Knowledge Bases

The four searchable knowledge libraries that ground the AI Copilot — sales playbooks, product info, support articles, competitive intel.

Knowledge Bases

A knowledge base is a searchable library of documents and articles the AI Copilot reads from when answering questions. HotCRM ships with four built-in knowledge bases, each tuned for a specific kind of question.

This is the layer that makes the Copilot's answers grounded — every fact, talking point, or article suggestion is traceable to a source document.

The four built-in knowledge bases

Knowledge baseUsed byWhat's in it
📘 Sales KnowledgeSales CopilotPlaybooks, objection handling, qualification frameworks, win/loss notes
📗 Product InformationSales + ServiceProduct datasheets, feature specs, release notes, FAQs
🎧 Support KnowledgeService CopilotKB articles, troubleshooting guides, runbooks
⚔️ Competitive IntelSalesCompetitor profiles, battle cards, win/loss analyses

How they work (in plain terms)

Each knowledge base is a library of documents that the Copilot can search semantically — meaning it finds documents by meaning, not just keyword. When the Copilot needs to answer a question, it:

  1. Searches the relevant knowledge bases for matching content.
  2. Picks the top 3-5 most relevant passages.
  3. Grounds its answer in those passages and cites them in the response.

If the knowledge base doesn't have a good match, the Copilot says so rather than guessing. This is the "I don't know, but here's what I found" behavior — far safer than hallucination.

📘 Sales Knowledge

The sales rep's playbook, in searchable form.

Typical contents:

  • Discovery-call frameworks (BANT, MEDDIC, SPIN).
  • Objection-handling cheat sheets.
  • Email templates by stage and persona.
  • Deal-stage exit criteria.
  • Forecast definitions (commit vs best case vs pipeline).
  • Win-stories — examples of successful deals to mirror.

Used by:

  • Lead Qualification skill — to score on the agreed framework.
  • Email Drafting skill — to match your company's voice.
  • Revenue Forecasting skill — for the methodology.

📗 Product Information

Everything about what you sell. Read by both sales and service copilots.

Typical contents:

  • Product datasheets (auto-pulled from the product catalog).
  • Feature specifications.
  • Pricing and packaging.
  • Release notes and changelogs.
  • Integration guides.
  • Common how-to articles.

Used by:

  • Email Drafting — to insert accurate product details into customer emails.
  • Case Triage — to find the right product context for a case.
  • Customer 360 — to summarise what a customer owns.

🎧 Support Knowledge

The service team's article library.

Typical contents:

  • Troubleshooting articles ("Symptom → Cause → Fix").
  • Step-by-step setup guides.
  • Error-code explanations.
  • Known issues with workarounds.
  • Internal runbooks (private to staff).

Used by:

  • Case Triage skill — to recommend the top 3 matching articles.
  • Email Drafting skill — to ground customer-facing fixes in approved guidance.

See also: Service › Knowledge Base for how agents read, write, and rate these articles.

⚔️ Competitive Intel

Battle cards and competitor research — typically maintained by product marketing.

Typical contents:

  • Competitor profiles (positioning, pricing, target market).
  • Feature-by-feature comparisons.
  • Battle cards by competitor (strengths, weaknesses, talk-tracks).
  • Win/loss analysis themes.
  • Quotes from won deals against each competitor.

Used by:

  • Email Drafting skill — when a customer mentions a competitor.
  • Lead Qualification skill — when a lead's source suggests they're evaluating alternatives.

How content gets into a knowledge base

Three ways:

  1. Manual upload — admins or knowledge managers upload PDFs, Word docs, Markdown.
  2. Sync from objects — product datasheets sync automatically from the product catalog; KB articles sync from the Knowledge Base object.
  3. External integrations — pull from Confluence, Notion, Google Drive, SharePoint (via connectors).

How content stays fresh

  • Auto re-indexing runs nightly to pick up edits.
  • Manual re-index button in the admin console for urgent updates.
  • Version tracking — the Copilot always cites the latest version.

What every knowledge base respects

  • Permissions — internal-only articles aren't surfaced to customer-facing answers.
  • Recency — newer documents are weighted higher than stale ones.
  • Source citations — every answer links back to the originating document so the user can verify.

Tips for admins

  • Audit the citations — at least monthly, review what articles the Copilot is citing. Frequently-cited articles deserve extra polish; never-cited articles may be dead weight.
  • Watch the "no result" log — questions the Copilot couldn't answer are gold for prioritising new content.
  • Tag content with audience (sales / service / customer-facing / internal) so the Copilot picks the right voice.

Tips for knowledge managers

  • ✅ Keep articles short and scannable — the Copilot extracts passages, not whole documents.
  • ✅ Include explicit symptoms ("Error 503", "Cannot login") at the top — improves match quality.
  • Date your release-note articles clearly — the Copilot uses dates to prefer current info.

See Customization › AI Skills for how to wire a new skill to a knowledge base.

On this page