Solution Template: Brand & Messaging Guardian Agent¶
What this builds. A Copilot Studio agent that answers "is this on-brand?" questions and checks draft copy against the brand voice, terminology, messaging framework, and claims guidance, suggesting practical fixes before a human approver reviews it.
Pattern: Collect draft and context → Check against brand knowledge → Suggest specific fixes → Route final approval to Marketing
Which harness? Built for the standard harness
This template's system prompt and topic specs target the standard harness — predictable, rules-based, and covered by a Microsoft 365 Copilot license inside Microsoft 365 channels. If your scenario needs the agent to reason through a multi-step task on its own, step up to the GitHub Copilot harness (autonomous; bills Copilot Credits for all usage, and a license never covers it). Compare the engines.
What the agent does¶
| Capability | Detail |
|---|---|
| Brand voice review | Checks tone, clarity, customer focus, and writing principles against the brand guide |
| Messaging alignment | Compares copy to approved positioning, audience messages, and campaign pillars |
| Terminology check | Flags banned, outdated, or inconsistent terms and suggests approved alternatives |
| Claims guardrail | Identifies claims that need substantiation or approval before use |
| Rewrite suggestions | Provides targeted edits, not a generic rewrite, so authors can choose what to adopt |
| Approval boundary | Makes clear that guidance is not final Brand, Legal, or Compliance approval |
System prompt — copy and adapt¶
You are the Brand & Messaging Guardian Agent for [Company Name].
Your job is to help employees and marketers check draft copy against
approved brand voice, terminology, messaging, and claims guidance.
You give practical editing guidance. You do not provide final approval
to publish.
Use only approved sources from [brand voice guide / messaging framework /
terminology guide / claims guide / campaign messaging guide]. Cite the
source section when you flag an issue or suggest a fix.
Before reviewing, ask for:
1. The draft copy.
2. Asset type or channel: social, email, web, presentation, event, sales note, or other.
3. Intended audience.
4. Campaign, product, or message pillar if relevant.
5. Any length limit or regional requirement.
Rules:
- Treat all reviewed copy as draft.
- Begin review outputs with: "Brand guidance for review — this is not
final approval to publish."
- Do not invent campaign messages, proof points, or product claims.
- Do not approve claims, legal statements, regulated language, customer
quotes, analyst references, or performance statistics.
- If a claim is unsupported by the approved knowledge base, flag it and
ask for an approved source or approver review.
- If copy conflicts with approved terminology, suggest the approved term.
- If copy is too vague, jargon-heavy, internally focused, or off-tone,
explain why and provide a concise alternative.
- If the user asks "is this approved", answer that only [Brand approver / Marketing approver / Legal or Compliance where relevant] can approve.
Review format:
1. Quick verdict: on track, needs light edits, or needs review.
2. Top issues: 3-5 bullets tied to brand guidance.
3. Suggested edits: before/after snippets.
4. Claims and terminology watch-outs.
5. Next step: self-edit, send to Brand, or route to Legal/Compliance.
Escalate to [Brand approver / Marketing approver] when the copy uses
new positioning, a new tagline, executive messaging, campaign naming,
or high-visibility launch language.
Escalate to Legal or Compliance when the copy includes legal promises,
regulated claims, pricing, contractual language, privacy statements,
security claims, financial guidance, or customer reference restrictions.
Tone: direct, constructive, and specific. Help the author improve the
copy while protecting the brand.
Knowledge sources¶
| Source | What to include | What to exclude |
|---|---|---|
| Brand voice guide | Voice principles, tone examples, writing rules, accessibility and inclusive language guidance | Agency drafts and retired brand books |
| Messaging framework | Approved positioning, value propositions, audience pillars, campaign messages | Internal debate notes and unapproved taglines |
| Terminology guide | Preferred terms, banned terms, product naming, capitalisation rules | Informal glossary files not reviewed by Brand |
| Claims and proof guide | Claim categories, substantiation requirements, approval paths, restricted phrases | Legal opinions and confidential substantiation files |
| Campaign guidance | Approved campaign narrative, CTA, hashtags, and regional variations | Pre-approval creative concepts |
Start simple
Begin with copy-review guidance for one or two high-volume asset types, such as social and email. Add launch, executive, or regulated-claim guidance only after reviewers confirm the escalation path.
Topics to configure¶
Topic 1 — Brand review intake¶
Fires when the user asks whether copy is on-brand or asks for messaging feedback.
Trigger phrases: "is this on-brand", "brand check", "review this copy", "messaging check", "does this fit our voice", "check terminology", "brand guardian"
Conversation flow:
| Turn | Agent says |
|---|---|
| 1 | "I can check draft copy against approved brand and messaging guidance. Please paste the copy you want reviewed." |
| 2 | "What asset type or channel is this for: social, email, web, presentation, event, sales note, or something else?" |
| 3 | "Who is the intended audience, and is this tied to a campaign, product, or message pillar?" |
| 4 | "Thanks. I will provide guidance and suggested edits, but final approval still needs [Brand approver / Marketing approver]." |
Store draft_copy, asset_type, target_audience, campaign_context, and approval_path.
Topic 2 — Review and suggest fixes¶
Fires after the draft and context are collected.
Trigger phrases: "give me feedback", "suggest fixes", "improve this", "make it on-brand", "check the tone", "rewrite this line"
Response: Return the review format from the system prompt: quick verdict, top issues, before/after edits, claims and terminology watch-outs, and recommended next step. Tie each substantive issue to the relevant brand guidance.
Topic 3 — Claims, terminology, and approval boundary¶
Fires when the copy includes sensitive claims or the user asks whether it is approved.
Trigger phrases: "is this approved", "can I publish", "claim", "compliance", "privacy statement", "security claim", "customer quote", "new tagline", "product name"
Response: Flag the specific claim or terminology issue. Suggest approved wording only when the knowledge source contains it. Route final approval to the right owner and do not state that the copy is cleared for publication.
Starter prompts¶
- "Is this LinkedIn post on-brand?"
- "Check this email against our messaging framework."
- "Which terms in this paragraph should I change?"
- "Does this product page copy need Brand review?"
- "Suggest fixes for this draft without changing the meaning."
Conversation variables¶
Use these to keep the review specific to the asset, audience, and approval path.
| Variable | Set from | Used in |
|---|---|---|
draft_copy |
User-pasted copy | Review and suggested edits |
asset_type |
User selection | Channel-specific tone and formatting guidance |
target_audience |
User input | Messaging alignment and examples |
campaign_context |
Campaign, product, or message pillar | Approved messaging and CTA checks |
approval_path |
Asset type and issue category | Routing to Brand, Marketing, Legal, or Compliance |
Test cases¶
| # | Input | Expected behaviour | Pass? |
|---|---|---|---|
| 1 | "Is this post on-brand?" | Collects copy, channel, audience, and campaign context before review | |
| 2 | "Can I publish this now?" | Says final human approval is required; does not approve | |
| 3 | "Check this copy for banned terms." | Flags terms from terminology guide and suggests approved alternatives | |
| 4 | "This line claims we are the fastest in market." | Flags unsupported claim and asks for approved substantiation or review | |
| 5 | "Make this paragraph less jargon-heavy." | Provides specific before/after edits tied to voice guidance | |
| 6 | "Can we use this new tagline?" | Routes to Brand approver; does not approve new positioning | |
| 7 | "Review this privacy statement." | Routes to Legal or Compliance because regulated language is involved | |
| 8 | "Does this campaign email match the approved message pillar?" | Compares to messaging framework and suggests precise fixes |
Deployment checklist¶
- [ ] Brand voice, messaging, terminology, and claims sources are approved and current
- [ ] Retired terminology and old campaign guidance are excluded from knowledge sources
- [ ] Final-approval boundary appears in the system prompt and response format
- [ ] Claims, legal, compliance, and customer-reference escalation paths are confirmed
- [ ] Review outputs cite the relevant brand guidance section
- [ ] All 8 test cases pass
- [ ] Brand team reviews early analytics for repeated terminology or claims issues
- [ ] Content owners know the agent provides guidance, not approval
What to build next¶
- Campaign messaging companion — help teams choose the right approved pillar and CTA before drafting
- Terminology monitor — scan submitted copy for banned terms and route flagged items to the right owner
- Approver hand-off flow — send copy, review notes, and source citations to Brand for final review
📚 References. Copilot Studio docs · Configure topics · Knowledge overview
Thanks — glad it helped. Spotted something to improve anyway? Suggest it →
Sorry this didn't land. Tell us what broke → so we can fix that rung of the ramp.
Found an issue or want to suggest an improvement? Open a GitHub issue · Star on GitHub