Use cases

Internal knowledge base AI search

Let ops, sales, and engineering ask questions over handbooks and runbooks — with citations so answers are auditable.

7 min read

Internal wikis rot. Slack pins disappear. New hires ask the same questions in #general every Monday. Internal knowledge base AI search gives employees one place to ask natural-language questions — with page-level citations back to Notion exports, Google Doc PDFs, and runbooks.

What to index first

Prioritize docs with high reread cost:

  • Security and access policies
  • Incident runbooks
  • Sales battlecards and pricing guardrails
  • Engineering architecture decision records (ADRs)
  • HR policies employees cite incorrectly

Skip ephemeral Slack threads unless exported to Markdown with context.

Project structure

Use separate Oprag projects by sensitivity:

  • Company-wide — benefits, holidays, office info
  • Engineering — runbooks, on-call guides
  • GTM — competitive intel (restrict dashboard access)

Isolation on Oprag’s AWS is per project — align projects with your permission model.

Rollout inside the company

Phase 1 — Champions (2 weeks)

  • 10 power users in support and ops
  • Golden internal questions in dashboard
  • Fix doc gaps before company-wide launch

Phase 2 — Slack or intranet embed

Phase 3 — API in internal tools

Wire /v1/chat into:

  • /help slash command in Slack (via your bot)
  • Internal Chrome extension
  • On-call dashboard sidebar

REST integration guide covers server-side proxy patterns.

Trust mechanics for employees

Internal users trust citations more than polished prose. Require:

  • Visible source doc name and page
  • “Last indexed” note in your UI (from upload timestamps)
  • Feedback button that routes to doc owners, not IT

Keep content fresh

Assign doc owners per section. Monthly:

  • Re-upload changed policies
  • Re-run golden questions (validation checklist)
  • Archive projects for deprecated products

Security note

Internal search is still sensitive — treat API keys like production secrets. Data stays on Oprag’s AWS with tenant isolation; it is not deployed into your corporate AWS account unless you separately choose that integration style.

Measure impact

  • Time-to-answer for new hires (survey)
  • Repeat questions in #ask-* channels
  • Citation clicks on runbooks during incidents

Ready to try it in your workflow?