AI-Powered Customer Feedback Analysis for a Fintech App
LLM integration for a better understanding of what customers say.
AI development
Fintech
7 weeks
Web
About the project
Our client was a fintech startup that offered digital banking services across Europe. They had a high volume of user feedback on their mobile apps via app reviews, support tickets, and in-app surveys, but didn’t have enough manual power to analyze it. The company partnered with us to enhance their internal web platform with AI-powered features that would analyze customer feedback from multiple channels, detect sentiment, and provide the team with actionable product and support insights.
The client had
A working web platform for internal use
We were responsible for
Developing and integrating AI-powered sentiment analysis features
Project Team
Project manager
AI engineer
Backend engineer
Frontend engineer
QA engineer
Image
Key features we worked on
Development process
Here is how we created the web tool
Discovery and planning
Requirements gathering.
Mapping feedback sources (App Store, Play Store, Intercom, in-app surveys).
Data integration
Connecting to APIs and exporting historical feedback.
Creating a unified data pipeline to standardize formats.
NLP model development
Sentiment analysis model (positive, negative, neutral).
Topic modeling to cluster feedback by themes.
Dashboards
Real-time dashboard for product managers and support teams.
Weekly reports with AI-generated summaries.
Testing and iteration
Internal testing with labeled datasets.
Continuous learning with human-in-the-loop feedback for edge cases.
Tech stack
The technologies we used to realise the LLM integration smoothly.
Development challenges and solutions
How our team dealt with a range of development challenges.
Unstructured data
Problem: Customer feedback was sometimes vague, inconsistent, and filled with typos, sarcasm, or irrelevant content.
Solution: We used NLP preprocessing and filtered out low-value feedback with the help of heuristics (minimum word count, presence of relevant keywords).
Integration with legacy systems
Problem: The existing internal platform had a rigid architecture and limited API support.
Solution: We used microservices to decouple AI logic from the main app and introduced a REST proxy service.
Result
80% of feedback was classified automatically (topic + sentiment) with high precision.
3x faster identification of critical issues.
22% drop in ticket response time for product-related issues.
