We document your brand's voice and set up a language model to follow it. You get fewer edits, a consistent tone across posts, emails and your website, and prompts your team can reuse.
Quote and schedule after your brief
We collect your best writing and pull out the patterns: sentence length, favourite and banned words, how you address the reader. These become a voice guide and a set of examples for the model.
We write a system prompt (a standing set of rules the model sees before every request) and a prompt library for your recurring tasks. Your team reuses it instead of improvising.
One guide with do and don't examples. A new writer or an AI agent lands in the right tone from the first draft.
We add forbidden phrases and fact-checking rules to the instructions. A person still reads every text before it goes out, and we show you what to check.
Use a package to describe what you need. We agree on the quote and schedule after reviewing your brief and requirements.
One brand, 1–2 content types, a team of up to 3
A brand with several channels: social, email, website
Businesses that want their voice in an AI chatbot or AI agent
You describe the task and current workflow. We clarify requirements, choose a first stage and prepare a quote.
You send texts you consider on-brand, plus examples that miss. Customer reviews and internal style rules help too.
We describe the voice and write the system prompt and prompts. You approve the guide before testing.
We generate test texts and you rate them blind. We agree on a measurable target upfront, such as the share of texts you accept without edits.
We train your team and hand over all files and access. Payment is made in stages.
At least 10–15 texts you consider exemplary, ideally from different channels. If you have fewer, we help you write a few reference pieces together.
Usually not. For most tasks a guide, examples and a system prompt are enough. Fine-tuning is rarely needed, and we'll say so honestly in the estimate.
That depends on account settings. We recommend plans and modes where your data isn't used for training and show you where to switch it on. The account is yours.
Yes, we can build a voice for several languages. Each gets its own examples, because tone is usually lost in translation.
In the pilot we agree on a target, for example how many test texts you accept with minimal edits. We don't promise a number upfront. We set it with you and record it.
It happens. We agree on the schedule after clarifying scope and dependencies. The first adjustment after an update is included in the month of support in the full package.
Describe your business, the outcome you need and your current workflow. We will clarify the requirements and prepare a quote after your brief.