Support chatbot
Prototype

Support Chatbot

ActiveTickets

Direct answers with source references, pragmatically built in just weeks

// OVERVIEW

The challenge

ActiveTickets has an extensive support portal with hundreds of articles. The search function works, but customers still need to scroll through results and find the right information themselves.

A chatbot that provides direct answers, including a link to the source article, would improve the support experience. But a "proper" API integration with the knowledge base? The development team has a considerable backlog with other priorities.

// APPROACH

Pragmatic building

Due to limited development capacity, we chose a pragmatic approach: scrape the support portal and use that content as the knowledge base. Works great for a prototype. If it goes to production, a real-time integration will follow.

Step 1

Scrape

Scraped support portal: collected all articles with URLs.

Step 2

Process

Content written to separate text files with corresponding URL.

Step 3

Index

Files loaded into OpenAI library for vector search.

Step 4

Build

Front-end and API integration built with ChatGPT and Claude Code.

// DEMO

How it works

Ask a question in natural language. The chatbot searches the knowledge base, formulates an answer, and provides a direct link to the relevant support article.

No menus, no search results to scroll through. Just an answer with the source included.

Example question:
"How can I request a refund?"
Answer:
To request a refund, you can... [explanation]
→ [support link]
// STATUS

Working prototype

What works

  • ✓Answering questions based on the knowledge base
  • ✓Source references with direct links
  • ✓Natural conversation in Dutch
  • ✓Logging of all questions for analysis

Still needed for production

  • ○Anonymizing incoming data
  • ○Integration with existing support environment
  • ○Performance optimization for scale
  • ○Monitoring and alerting
Data collection

Web scraping of support portal

AI Platform

OpenAI GPT with file search / RAG

Development

Front-end and logic with ChatGPT + Claude Code

Knowledge base

Text files with URL references

// LEARNINGS

Prove it first

Don't wait for the ideal situation. A "proper" API integration could take months. By scraping, we had a working prototype within days that proved the concept's value. That pragmatic approach makes the difference between "interesting idea" and "working demo".

// MORE_CASES

Check out these cases

Snuffl
CASE

Snuffl.app

From breed recognition to dog hub with AI-driven development

Museon-Omniversum
CASE

Museon-Omniversum

AI phone assistant for visitor questions during peak times

// CONTACT

Improve support with AI?

Curious what an AI chatbot could mean for your support environment? We'd love to build a prototype to explore the possibilities.