Product docs

Knowledge base documentation

The Zynli knowledge base is the documentation layer that grounds AI support answers in your actual policies, product info, and help content. This page explains how KB content helps AI answer support questions, what types of content belong in KB, how KB connects with workflows, and when to escalate instead of relying on KB alone.

Book a demoKB and ticketing overview

How KB content helps AI answer support questions

Zynli uses RAG (retrieval-augmented generation) to find relevant passages from your knowledge base and generate grounded answers. This reduces hallucination and keeps responses aligned with your actual policies and product documentation.

  • Website crawl: index your help center, policy pages, and product docs
  • File uploads: PDFs, docs, and spreadsheets with your support content
  • Google Drive: connect shared folders for live doc sync
  • Citations: AI answers include source references agents can verify

What types of content belong in KB

Return and refund policies

Return windows, eligibility rules, refund timelines, exchange procedures, and restocking fees.

Shipping and fulfillment

Processing times, carrier options, international shipping, delay expectations, and SLA commitments.

Product documentation

Sizing guides, compatibility charts, ingredient lists, care instructions, and feature specs.

Billing and payment

Payment methods, charge descriptions, subscription terms, and invoice request procedures.

Account and order help

How to update addresses, cancel orders, track packages, and contact support.

How KB connects with support workflows

Knowledge base and workflows work together — not as separate tools. KB answers policy questions; workflows fetch live data; both can appear in the same conversation.

KB-first for policy questions

"What is your return policy?" — answered directly from KB without an API call.

KB + API for mixed questions

"Can I return order #12345?" — KB explains policy, workflow verifies order date via API.

Workflow fallback to KB

When API lookup fails, workflow branches to KB processing-time guidance before escalating.

When KB alone is not enough

Documentation cannot answer questions that require live system data or human judgment. Know when to route beyond KB.

  • "Where is my order?" needs live order/tracking data via API workflow
  • Billing disputes need human review even when policy is documented
  • Damaged or lost shipments need agent investigation beyond KB SLAs
  • Edge cases outside documented policy should escalate, not guess

How unclear cases should escalate

When KB retrieval confidence is low, the question is ambiguous, or the customer disputes the AI answer, Zynli should escalate to a human agent with the KB citations attempted and full conversation context.

  • Configure low-confidence thresholds for automatic escalation
  • Agents see which KB sources were used before taking over
  • Tickets inherit the chat thread and KB citation history
  • Use escalation to improve KB content based on recurring gaps

From KB gaps to Content Opportunities

Recurring questions your KB cannot answer confidently become content opportunities. Zynli Content Opportunities (Pro+) clusters those questions from widget conversations, scores KB coverage, and generates SEO/AEO briefs — so support data drives documentation improvements.

  • Analyze visitor messages over a configurable date range
  • See gap labels: none, weak, partial, or strong per cluster
  • Generate briefs with title, outline, FAQ, and direct-answer blocks
  • Scope is tenant conversations and KB — not Zynli marketing pages

KB vs FAQ-only chatbot

FAQ-only chatbot

Matches keywords to static FAQ entries; cannot combine with live data or create tickets.

Zynli knowledge base

RAG over your full docs, connected to workflows for live data and handoff for exceptions.

Support workflows you can automate

Configure AI support for orders, shipping, returns, billing, and product questions — grounded in your knowledge base and connected to your APIs.

Order Agent

Order status, delivery updates, and order lookup.

Example questions

  • · Where is my order?
  • · Has my order shipped?
  • · Can I change my delivery address?

Example actions

  • · Check order status
  • · Fetch tracking number
  • · Escalate delayed orders

Billing Agent

Charges, invoices, payment methods, and billing disputes.

Example questions

  • · Why was I charged twice?
  • · Can I update my payment method?
  • · Where is my invoice?

Example actions

  • · Look up charge details via API
  • · Explain billing policy
  • · Escalate billing disputes

Returns Agent

Return eligibility, refund status, and exchange requests.

Example questions

  • · How do I return this item?
  • · When will I get my refund?
  • · Can I exchange for a different size?

Example actions

  • · Apply return policy from KB
  • · Check refund status via API
  • · Open return ticket for review

AI resolves most issues. Agents handle the rest.

When automation can't resolve an issue — damaged goods, billing disputes, or policy exceptions — escalate to a human agent with full thread history and optional ticket creation.

  • Low-confidence answers trigger handoff automatically
  • Agents see chat history, citations, and workflow run logs
  • Escalations become numbered tickets on your kanban board
  • Customers track progress via magic-link portal

Visitor

My package arrived damaged. I need a replacement.

AI workflow

Connecting you with an agent — your order details and photos are included.

Human agent

Replacement arranged. Ticket created for tracking.

Frequently asked questions

What KB sources does Zynli support?

Website crawl, file uploads (PDF, docs), and Google Drive folder connection. Content is indexed for RAG retrieval by AI agents.

Can KB answer order status questions?

KB can explain processing times and shipping SLAs, but live order status requires an API workflow. KB and workflows complement each other.

How do I keep KB answers accurate?

Update your source docs when policies change. Re-crawl or re-upload content. Review escalated conversations to find KB gaps.

Does KB work without workflows?

Yes. KB-powered answers work on all plans. Workflows add live data resolution on Pro.

Can agents edit KB content?

KB content is managed via source connections in the dashboard — crawl targets, file uploads, and Drive folders. Agents focus on conversations and tickets.

See how Zynli automates your support

Train AI on your knowledge base, connect your APIs, and resolve order, shipping, billing, and return issues in chat.

Book a demoKB and ticketing overview

Product documentation

Resources

Content Opportunities docs

Close KB gaps with support-driven SEO/AEO briefs.

Learn more

Knowledge base and ticketing with AI

Platform overview for KB, inbox, and ticketing.

Learn more

Support workflows docs

How KB connects with API workflow actions.

Learn more

Human handoff docs

When KB alone is not enough — escalation paths.

Learn more

How Zynli works

KB as the first layer in the support resolution flow.

Learn more

AI product support

Product documentation in KB for sizing and compatibility.

Learn more