The challenge
As its e-commerce channel grew, Farby.sk faced a growing volume of customer inquiries and support requests. This reflected a broader retail trend in which higher online sales and expectations of immediate communication put more pressure on customer service teams. Retailers across the sector face challenges in scaling support and finding qualified agents while controlling the cost of maintaining service quality.
More than 65% of incoming tickets were repetitive, low-complexity questions such as “Where is my order?”, “What is the delivery time?”, or “How do I return an item?”. Handling these requests manually slowed response times, increased costs, and limited team capacity.
Key challenges:
-
Long response times of 8–10 minutes and no support outside operating hours.
-
Low customer satisfaction, with CSAT below 75%.
-
High operating costs caused by the need for seasonal agents.
-
Limited scalability as online sales grew quickly.
The solution
Aspecta designed and implemented an AI Smart Assistant using AWS generative AI services to automate routine interactions and provide immediate, accurate, and consistent responses across channels. The approach combined a high degree of automation with limited infrastructure management.
AWS provided the required scalability, security controls, and deployment speed. Amazon Bedrock provides access to state-of-the-art language models without requiring the client to train them directly, which shortened implementation and reduced costs. The serverless architecture using AWS Lambda and API Gateway also provides flexibility for future extensions.
Solution architecture
- Amazon Bedrock: generative models (Claude/Titan) for multilingual, context-aware responses.
- Amazon OpenSearch (Vector Engine): semantic search across FAQs, policies, and order information.
- API Gateway + AWS Lambda: serverless orchestration of requests between Amazon Bedrock, OpenSearch, and ERP/e-commerce systems.
- Amazon DynamoDB: conversation context management and session storage.
- Amazon S3: central storage for documentation, FAQs, and chat logs.
- CloudWatch, GuardDuty, CloudTrail: monitoring, security alerts, and audit records.
Integration with Farby.sk systems
The existing ERP and e-commerce systems remained on their current infrastructure. A secured middleware layer provides controlled access to order, payment, and delivery-status data.
Implementation
The project followed an iterative approach across three phases:
-
Analysis and solution design: identification of the most frequent questions, definition of domain prompts, and mapping of API integrations.
-
AWS architecture deployment: configuration of Amazon Bedrock, OpenSearch, and AWS Lambda orchestration in Multi-AZ mode with a CI/CD pipeline.
-
Pilot and optimization: A/B testing of responses, accuracy improvements, and multilingual support for SK, CZ, and HU.
Challenges addressed
-
ERP integration: limited access to historical data was addressed through proxy API layers.
-
Knowledge base updates: event-driven reindexing was combined with manual approval to support compliance requirements.
-
User adoption: a pilot campaign and direct integration of the chat component into the e-commerce interface supported adoption. Implementation proceeded without major issues.
Using fully managed AWS services and a prepared integration layer allowed the solution to reach production within several weeks. This shortened time-to-value and allowed KPI measurement to begin immediately.
Results and benefits
The AI Smart Assistant delivered measurable results during its first months of operation:
-
60% automation of recurring requests
-
>90% faster response time, from 8 minutes to immediate responses
-
+12 CSAT points in customer satisfaction
-
24/7 availability without expanding the team
-
Lower support costs and less seasonal hiring
The solution allowed online support services to scale without increasing staffing levels and created a foundation for personalized digital assistance.
Conclusion
The AI Smart Assistant for Farby.sk demonstrates how technology and customer experience can work together. The solution showed that generative AI on AWS can reduce operating costs, shorten response times, and improve customer satisfaction without requiring changes to existing core systems.
The project also shows the potential of automation in retail and e-commerce, where high volumes of customer interactions create both operational challenges and opportunities. Similar solutions can help companies scale services, provide more consistent responses, and allow employees to focus on higher-value work.
The Farby.sk case provides a practical example for organizations considering generative AI to improve efficiency, availability, and customer satisfaction as part of digital transformation.
