Guides
How to prepare artifacts
Good artifacts preserve the operational pattern while removing unnecessary sensitive data. Messy source material is fine. Irrelevant or unsafe source material is not.
You should know what to gather, what to redact, and how to handle unsupported file types.
Gather a tight source set
- 1
Collect examples from one issue
Pick tickets, threads, macros, call snippets, CRM notes, or docs that all point to the same recurring problem.
- 2
Keep the useful mess
Include contradictions, old guidance, escalation notes, and examples where the team handled the issue differently.
- 3
Remove unrelated material
Do not upload broad exports full of unrelated customers or unrelated processes just because they are easy to export.
Redact before upload
- Remove customer names, personal names when unnecessary, emails, phone numbers, addresses, payment data, account IDs, invoice IDs, tokens, API keys, private URLs, and regulated personal data.
- Use placeholders like [CUSTOMER_EMAIL], [ACCOUNT_ID], [COMPANY_A], [INVOICE_ID], [BUG_ID], [ACCOUNT_OWNER], or [CS_OWNER].
- Keep safe operational context such as issue category, plan tier, rough date, policy version, ticket status, resolution status, source type, and sanitized quote.
When a file cannot be parsed
If a PDF has no readable text layer, or if a DOCX or ZIP is accepted as metadata-only, paste the important text into the source material box or provide a text-like export. Company Brain should not pretend unsupported file types were parsed.