The customer chose to protect their confidential information
Multi-channel chatbot for a bank

About the client
In this business case, Andersen's customer was a Kazakhstan-based bank that is particularly active in the mortgage market.
![[object Object] on the map](https://content.andersenlab.com/common-data/client-maps/kazakhstan/1741687556.png)
About the project
The customer approached Andersen with a request to develop and deploy a scalable AI-driven chatbot that would enhance customer interactions across digital platforms. The chatbot was required to support multi-channel intent recognition for RAG databases, integrating AWS AI solutions for text recognition and document search (for both external and internal use).

Challenges
The cost-effective and modern solution suggested and implemented by Andersen resolved the following challenges:
- Optimizing LLMs through fine-tuning, prompt engineering, and domain-specific data to leverage their powerful chatbot capabilities;
- Ensuring multi-channel support to guarantee outstanding user experiences and seamless interactions across platforms;
- Achieving cost-efficiency by balancing performance with cost, depending on query volume and AI model selection.
Solution
Andersen engineered and deployed an AI chatbot for integration into the bank’s digital channels. The solution incorporates LLMs (Anthropic's Claude, Cohere, and HuggingFace embedding models) and enables building REST APIs for text queries and RAG database management. Our team designed a storage structure that was optimized for multi-channel adaptability. Additionally, we enhanced the chatbot's API with prompt engineering to improve content retrieval. The solution is built on a robust AWS foundation and uses, among other services:
- Amazon Bedrock to build and scale generative AI applications with foundation models;
- Amazon API Gateway to manage secure API requests, ensuring seamless interactions between the chatbot and external applications;
- AWS Lambda for serverless execution of chatbot logic, handling real-time text processing and query routing;
- AWS CloudFormation to automate infrastructure provisioning, a critical requirement for scalable and efficient chatbot deployment;
- Amazon OpenSearch Service to enable fast and relevant search queries, enhancing the chatbot’s ability to retrieve information from large datasets.
App functionality
Delivered functionalities include:
- Automated customer support;
- Personalized assistance;
- Intelligent document search;
- Smart loan advisory;
- Integration with human agents.
Project results
Outcomes include:
- High query processing volume: approximately 2 million queries were processed in the first month;
- Enhanced response accuracy: chatbot interactions were optimized across multiple digital channels;
- Robust security and compliance: customer data is protected via AWS KMS, AWS Secrets Manager, and full PCI DSS compliance;
- Accelerated deployment: rollout times were reduced from days to hours thanks to CI/CD pipelines and AWS CloudFormation;
- Increased user engagement thanks to the flawless multilingual support (English, Kazakh, and Russian are available);
- Monthly infrastructure costs are consistent even when traffic increases;
- Pay-as-you-go and auto-scaling options have lowered expenses by about 40% compared to on-premises solutions.
About Andersen
As an official AWS partner, Andersen employs a team of over 50 AWS-certified engineers, specializing in cloud-native development and seamless software migrations. With over a decade of experience in AWS consulting and application development, Andersen has successfully delivered high-performance, secure, and scalable cloud solutions to clients worldwide.
Let's talk about your IT needs
What happens next?
An expert contacts you after having analyzed your requirements;
If needed, we sign an NDA to ensure the highest privacy level;
We submit a comprehensive project proposal with estimates, timelines, CVs, etc.
Customers who trust us