Fern AI Assistant

AILLMCustomer Support
Year
2025
Duration
4 months
Team
3 engineers

An AI-powered customer support agent for a fintech company. Built with RAG, vector databases, and custom fine-tuned models. Reduced ticket volume by 60%.

The challenge

A fintech client was drowning in support tickets. 70% of questions were repetitive (password resets, transaction status, fee explanations), but the team couldn't just use a chatbot — customers needed accurate, compliant answers.

Our solution

We built a RAG system with a vector database of their knowledge base, fine-tuned a model on their tone and compliance requirements, and added guardrails to refuse to answer anything outside scope.

Results

  • ✓ 60% reduction in ticket volume
  • ✓ 94% accuracy on compliance questions
  • ✓ Average response time: 8 seconds
  • ✓ Customer satisfaction up 22%

Tech stack

PythonOpenAIPineconeFastAPILangChainDocker

Have a similar project?

We'd love to hear about it. Let's talk.

🌿
Thank you!
We'll be in touch soon.