as Digital Rights Management (DRM). Provenance is intended to provide information about content and its history; it is not designed to prevent you from using the content.

Where ownership, licensing or intellectual-property rights are relevant, those rights are determined by the applicable agreement, the nature of the work and the terms of the services used — not simply by the presence of a provenance record.

Will provenance change my content?

Usually, provenance information is designed to accompany or be associated with the content rather than change the underlying content itself.

Depending on the technology, provenance may be stored as metadata, embedded information, or associated with the content through another mechanism. Some systems may also use visible or invisible watermarking.

This means an AI-generated image, document, video, audio file or other output does not automatically become less usable simply because provenance information exists.

However, different AI tools and platforms handle provenance differently. Information may be preserved, changed, removed or lost when content is edited, converted, exported or processed by another system.

Why we support transparency

We believe clients should be able to use AI without having to hide the fact that it was involved.

Where AI-generated content is subject to transparency, disclosure or labelling requirements, we will take those requirements into account. Where provenance information is available, we may preserve it rather than deliberately removing it.

The goal is not to make AI-generated work less useful.

The goal is to make its origin clearer while keeping the focus on the value of the work itself.

There is no single AI service that is right for every project.

The tools we use may also change over time as AI technology develops. We choose tools based on what the project needs, not simply because a particular AI service is popular.