Cohesion: A CAFECITO Pillar

Lock in systematic voice consistency that AI tools can read and follow automatically.

The voice-first question

Does our cafecito taste identical shot-to-shot?

Perfect consistency requires systematic precision. Your voice rules need to be machine-readable so AI tools deliver identical personality every time.

Why AI needs systematic voice rules

Think of Cohesion like a Cuban barista's exact cafecito recipe: same beans, same grind, same timing, same technique every single shot. No improvisation. Without systematic voice rules, AI gets creative when you need consistency.

Brand-Voice JSON converts your voice guidelines into machine-readable instructions. Instead of hoping AI interprets "be warm but professional," you define exactly what warmth means in terms AI can follow systematically.

How to build your Brand-Voice JSON

Convert voice guidelines into an AI-readable format for consistency across all tools and platforms. Four sections. The Brand Voice JSON Guide walks through each field with a full worked example.

1. Voice foundation

Core identity and personality traits defined in precise, actionable terms that AI tools can interpret consistently.

{
  "brand_essence": "Miami-spirited AI strategists who help SaaS companies turn scrambled copy into voice-ready prompts that sound unmistakably authentic",
  "core_personality": {
    "warmth": 75,
    "professionalism": 60,
    "confidence": 80,
    "playfulness": 45
  },
  "cultural_elements": [
    "Miami energy and warmth",
    "Cuban coffee culture references",
    "Beach/ocean metaphors for flow"
  ]
}

2. Language patterns

Specific vocabulary preferences, phrase structures, and linguistic patterns that define how your brand communicates.

{
  "preferred_vocabulary": {
    "instead_of": ["utilize", "leverage", "optimize"],
    "use": ["use", "work with", "improve"]
  },
  "signature_phrases": [
    "voice-ready prompts",
    "sounds unmistakably you",
    "AI-assisted, human-vetted"
  ],
  "sentence_structure": {
    "length": "medium",
    "complexity": "conversational",
    "active_voice_percentage": 85
  }
}

3. Contextual adaptations

How voice adapts across different channels and situations while maintaining core personality traits.

{
  "channel_adaptations": {
    "email": {
      "formality": 40,
      "personal_touch": true,
      "signature_style": "Warm professional"
    },
    "social_media": {
      "formality": 25,
      "energy_level": 85
    },
    "support": {
      "formality": 60,
      "empathy_level": 90,
      "solution_focused": true
    }
  }
}

4. Voice guidelines

Specific rules and constraints that ensure consistency across all AI-generated content.

{
  "voice_rules": {
    "always_do": [
      "Include specific examples",
      "End with clear next steps",
      "Acknowledge challenges authentically"
    ],
    "never_do": [
      "Use jargon without explanation",
      "Make promises we can't keep",
      "Sound robotic or corporate"
    ],
    "tone_indicators": {
      "confident_but_humble": 70,
      "helpful_never_pushy": 90,
      "practical_over_theoretical": 85
    }
  }
}

Implementing Brand-Voice JSON

How to create, test, and deploy your machine-readable voice guidelines over three weeks.

Week 1: Convert your Voice Charter to JSON

Goal: first draft of Brand-Voice JSON.

Take your existing Voice Charter document
Break down personality traits into numeric scales (0-100)
Convert voice principles into actionable rules
Define channel-specific adaptations

Week 2: Test with AI tools

Goal: strong voice consistency scores.

Inject the JSON into 5 different AI prompts
Generate sample content across channels
Run team voice consistency tests
Identify gaps and inconsistencies

Week 3: Deploy across systems

Goal: JSON integrated into two or more workflows.

Integrate into custom GPTs and Claude projects
Add to prompt libraries and templates
Build into automation workflows (Zapier and similar)
Share with the team for manual prompt enhancement

What you walk away with

A Brand-Voice JSON file that any AI tool can read the same way every time. Machine-readable rules outperform written guidelines alone because there's nothing left to interpret: voice-related editing drops as outputs match brand personality from the start, and the team finally shares a concrete definition of what "sounds like us" actually means.

Cohesion