The challenge
I own UX Writing for BOLD's international products. That means several products, several locales, and more copy requests than one writer can draft from scratch.
Every project has its own voice, but the standards underneath are the same: purposeful, concise, conversational, and clear, and written so it still works after translation. I was re-explaining those standards in every review. I wanted them in the tool, not in my head.
What went into it
The four quality standards
Every string has to be purposeful, concise, conversational, and clear. The assistant edits in that order, one pass per standard, the same way I do by hand.
Writing for localization
No slang or idioms, active voice where possible, room for text expansion, and no overly excited tone, since enthusiasm doesn't land the same in every locale.
Our patterns and examples
Titles in sentence case with no end punctuation. Subtitles that add value and always end with a period. Buttons that say what happens next, never "Submit" or "Click here". Each pattern comes with real examples from our products, like "Add your resume" and "Upload one you already have or create a new resume."
Voice per project, tone per moment
Voice comes from each project's own instructions, so the same assistant can write for different products without blurring them together. Tone adapts to the moment: an error message and a success message shouldn't sound alike.
What it was meant to change
Efficiency: first drafts in seconds instead of starting from a blank text box, so my time goes to the decisions only a writer can make.
Output: more surfaces covered across more locales, without adding writers.
Consistency: the same terminology, patterns, and localization rules in every product, whether or not I reviewed the string.
Standards written once, applied in every project.
Sharing it with the other writers
A tool only one person uses isn't a process. I packaged the assistant as a skill file that any writer can load into their own AI tools, walked the writing team through how it edits, and set it up so each project's voice instructions plug in without touching the core rules.
When a standard changes, it changes in one place and every writer gets it on their next draft.
Draft. Results and a before-and-after example to be added.
Full case study available on request.