AI policy

Why this policy exists

Artificial intelligence is changing how design studios work. We think that change should be navigated thoughtfully – not avoided, but not adopted uncritically either. We work with organisations tackling some of the world’s most pressing challenges. The integrity of that work, and our honesty with the people we work with, depends on being clear about how we use AI, why we’ve made the choices we have, and where we’ve drawn the line.

This policy sets out our current position on AI use across the studio. It will evolve as the technology, our understanding, and better options develop.

Our approach in brief

We use AI as a tool for thinking, organising, and running our operations – including helping us draft and refine written documents. We do not use AI to generate the creative work that forms the core of what our clients pay us for. We are honest with our clients about how we use AI and give them the ability to opt out of AI-assisted workflows. We take the environmental and ethical dimensions of AI seriously – our default tool is chosen specifically for its environmental credentials, and client data is handled only through tools with the strongest available privacy protections.

How we use AI

We use AI to support our own processes and thinking. Specifically, we use it for:

• Collating, summarising, and organising notes and workshop outputs
• Synthesising research and background material to support our own thinking
• Drafting and refining written documents, including proposals, specifications, and instructional content
• Helping to shape our own marketing collateral, such as LinkedIn posts and website copy – never as finished, 100% AI output, but as a collaborator on writing we shape and own
• Internal studio operations: knowledge management, process documentation, and administrative tasks
• Project management support: structuring project information, summarising client briefs, and organising project documentation within our workspace

The boundary is simple. AI assists the scaffolding, human judgement shapes the work. The final creative and strategic output is made by people.

How we do not use AI

We do not use AI to generate the creative work our clients engage us for. This means:

• No AI-generated imagery, brand assets, visual design, or graphic output of any kind
• No AI-generated strategy, frameworks, or creative direction presented as our professional output
• No finished copy that has not been written or shaped by us

This position reflects something we believe about the work itself. Design that engages, influences, and activates people requires human authorship. Our clients are mission-driven organisations whose communications carry real weight – with their audiences, their funders, and the communities they serve. That work should be made by people, for people. AI cannot carry the interpretive, creative, and ethical judgement that good impact design requires.

We are also concerned about the broader ethics of generative AI. Many large models have been trained on the work of artists, writers, and designers without their consent or compensation, and we do not want to use or profit from tools built on those foundations. We recognise this is a fast-moving area and will continue to monitor how training data is sourced and licensed.

The tools we use

GreenPT – our default tool

GreenPT is our preferred tool for studio operations. It is hosted in Europe under strict GDPR compliance, operates with a zero-training policy on user inputs, and shows real-time energy and CO₂ metrics for every prompt. Its environmental credentials are a core reason we chose it, and we believe it is the strongest option we are aware of for environmentally responsible AI use.

GreenPT is the first-choice tool for any work involving client-sensitive or confidential information, including:

• Client project information, briefs, or data
• Workshop notes, research outputs, or any materials that identify a client or their work
• Any content carrying confidentiality obligations under our Letter of Engagement

Notion (in-app AI) – project workspace and knowledge management

Notion is our project management tool and knowledge workspace. It holds client project information, contact details, proposals, and contracts, and we use its built-in AI for structuring research, summarising documents, organising project information, and drafting documentation. Notion has contractual agreements with all of its model providers that prohibit the use of customer data to train their models, and holds SOC 2 Type 2 and ISO 27001 certifications. The AI honours existing workspace permissions – it cannot access pages a user does not already have permission to see.

As directors, we have agreed which AI models within Notion are approved for use, based on the governance, safety, and ethical track record of the labs behind them.

Claude (Anthropic) – operational and complex tasks

Claude is used for studio operations and tasks that require more advanced capability than GreenPT currently provides. It is accessed through a paid account with data privacy commitments in place, and is not used for any work involving client-sensitive or confidential information.

Gemini (Google) – specific supplementary tasks

Gemini is used for specific tasks where its capabilities are best suited, such as integration with Google Workspace tools. It is not used for any work involving client-sensitive or confidential information.

Data handling and privacy

Our data handling rules are:

• Client or confidential data: GreenPT (first choice, EU-hosted, zero-training, GDPR-compliant) or Notion’s in-app AI (contractual no-training with all model providers, 30-day retention or less, SOC 2 Type 2 and ISO 27001 certified).
• Internal operational content with no client-sensitive information: any of the four approved tools (GreenPT, Notion, Claude, Gemini)
• Free-tier or unapproved AI tools: not permitted for any studio or client work

Any decision to adopt a new AI tool, or to enable a new model within an existing tool, requires review and agreement from both directors before use, including an assessment of its data handling terms.

Environmental and ethical considerations

AI tools carry an environmental cost in both training and operation – energy, water, and infrastructure. That sits alongside our environmental commitments, and we take the tension seriously.

Our response is practical rather than performative:

• GreenPT is our default because of its environmental credentials – renewable energy, EU hosting, and visible per-prompt impact
• Where we use Claude or Gemini, we do so for specific capability reasons and acknowledge these tools carry a higher environmental cost. We minimise unnecessary use and continue to evaluate whether GreenPT can cover these uses over time
• Notion’s in-app AI runs on models from several providers, each with different environmental footprints. We have limited our approved model set partly on these grounds, and the choice of model within Notion is a decision we make deliberately rather than leaving it to default selection
• We remain concerned about the ethics of large model training – particularly the use of creative works without consent – and factor this into how we evaluate tools and select models within them

Our goal is to reduce our dependence on higher-impact tools as lower-impact alternatives develop comparable capability.

Transparency with clients

We tell clients that we use AI tools across our studio operations, and this is part of our standard engagement. We are clear with clients about what AI is used for, what it is not used for, which tools we use, and how their data is handled.

How it works in practice: AI is embedded in how we run our day-to-day operations and client projects, as it now is for most design studios. It helps us work efficiently and keeps our fees down. Where a client would prefer AI not be used in their project, we will happily discuss this, and the project scope and fee structure will be revised to reflect the additional human time required. This is agreed in writing before work begins.

–

Approved by both directors of Made for the World, September 2026. This policy is reviewed annually and updated to reflect changes in our AI use, the tools available, and our understanding of their environmental and ethical dimensions.

Questions about this policy can be directed to hello@madefortheworld.studio.