Concepts
Source material
Source material is the set of files, tickets, notes, transcripts, exports, and pasted excerpts from one bounded team or function corpus. It is the evidence Company Brain uses to draft operational knowledge.
You should know what counts as useful source material and which file types have limited parsing today.
Useful source types
- Resolved support tickets showing the customer question and agent response.
- Support macros, saved replies, and help-center drafts.
- Slack or Teams exports where teammates explain the real process.
- CRM account notes, implementation notes, and onboarding handoff records.
- VTT or SRT call transcripts.
- Existing SOP drafts, internal FAQs, and policy notes.
Recommended source count
Bring enough relevant files, tickets, notes, transcripts, exports, or pasted excerpts to show repeated patterns. A good first corpus is bounded by team, queue, operating area, and reviewer; broad exports blur the scope unless they are split first.
Current parsing limits
- Text extraction is browser-readable for TXT, Markdown, CSV, JSON, HTML, VTT, SRT, text-layer PDFs, readable DOCX files, and supported text-like entries inside ZIP uploads.
- Scanned or image-only PDFs are not OCRed; paste readable text or upload a text export for those files.
- Empty, protected, or image-only DOCX files fail clearly; paste readable text or upload a text export for those files.
- ZIP parsing is limited to supported text-like entries; nested archives, unsupported files, unsafe paths, over-limit entries, corrupt archives, encrypted archives, and password-protected archives are skipped or failed with visible copy.
- Unsupported or failed files must be removed before extraction.
Do not assume unsupported, failed, skipped, or metadata-only files were parsed. Paste readable text for the important parts.
Read next
Prepare source filesGather relevant files, tickets, notes, transcripts, exports, or pasted excerpts from one bounded team corpus before uploading them into Company Brain.Data and securityCurrent Company Brain trial data handling, security boundaries, source-material tiers, and restricted-material expectations.