Brand Guardrail Lock (Onboarding)

Brand Guardrail Lock (Onboarding)

1. 01

   Collect the client's locked brand assets before any generation tool is connected

   Pull the current logo files, hex codes, licensed typefaces, and the last two approved campaign decks. Anything the client cannot supply in a reusable format becomes a documented gap, not an assumption.

2. 02

   Write a one-page voice sheet with three banned phrasings the client has rejected before

   Most brand damage from generated copy comes from tone drift, not from wrong facts. Naming the specific phrases a client vetoed gives reviewers a fast pass or fail test.

3. 03

   Load the guardrails into the tool's brand memory or voice profile and screenshot the configuration

   Platforms such as AdMove store voice, colors, fonts, and vocabulary per client in a Brand Hub, while Anyword centralizes brand voice for scoring. The screenshot is the audit artifact if output later drifts.

4. 04

   Run a ten-variant calibration batch and score it against the voice sheet

   Generate across at least two formats (static and short video) so the review covers both visual and textual output. Reject any variant that fails the banned-phrasing test outright.

5. 05

   Have the client sign off on the calibration batch in writing before production volume begins

   A signed batch becomes the reference set for every future QA pass. Without it, disputes about 'off-brand' output turn into taste arguments with no baseline.

6. 06

   Record the guardrail version and date in the client's delivery file

   Brand rules change when a client rebrands or hires a new CMO. Versioning the guardrail set makes it obvious when a retainer deliverable was produced against stale rules.