Companion videos

Design an AI customer support agent: triage, knowledge, actions, and escalation — companion videos

The article explains how to derive an automation target from your own support queue, then design triage, knowledge retrieval, action-taking, evaluation, and clean escalation around that evidence. These videos provide build-from-scratch examples. Their demonstrations are implementation references, not proof that the same resolution rate or safety profile will transfer to your queue.

Primary pick

3:50:39

How to Build & Sell AI Agents: Ultimate Beginner's Guide

Liam Ottley

The course covers RAG-grounded chatbots, escalation logic, and a no-code stack built with tools such as Botpress, Voiceflow, Make, and n8n. Its "anatomy of an agent" section gives a concrete implementation to compare with the article's triage → evidence → proposed action → policy gate → handoff responsibilities.

What you should get from this: See how a support agent combines retrieval, escalation and workflow tools before production hardening.

Watch or know first: No coding needed — the stack is no-code; just budget for the near-four-hour runtime.

AI Expert note: It's a build-and-sell agency course, so skip the selling chapters if you only want the build — the 'anatomy of an agent' section is the piece that maps onto the article.

Open video page

Also worth watching

31:19

How To Build An AI Customer Support Agent with n8n (free template)

Bart Slodyczka

The n8n workflow connects to help-desk products, drafts replies from retrieved material, and turns solved tickets into candidate knowledge-base content. Before using that ingestion pattern, add human approval, redact personal data and secrets, preserve the source ticket and content version as provenance, and run retrieval evaluations against a held-out ticket set.

What you should get from this: See how an n8n support workflow drafts retrieval-backed replies and proposes knowledge-base updates, then identify the controls needed before production use.

Watch or know first: Basic n8n familiarity and access to a helpdesk like Zendesk, Gorgias or Freshdesk.

AI Expert note: Kept despite the small channel because it exposes a useful ingestion decision for review: solved tickets should become candidate material, not trusted knowledge, until an authorised reviewer approves a redacted, sourced, versioned entry and retrieval tests pass.

Open video page