Cavaridge Academy
Cavaridge AI for Service Teams
Module 4 of 5

Content generation with brand voice

Generate customer-facing content that respects brand chrome, tone, and accuracy guardrails.

Video — pending production
Read the transcript below. Once recording is complete, the video will replace this notice.
--- title: Content generation with brand voice status: draft note: AI-generated first-pass transcript pending video production + SME review. --- Content generation is what most people think of when they think of generative AI: emails, summaries, customer-facing reports. The platform makes this a controlled surface — brand chrome, citations, audit trail. This lesson is how to use it well. ## Brand chrome is automatic Every Cavaridge AI generation pulls brand tokens, footer rules, and tone guidance from `tenantConfig`. You don't paste branding into the prompt; the platform applies it at render. This means: - Customer-facing outputs include the right logo, footer, and disclaimers for the tenant context. - For tenanted MSP→client work, the **client's** branding renders, not the MSP's. - White-label tenants get their own brand — never Cavaridge™'s. Don't fight this by trying to override branding in the prompt. The gateway strips override attempts and warns you. ## The three-line discipline When generating customer-facing content, every output needs: 1. **A subject + summary** that an editor can scan in 5 seconds. 2. **The body** with citations on any factual claim. 3. **A "model + run id" footer** so a reviewer can trace provenance. Studio templates enforce this. Use them. ## Confidence flags When Studio is uncertain about a claim, it inserts a confidence flag — a small bracket like `[verify: claim about ContosoCare's last audit date]`. Treat these as todo items for review. Don't ship content with flags still present. The platform's pre-publish gate will block you anyway. ## Tone Tone is a `tenantConfig` setting per audience. Studio knows that the Pulse status page tone is different from a customer security report tone is different from a marketing landing-page tone. If you find yourself fighting the tone, the right answer is usually to change the audience setting at the workspace level — not to override the prompt. ## When to use a template vs free-form - **Template** when the output shape is consistent (security report, weekly summary, partner update). Templates have validators that catch missing sections. - **Free-form** for one-offs, exploratory drafts, internal-only notes. Free-form output should never go directly to a customer. ## Hands-on The **forge-content-brand-pack** seed gives you a tenant with two brands (an MSP and a tenanted client) and three audiences (internal, client-admin, executive). Generate the same content under each context and compare the outputs. You'll see brand chrome and tone shift without you changing the prompt.
Hands-on sandbox
forge · seed: forge-content-brand-pack · 60 min

Knowledge check

  1. Question 1 · select one
    Brand chrome on outputs is enforced via
  2. Question 2 · select one
    When generating customer-facing copy, which is the right escalation if you're uncertain about a claim?
  3. Question 3 · select all that apply
    Which features make Studio outputs auditable for marketing review?