AI POLICY
We use AI.
We also use our judgement.
AI is part of how modern marketing gets researched, explored and made. We use it where it makes the work better or faster, but not as a substitute for expertise, responsibility or common sense.
Rethink Social uses AI to support research, analysis, idea development and drafting, while keeping human judgement, confidentiality and accountability with the people doing the work.
Useful tool. Terrible boss.
JUMP TO
HOW WE USE AI
For the useful bits. Not because it’s shiny.
We may use AI tools to support research, analyse information, explore patterns, organise thinking, develop ideas, test different approaches and assist with drafting.
We may also use AI to speed up repetitive work or help us examine larger amounts of information than would be practical manually.
The point is not to produce more for the sake of it. We use AI when it gives us a better way to understand the problem or improve the work.
Faster can be useful. Better matters more.
Human Judgement
AI can suggest. People decide.
AI can surface patterns, suggest routes and produce plausible answers very quickly. Plausible isn’t the same as right.
We review the work we use. Important recommendations, interpretations and decisions are made with human judgement, not handed over to a model.
That matters particularly where the context is messy, the evidence is incomplete or the consequences of getting something wrong are significant.
Nothing really beats real world experience like a human.
CLIENT & CONFIDENTIAL DATA
Confidential means confidential.
Client work can involve strategy, research, customer information, commercial plans and other material that was never intended for the wider world.
We don’t casually paste confidential information into public AI tools. Where AI is useful for client work, we choose tools and settings appropriate to the information involved and limit what we provide to them.
Where practical, we remove or reduce identifying and sensitive detail before using AI. We also follow any confidentiality, data-protection or AI requirements agreed with the client.
What we won’t do
We won’t knowingly use confidential client material to train a public AI model, publish it through an AI service or reuse it for another client.
If client material includes personal information, our handling of that information also needs to comply with the relevant data-protection requirements and our agreements with the client.
Cannot say this enough, this is never about convenience. For us AI is a tool, not a brain.
ACCURACY & SOURCES
Confidently wrong is still wrong.
AI can be useful for finding patterns, summarising information and suggesting where to look next. It can also make mistakes, invent details or present uncertainty as fact.
We check important claims before we use them. Where evidence matters, we look for the original source, understand the context and make sure the conclusion is actually supported.
We don’t knowingly use fabricated citations, invented statistics or AI-generated “facts” as evidence.
When the stakes are higher
The more important the recommendation, claim or interpretation, the more scrutiny it gets. AI output is a starting point for judgement, not proof on its own.
Simply put, if we can’t stand behind it, it doesn’t go out.
CONTENT & CREATIVITY
AI can start the work. It doesn’t get to finish it.
We may use AI while developing ideas, exploring different angles or working through early drafts. It can be genuinely useful for getting somewhere faster.
But generated content isn’t automatically good content. Anything Rethink publishes or delivers still needs judgement, editing and a reason for existing.
We also take reasonable care not to ask AI to imitate another person’s work or knowingly reproduce protected material. Where we use third-party material, the usual copyright and attribution rules still apply.
Images and other generated media
We may use AI-generated or AI-assisted imagery where it suits the work. We won’t deliberately present synthetic material as documentary evidence of something that really happened.
Where the fact that something is AI-generated would materially affect how someone understands or trusts it, we’ll make that clear.
“The AI made it” is not an editorial standard.
transparency
If AI matters to the answer, we’ll say so.
We don’t think every use of AI needs a flashing warning label. But where AI materially affects the work, the evidence or how something should be understood, we’ll be open about it.
That includes situations where AI is used in a way that could affect someone’s rights, where personal information is involved, or where the synthetic nature of content would change how a reasonable person interprets it.
If AI contributes to a recommendation or decision that matters, we’ll make sure the human role is clear and that the output can be challenged rather than simply accepted because “the model said so”.
When personal data is involved
If we use AI to process personal information, our Privacy Policy explains the relevant purposes, retention and sharing arrangements. Where automated decision-making has legal or similarly significant effects, additional transparency and safeguards may apply.
No smoke. No mirrors. No pretending the robot did it all.
KEEPING THIS HONEST
AI will change. So will this policy.
The tools we use, what they can do and the rules around them will keep evolving.
We’ll review this policy as our use of AI changes and update it when something material changes in how we work.
Last updated: September 2026
New tech. Same responsibility.