New Feature: Brand Voice Scoring & Compliance

Priority: 🟡 Pre-launch
Area: Governance / Quality
Competitive Reference: Jasper (Brand IQ + Brand Voice + Style Guide + Visual Guidelines), Copy.ai (Brand Voice + Infobase)

Why This Matters

Jasper’s entire competitive moat is built on brand governance — every piece of content is scored against the brand’s voice, style guide, and visual guidelines before delivery. This is exactly what a digital marketing agency needs: guarantee that AI-generated content sounds like the client’s brand, not generic AI output.

Leadmetrics has brand assets (colors, logos, fonts) and skill files, but no automated scoring or compliance check on generated content.

What Leadmetrics Has Today

• ClientContext model — stores business info, industry, target audience (free text)
• Brand assets — logos, colors, fonts, visual style stored per tenant
• Skill files — .md files injected into agent prompts as context
• RAG knowledge base — tenant documents indexed in Qdrant for agent retrieval
• No brand voice definition — no structured tone-of-voice, vocabulary, or style rules
• No content scoring — generated content is not automatically scored for brand alignment
• No compliance check — no automated check for banned words, required disclaimers, character limits

What Competitors Offer

Proposed Implementation

Step 1 — Brand Voice Definition Model

Add to Prisma schema:

model BrandVoice {
  id              String   @id @default(cuid())
  tenantId        String   @unique
  tonePrimary     String   // e.g., "Professional", "Friendly", "Authoritative"
  toneSecondary   String?  // e.g., "Witty", "Empathetic"
  personality     String[] // e.g., ["Innovative", "Trustworthy", "Approachable"]
  vocabularyUse   String[] // Words/phrases TO use
  vocabularyAvoid String[] // Words/phrases to AVOID
  writingStyle    String   // e.g., "Active voice, short sentences, no jargon"
  exampleContent  String?  // Sample paragraph in the brand's voice
  targetAudience  String?  // Who the content speaks to
  createdAt       DateTime @default(now())
  updatedAt       DateTime @updatedAt
  tenant          Tenant   @relation(fields: [tenantId], references: [id])
}

Step 2 — Inject Brand Voice into Agent Prompts

Enhance the context-file-writer agent output to include a “Brand Voice” section that all content agents read:

## Brand Voice
- Tone: Professional yet approachable
- Personality: Innovative, trustworthy
- Always use: "transform", "empower", "drive results"
- Never use: "cheap", "best in class", "synergy", "leverage"
- Style: Active voice, short paragraphs, no jargon

Step 3 — Post-Generation Brand Score

After any content agent completes, run a brand scoring pass (lightweight LLM call):

const scorePrompt = `
Score this content 0-100 for brand voice alignment.
Brand voice: ${brandVoice}
Content: ${generatedContent}
Return JSON: { score, issues: [{ text, reason, suggestion }] }
`;

Store the score on the deliverable:

• brandScore: Int on BlogPost, SocialPost, etc.
• Below threshold (e.g., < 70) → auto-flag for review with specific issues

Step 4 — Dashboard UI

• Brand Voice Settings page under tenant settings
  • Guided setup wizard: “Describe your brand in 3 words” → AI generates voice definition
  • Or manual input for each field
• Brand Score badge on every content card in the deliverables list
  • 🟢 80–100: On-brand
  • 🟡 60–79: Needs review
  • 🔴 <60: Off-brand
• Compliance details panel on content detail page showing specific issues
• Brand Health dashboard — aggregate brand consistency score over time

Step 5 — Brand Compliance Audit (Jasper-style)

A one-time or periodic scan of the tenant’s website:

1. Crawl the tenant’s live website (using existing website-crawler agent)
2. Score each page against the defined brand voice
3. Generate a “Brand Compliance Report” with page-by-page scores and recommendations

Related

• Content Repurposing Engine — derivatives must maintain brand voice
• AI Image Generation — visual guidelines for image generation