How to Automate Customer Success with a Chatbot

The Rundown: If you’ve ever had people constantly asking for “that one tutorial” or “the guide on X, about your company’s product/service” you know it can get messy fast. Instead of sending them off to dig through archives, you can give them exactly what they need instantly - through a simple chat interface. Today, I’ll walk you through setting up Chatbase so your users can find the right content or information in seconds.

🧰 Who is this useful for:

Course creators who want students to find resources quickly
SaaS companies needing 24/7 customer support
Agencies that get the same questions over and over
Anyone with a library of content, guides, or documents to make searchable via chat

STEP 1: Getting Started

Head to the Chatbase website. Before building your agent,, check out their pricing.

Free plan: Includes fast models (GPT-4o mini, Claude Haiku, Gemini Flash) - great for quick, cost-effective responses.
Comes with one test agent so you can try it before committing.
Paid plans (Hobby, Standard, Pro) open up all models, increase message credits, and allow for more user-AI conversations.

Once you’ve chosen a plan, click into the Agent creation interface. Think of this as your playground: the place where you build, configure, and test your chatbot.

STEP 2: Add Your Training Data

Here’s where Chatbase learns what it needs to answer questions accurately. You can feed it:

Files (PDF, DOC, TXT) - drag and drop
Raw text you paste in
Websites - either crawl an entire site or add a single link

For my example, I used the Rundown AI University Guides page - short, 5-minute tutorials solving specific problems like “AI presentations” or “SEO workflows.”

I added the direct link so Chatbase would pull only that content. You’ll see the link listed in your interface, with options to edit, delete, or restore it.

STEP 3: Configure and Create Your Agent

Before hitting “Create Agent,” you can tweak:

Q&A pairs for extra context
Notion integration if your data lives there
Model selection - newer models are smarter but slower and cost more credits
Temperature - 0 = precise and predictable; higher numbers = more creative but more “hallucinations”
AI Actions - connect your bot to do real-world tasks (check CRM, update Stripe payments, etc.)
System prompt - guide your bot’s personality, focus, and restrictions
Role restrictions - e.g., only act as a sales rep or lead conversion bot

Now click Create Agent. Chatbase will train your bot on your chosen content (takes a few minutes).

STEP 4: Test Your Agent

Example test prompts I ran:

“Find me the best guides on AI presentations.”
- Returned top matches like "Automate Deep Research and Presentations with ChatGPT Agents" and "Create Self-Updating Slide Decks with Manus" - each linked directly to the resource.
“Show guides for Claude Code.”
- Pulled up "How to Create Claude Code Sub-Agents" and "Build Apps with Open Source Claude Code", skipping irrelevant results.

If something comes back that doesn’t fit, you can give feedback like:

“Exclude this result when users ask for Claude Code guides.”
The agent will remember and improve over time.

STEP 5: Deploy & Refine

Once you’re happy, you can:

Embed your chatbot on your site
Give it to your customer support team
Keep refining its responses based on real user queries

This means users skip the whole “digging through pages” step and get exactly what they need - instantly.

Pro Tip:
Start small with one focused content set, like your most requested guides. Once it works well there, expand to other areas. Chatbase improves with feedback, so keep training it just like you would a team member.