Little Language Models

Challenge

The client, a leader in the enterprise machine learning model-building product space, wanted to standardize content across their platform, but lacked unified product content guidelines.

Without documented standards for voice, tone, terminology, and UI copy patterns, design and writing teams worked inconsistently, slowing the path to engineering readiness and creating friction for users encountering varied messaging.

Discovery

I conducted a foundational audit of the existing design system and documentation standards. I analyzed voice, tone, and writing style parameters.

During discovery, I also audited the trial experience, creating a separate guide with notes on gaps and improvement opportunities.

The comprehensive style guide I created included annotated examples and decision trees for edge cases.

Approach

• 70-page annotated language style guide with foundational patterns and best practices
• 6-part structure covering voice, tone, grammar, UI elements, and information architecture
• “Avoid + Verify” sections for quick reference on common terminology pitfalls

Example of a tactical IA improvement recommended: “Avoid + Verify” sections for quick reference on common terminology pitfalls

– Avoid: Terms that create confusion, anxiety, or accessibility issues (e.g., “instantly,” “abort,” “in the app”)
– Verify: Standardized versions of frequently misused terms (e.g., “Accepted values” vs. “Valid values”; capitalize user roles like “Creator,” “Viewer,” “Owner”)

These sections are findable, clear, relevant, and timely, reducing back-and-forths over common disagreements.

Impact

Result: A single source of truth that enables teams to move faster, maintain quality, and scale confidently.