Published policy
Our AI marketing policy: where the machine works and where a person signs
Most agencies now use AI and very few will tell you where. You are entitled to know exactly what a machine touched, what it did not, and who checked it.
The short version
LAUNCHT is a marketing agency. Our production runs on automation, which is why work reaches you in days rather than months and costs what it does. The automation does the production. A person does the judgement, and that person is accountable for every piece that leaves here.
This AI marketing policy is published rather than internal, so you can hold us to it. Last updated 13/08/2026.
Automation is why the work ships. It is not what the work is. Every piece that leaves here has been read by a person who is accountable for it, against a written brief for your brand, and rejected if it reads like it could belong to anyone else. We would rather send you one thing that works than four that fill a calendar.
And the part that makes it more than a claim: if we ever send you something that reads like it came out of a machine, tell us, because that is a failure of our process and we will show you what we changed.
Accuracy
What stops us sending you something that is wrong
The honest risk with AI is not that it writes badly. It is that it writes plausibly while being wrong, and a plausible mistake is harder to catch than an obvious one.
- Every factual claim in your copy is traced to a source before it ships. Figures come from your own analytics, your own store, your own platform, or a named public source. If a number cannot be traced, it does not go in
- We do not invent results. No case-study figure and no claim about what we achieved for anyone appears unless it came out of a real account and can be shown to you
- Anything published to your site is checked against the live platform after it goes up, not only before
- We tell you when we are unsure. A flagged uncertainty is cheap. A confident error you repeat to a customer is not
Where AI is not used at all
- Deciding your strategy
- Choosing which products to feature
- Pricing
- Anything requiring judgement about your market or your customers
Those are decisions, not production.
Your data
What goes into a model, and what never does
Never, under any circumstances
- Your passwords, API keys, tokens or platform credentials. These live in an encrypted store and are used by software to talk to your platforms. They are never typed into a chat, a prompt or a model
- Personal data about your customers beyond what a specific task genuinely requires. We do not load your customer list into a model to see what it finds
- Anything you have marked confidential. Tell us once and it is excluded
Your data is not used to train anything. We use commercial AI services under business terms that exclude customer content from model training. We do not fine-tune models on your material, and we do not use one client's work to produce another client's.
POPIA. We process personal information only where there is a lawful basis and only for the purpose you engaged us for. If a task would involve customer personal information, we tell you first and agree the scope. That is a plain statement of how we work rather than legal advice, and your own obligations as the responsible party remain yours.
Where your data sits: in the platforms you already own, and in our client portal. We do not create shadow copies of your systems.
Ownership
You own the work, and it never leaves your accounts
Everything produced for you - copy, images, designs, code, strategy documents, reports - is yours on payment, without conditions.
- We work inside your own accounts. Your website, your store, your ad accounts, your analytics, your Google Business Profile. You keep the logins
- If we stop working together, nothing breaks and nothing has to be handed over, because it was never anywhere else
- We never generate your logo or your trademark with AI. Those are made properly, precisely because ownership must be unambiguous
On AI-assisted work and copyright, stated honestly: purely machine-generated output has an uncertain copyright position in most places, South Africa included. In practice our work is written and directed by a person and edited by a person, which is what carries the authorship.
Where a piece is heavily machine-generated - a background, a decorative illustration - we tell you, so you can decide whether to use it somewhere that matters, like a trademark.
The checkpoints
Where a human signs, and where the record lives
Every piece of work passes these gates. The record of each one is in your portal, which makes it your evidence rather than our report about ourselves.
| Checkpoint | What is checked | Where the record lives |
|---|---|---|
| Brief | That the work is written against your brand brief and voice guide, not a generic one | The brief on the deliverable |
| Facts | Every figure and claim traced to a real source. Nothing invented | Noted on the deliverable |
| Voice | Read against your own voice guide, and rejected if it could belong to another business | Internal review status on the item |
| Internal sign-off | Nothing reaches you until it has been reviewed here first | Timestamped status change |
| Your approval | You see the actual work, and comment or approve | Your comment and approval, timestamped, in your portal |
| Published | Checked live on the real platform after publishing, not only in the document | Activity record on your account |
| Reported | What was published and what it produced | Your Reports page, monthly |
Nothing is published to your website, your store or your social accounts without your approval, with one exception you set yourself: where you have explicitly asked us to publish on an agreed schedule without a per-item approval, which some clients prefer for routine content. That is your choice and it is reversible at any time.
Disclosure
What we tell you, and what you tell your audience
To you: everything. Ask about any piece of work and you will be told how it was made. We do not charge differently for AI-assisted and human work, because you are buying the outcome and the accountability, not the number of hours it took.
To your audience: your call, and we follow it. Around three quarters of people say they want to know when content was created with AI, so the question is real. If you would like an explicit disclosure on your content, tell us and we add it as standard.
- Photography of real people, real places and real products is real. We never present a generated image as a photograph of your business, your staff or your products
- Generated imagery that is illustrative needs no label, in the same way a stock photograph never did
- Anything presented as evidence - a testimonial, a review, a customer photograph, a result - is real, always. This is the line that actually matters and we will not cross it even if asked
- A person's words are that person's words. We do not put a generated quote in a real person's mouth, including yours
Hold us to it
If something on this page is not being honoured, say so. The whole point of publishing it is that you can check rather than take it on trust.
This policy does not make us right about everything. It makes it clear who is accountable when we are wrong, and gives you a record you can check.