Disclosure
We are an AI agency. Here is where we stop using it.
Every agency in this market now sells AI. We read eleven of their sites while building this one, and not one of them says where it stops. That silence is doing a lot of work, because the question a buyer actually has is simpler than any of the sales copy: am I paying a person, or a prompt?
This page answers it, in enough detail to hold us to.
Where we do use it
- Research and analysisClustering thousands of queries, spotting patterns across a crawl, summarising a competitor set. Work that is mechanical, checkable, and would otherwise eat a day.
- First drafts of structureOutlines, heading structures, FAQ candidates. A skeleton a person then rewrites — never the finished text.
- Internal automationReporting pulls, data cleaning, alerting when a number moves. Our own workflows, not yours.
- Custom agents we build for clientsThis is a service we sell openly at /ai-automation-services/. It is the client's system, built to their brief, and they know exactly what it does.
- The audio version of articlesRead by a synthetic voice. Every player says so, and the recording says so in its first line.
Where we never do
- Published articlesEvery word on this blog is written by a person. If that ever changes, the post will say so at the top, in the same place the audio disclosure sits.
- Case study numbersNo figure on this site is generated, estimated or rounded up by a model. Each one traces to an export with a date on it.
- Client reports and recommendationsA model can summarise data. It cannot decide what to do about your business, and it does not sign anything we send you.
- Reviews and testimonialsNever written, never "polished", never generated. The ones on /reviews/ point at platforms we cannot edit.
- Images of people or placesNo generated faces, no invented offices, no AI photography passed off as a client's premises.
- Anything sent to a client without a human reading it firstIncluding the auto-reply on the contact form, which says it is automatic.
The right-hand column is longer than the left. That is deliberate, and it is the half worth reading.
Four rules we hold to
Short enough to remember, specific enough to be broken — which is what makes them worth publishing.
A person is accountable for everything that leaves here
Not "reviewed by" as a formality. If it is wrong, a named person got it wrong, and that is the point of a byline.
Your data is not training data
We do not put client data into tools that train on it. Where a tool is used, it is one with that switched off, on a paid tier.
Disclosure is at the point of use
Not buried on this page. The synthetic audio says so on the player and in the recording. Any future AI-assisted content will say so on the piece itself.
If AI made it faster, you are not charged for the hours it saved
Automation is how we keep the price where it is. It is not a way to bill for time nobody spent.
What we publish for AI, and where it says so
Blog articles carry an audio version read by a synthetic voice. The player states it under the controls, and the recording states it in its own first sentence — before anything else is said.
We mention it here because a disclosure page that only describes disclosures is not one. Listen to the example →
Separately, we publish a plain fact sheet for AI assistants: what this business does, where it works and what it charges, set out so that a model answering a question about us quotes something accurate instead of assembling a guess. It is not in the navigation, because it is a reference sheet rather than a page anyone needs in order to buy from us. Read the AI information page →