Knak / 2024-2026
AI Brand Voice
Made on-brand AI output automatic for marketers while giving administrators one place to govern voice and language, contributing to 25% growth in AI feature usage.
Design Highlights
Overview
I designed AI Brand Voice as a foundational control layer for Knak's AI-generated content. It carries brand context across text generation, subject lines, and full email generation so teams get consistent output without repeating the same instructions.
Challenge
Before Brand Voice, users had to manually add brand context every time they prompted an AI generation feature. There was no persistent memory or shared context across AI interactions, no enforcement of brand standards and no consistency between generated outputs. This created duplicate effort for every generation request and meant that brand compliance depended entirely on the user remembering to include the right context each time. The result was inconsistent AI outputs that required manual rewriting to match brand guidelines.
Approach
I designed one system for two distinct needs: admins configuring brand standards and marketers generating content. Admins can define voice attributes and flexible rules for tone, personality, language, and syntax. For end users, I made the integration invisible so brand context is applied automatically without adding steps to the creative flow.
Solution
I built a configuration experience where admins define voice attributes and encode standards as flexible rules, from broad tone direction to precise language and formatting conventions. Once configured, every Knak AI feature pulls those attributes and rules into its context automatically.
Impact
Marketers receive content that starts closer to their brand without repeating the same instructions or rewriting every result. Customer administrators can define voice and language rules once, then apply them across Knak's content-generation experiences. For Knak, Brand Voice became a shared control layer for the AI portfolio and contributed to a 25% increase in MAU/WAU for AI features.
Key Learnings
I learned that the best AI guardrails are often the ones end users never have to think about. Flexible rules worked better than rigid settings because every brand expresses standards differently, while separating admin configuration from everyday generation kept the experience fast.