admin-baked/skills/org/brand-voice

Brand Voice

Purpose

Encode a dispensary or brand's approved voice, tone, and language patterns so any agent or human producing content knows what "on-brand" looks and sounds like — without needing to ask every time.

When to Use

• Before generating customer-facing copy for a specific brand account
• When reviewing draft content for voice consistency
• When onboarding a new brand account and documenting their standards
• When an operator reports that copy "doesn't sound like us"

When NOT to Use

• Compliance review — voice is not a compliance gate; use Deebo for regulatory checks
• Factual claims about products — voice does not override accuracy requirements
• Internal-only communications — slack messages, code comments, docs; voice standards are customer-facing only
• Deterministic field values (prices, hours, addresses) — these are facts, not voice decisions

Required Inputs

• org_id or brand_slug — to look up stored brand guide
• content — the copy to review or the context to generate within
• channel — email, SMS, social, web, in-app (tone calibration varies by channel)
• Brand guide data (pulled from extractBrandData() or stored brand config)

Reasoning Approach

Every brand has three voice dimensions. Assess all three before writing or reviewing:

1. Personality axis Where does the brand sit?

• Chill & approachable ↔ Professional & authoritative
• Playful & punchy ↔ Warm & nurturing
• Educational & expert ↔ Conversational & peer

2. Language patterns

• Approved words / phrases the brand uses consistently
• Words or phrases the brand explicitly avoids
• Sentence length preference (short punchy vs. flowing descriptive)
• Emoji usage (never / occasional / frequent)

3. Channel calibration | Channel | Tone adjustment | |---------|----------------| | SMS | Shortest, most direct — 160 chars max; urgency-appropriate | | Email | Warmer, more narrative; subject line punchy, body conversational | | Social | Match the brand's existing social personality; hashtag discipline | | In-app chat | Most casual, real-time feel; quickest responses |

BakedBot platform defaults (apply when brand-specific guide is absent):

• Warm, community-forward, never corporate
• Cannabis-literate without being exclusive
• Inclusive — never assumes prior knowledge
• Compliant framing as default — no medical claims embedded in copy even casually

Output Contract

For voice review:

VOICE REVIEW — [Brand] — [Channel] — [Date]

ALIGNMENT: ✅ On-brand / ⚠️ Partial / ❌ Off-brand

FLAGGED PHRASES:
| Original | Issue | Suggested Revision |
|----------|-------|-------------------|

OVERALL NOTES:
[1–2 sentences on what's working and what to adjust]

For voice-guided generation: Produce the requested copy artifact with a one-line note: [Voice applied: warm/approachable/community-forward — adjusted for SMS brevity]

Edge Cases

• No brand guide on file: Use BakedBot platform defaults; note "brand guide not configured — using platform defaults"
• Brand guide conflicts with compliance: Compliance wins. Flag the conflict to the operator.
• Multiple brands in one org: Apply the brand guide for the specific brand_slug in context
• New brand account: Generate a brand guide draft from extractBrandData(url) output; present for operator review

Escalation Rules

• Brand guide missing and content is high-visibility (campaign, launch): Ask operator to confirm voice before proceeding
• Copy passes voice but triggers compliance concern: Route to Deebo before delivery

Compliance Notes

• Brand voice never overrides regulatory requirements
• "Sounds like us" is not a compliance defense
• Medical-adjacent language is always flagged even if it matches brand personality