Guides · 06

Guides by task

Start with the outcome you want to achieve.

Bring your brand guidelines into AI

The goal is for AI to use reviewed knowledge, beyond receiving an attachment.

  1. Import current guidelines into the correct brand.
  2. Review products, customers, voice and boundaries in the profile.
  3. Fill gaps or correct extraction errors before using it.
  4. Connect your AI tool and try the read-only prompt in Quick install.
  5. Compare the answer with sources and record inaccuracies to correct in the profile.

The useful outcome is an accurate profile and an answer traceable to its sources, rather than a count of imported pages.

Write content for a new product

Prepare confirmed product details: name, audience, features, evidence, conditions of use and prohibited claims. If the product is absent from the profile, distinguish new draft inputs from approved changes to the shared profile.

Read the brand profile and draft a launch brief for [product], for [audience] on [channel]. Confirmed product details: [data]. List missing information first and do not invent claims. Write the content draft only after I approve the brief.

Brand Agent in the web app has brief and content drafting tools. An external AI tool can draft from the profile it reads, but through MCP it can only read briefs and strategy, not sign them off. Review the brief first, then the draft's voice, evidence and boundaries. Update the profile with appropriate permissions when product information should be reused, rather than leaving it in one conversation.

Review a campaign idea

Open Brand Agent in the web app: scored idea review runs there, not through MCP. Provide the idea, goal, audience, channel and execution conditions. Ask AI to compare it with the profile and identify missing evidence needed to evaluate it.

Idea: [description]. Goal: [goal]. Audience: [audience]. Review it against the brand profile, separate boundaries from evaluation dimensions, identify two things to verify and suggest one improvement that preserves the main idea.

Read each dimension instead of seeking one total score. Suggested improvements are options to discuss; the responsible person still decides budget, execution and publishing. Record experiments to learn from real feedback. In an external AI tool, antiem_check_brand_content can check a finished draft against the brand's boundaries without using a model.

Hand off brand knowledge to collaborators

Before inviting someone, decide whether they need to read, create content or edit the profile. An owner or admin sends invitations: open Cài đặt (Settings, in the user menu at the bottom of the sidebar), then your brand's section and Thành viên (Members). You can invite someone as an admin or a member. Share the relevant brand only.

Give collaborators the task's goal, where to read the profile and what they must review. For collaborators using AI, share connection instructions and ask them to choose the right brand before each task.

Check the handoff with a small task, such as summarizing the voice with sources. Review member permissions and connections when collaboration ends, beyond deleting a link from chat.

Manage multiple client brands

Separate documents, profiles and tasks by brand. Use names that distinguish clients for both people and AI, especially when an account belongs to multiple brands.

Check the brand name before importing or writing in the app. In external AI tools, name the brand explicitly and ask for a fresh profile read. Use separate conversations for sensitive client work to avoid reusing another client's context.

Server permissions protect access; organizing conversations helps people and AI use information for the right purpose. Switching brands on the web does not clear an external AI conversation's context. A connected AI tool can reach every brand your account can access, so name the brand every time.