Start With What You Block
The most common mistake teams make when setting up brand governance is trying to define what good looks like. Start instead with what bad looks like.
Your first rules should be things your brand absolutely never does:
- Forbidden phrases - Words or phrases that violate your brand voice ("guaranteed results", "100% safe", superlatives without evidence)
- Tone violations - Slang, ALL CAPS, excessive punctuation that does not match your professional standard
- Format constraints - Emoji usage, hashtag policies, URL shortener restrictions
These rules are clear, unambiguous, and easy to enforce deterministically.
Layer Severity Correctly
Not every rule violation is equal. Torobari uses two severity levels:
Block Rules
Reserve block severity for violations that should never reach production under any circumstances. These are your hard constraints:
- Forbidden phrases that could create legal liability
- Content that violates regulatory requirements
- Formatting that breaks your brand guidelines absolutely
Escalate Rules
Use escalate severity for violations that need human judgment. These are your soft constraints:
- Slang that might be acceptable in some channels
- Length limits that depend on context
- Claims that are not strictly forbidden but need verification
Channel-Specific Rules
Your brand voice on Twitter is different from your brand voice in a legal document. Use channel scoping to apply rules where they matter:
- Social - Might allow more casual tone but still block forbidden phrases
- Ads - Strictest rules, requiring disclaimers and blocking all unverified claims
- Email - Moderate formality with length constraints
- Web - Standard brand voice with full rule coverage
- Print - Formal tone with additional formatting requirements
Avoid Over-Engineering
The temptation is to create dozens of rules covering every edge case. Resist this. Start with 5-10 rules that cover your most important constraints. Watch what gets blocked and escalated. Add rules based on real patterns, not hypothetical concerns.
A small, well-tuned ruleset is more effective than a large, noisy one. If your team is ignoring escalations because there are too many, you have too many soft rules.
Measure and Iterate
Track your decision distribution over time:
- High PASS rate (>90%) - Your rules are well-calibrated and your AI outputs are generally on-brand
- High BLOCK rate (>30%) - Either your rules are too strict or your AI prompts need improvement
- High ESCALATE rate (>20%) - Consider whether some escalate rules should be blocks or should be removed
The goal is not zero blocks. The goal is that every block prevents a real problem and every pass represents genuine compliance.
The Ruleset Hash
Every time you change your rules, the enforcement engine computes a new ruleset hash. This hash appears in every decision record, creating a versioned link between your governance posture and the decisions it produced. When you audit past decisions, you can verify they were evaluated against the rules you intended.
Good governance is not about perfection. It is about consistency, transparency, and continuous improvement.