DataLife
LLM Bot

AI Customer Support Bot

ClientNutriFlow App
IndustryHealth & Nutrition SaaS
Duration4 weeks

Key Result

70% of support queries handled automatically

The Challenge

What needed solving

NutriFlow's customer success team was drowning in repetitive support tickets — subscription billing questions, meal plan FAQs, and app navigation issues made up over 70% of all incoming messages. Response times had ballooned to 18 hours and churn was climbing. Hiring more agents wasn't an option with their runway.

The Solution

How we solved it

I designed and deployed a RAG-based support bot powered by GPT-4o-mini, grounded in NutriFlow's help centre articles and subscription policy docs loaded into ChromaDB. The bot handles tier-1 queries autonomously, escalates to a human when confidence is below threshold, and logs every session to Supabase for ongoing fine-tuning. A Slack webhook notifies the support lead whenever the bot escalates.

Results

The numbers tell the story

70%

Queries resolved without human touch

2 min

Average first response time (was 18 hrs)

41%

Reduction in support ticket volume

4.6/5

User satisfaction score post-launch

Tech Stack

Tools used on this project

LangChainGPT-4o-miniChromaDBSupabaseVercel AI SDKNext.js
We went from drowning in tickets to handling the majority of queries on autopilot within days of launch. The bot is eerily good — users often don't realise they're not talking to a human.

Anika Mensah

CEO, NutriFlow App

Ready for results like these?

Tell me about your project and I'll come back with a clear scope and quote within 24 hours.

Get a Free QuoteView All Case Studies