Steps to build a custom brand voice bot
This resource is based on Your brand voice guide is your best AI training data, featuring Samantha Becker of SAB Creative & Consulting, published on the AI Lab by ActiveCampaign.

Get the quick-start guide
Before you start
Make sure you:
- Identify your AI platform: Becker uses ChatGPT Enterprise, but this workflow is platform-agnostic (Claude, Gemini, or any custom GPT builder works)
- Confirm brand guidelines exist: you need at least a style guide, brand voice description, or examples of on-brand content
- Gather access to past content: social posts, articles, emails, or transcripts that represent your organization’s voice
The workflow
Phase 1: Compile your knowledge base
- Collect style guides and brand position documents: include voice attributes, tone descriptors, and any “do/don’t” writing rules
- Pull transcripts from leadership: talks, interviews, or presentations that capture how key leaders communicate
- Gather exemplary content: past social posts, owned media articles, and emails that nail the organizational voice
- Include detailed style guidelines: formatting rules, preferred terminology, audience-specific messaging
- Convert all materials to machine-readable formats: Word docs, PDFs, or plain text files
Phase 2: Consolidate strategically
- Combine all voice attributes and brand positions into a single document: AI platforms limit file uploads, so fewer, denser files work better
- Make the document accessible to stakeholders: store it where team members can review and suggest changes
- Add individual voice profiles: if relevant, include writing samples and style notes for specific leaders or spokespersons
Phase 3: Feed materials into your platform
- Upload consolidated documents to your AI workspace: in ChatGPT, this means creating a custom GPT and adding files to its knowledge base
- Write clear system instructions: tell the bot its role (e.g., “You are a communications assistant for [organization]. Use only the uploaded brand guidelines and voice documents to generate content.”)
- Test with a basic prompt: ask for a LinkedIn post or email draft and compare against the same prompt in generic ChatGPT
Phase 4: Train for individual voices
- Add voice samples for specific leaders: upload writing examples from executives or key communicators
- Test voice differentiation: prompt the bot to write in one leader’s style, then another, and verify the outputs sound distinct
- Share access with the team: give communicators across departments access to the bot so they can generate on-brand drafts
- Share access with the team: give communicators across departments access to the bot so they can generate on-brand drafts
Quick reference
- Total time: 2–4 hours for initial setup (ongoing maintenance: 30 min/month)
- Tools needed: ChatGPT Enterprise (or any custom GPT platform), existing brand/voice documents
- Key output: A custom AI bot that generates on-brand content matching your organization’s voice — and individual leaders’ voices
Ready for the full story?
Read Your brand voice guide is your best AI training data, featuring Samantha Becker of SAB Creative & Consulting, published on the AI Lab by ActiveCampaign.
Related
More data from the AI Lab.