Skip to main content

Automate support with your knowledge

A large share of support tickets ask something your organization has already answered somewhere: in the documentation, in past tickets, in internal wikis, in Slack threads. Those tickets do not need a human, they need the answer found. With that knowledge indexed into one knowledge base, you can automate in shades: resolve routine tickets entirely automatically, deflect them before they are even filed, or draft grounded replies for a human to review and send, faster than writing from scratch. Your team keeps the questions that genuinely need them.

Ways to build this with Kapa​

  • Deflect tickets before they are filed with the Support Form Deflector: it answers on your support form while the user types, and only tickets the answer does not resolve get submitted.
  • Plug Kapa into your support platform's native agent over MCP: platforms like Intercom and Front let their built-in AI agents use external tools, so your knowledge grounds the replies they draft.
  • Build your own automation outside your support platform. In a workflow tool like n8n, react to incoming tickets however you want, with your knowledge available at every step through the hosted MCP server or the Retrieval endpoint.
  • Draft replies for your team to review. Wire Kapa into your support platform's own workflow tools so every new conversation arrives with a knowledge-grounded draft; the Resources below work this out for Intercom Fin, Front, and Zendesk. In Zendesk, the Zendesk Agent does the same with no workflow building required.

What to index​

What to index depends on who sees the answers. An automation that responds directly to customers with no human in the loop should rely only on your public sources, such as your documentation site via Website Crawl. With a human reviewing drafts before they are sent, your private knowledge comes into play, and it is often where the best answers live: past tickets via Zendesk Tickets or Salesforce Cases, and internal knowledge in Slack, Notion, or Confluence. Keep private sources in an internal project. Start with Index your first source.

Resources​