Our team will reach out within 24 hours to gather your project requirements, clarify your business objectives, and outline the next steps in our collaboration.
Client
Studytube is the #1 corporate learning platform for organizations and employees that helps maximize time savings in organizational learning.
Challenge
The client’s customer service team had to search for answers in the knowledge base manually, spending lots of time and manual effort. Studytube needed a way to automate the search process and improve the overall response speed of the team to improve both operational efficiency and customer satisfaction.
Solution
DataArt developed a GenAI-powered RAG solution to help Studytube customer service agents automate replies to certain customer inquiries. The tool drafts answers to specific customer questions, taking into account data stored in the FAQs and other documents available. It generates the most optimal response on the basis of given prompts and response parameters with the help of Claude 3, while also providing information about documents used as reference. Support agents can edit the suggested answer before sending.
Attaining high accuracy and usability of quoting the source documents was the main challenge. The model also had to be trained to tell the user when it does not have the answer, or the information is absent. Prompt engineering was applied to overcome the challenge – response examples and a set of instructions were provided to the model so that it can draft the most relevant response.
Technologies
Bedrock
Lambda
OpenSearch
ECR
ECS
DynamoDB
API Gateway
S3
Kinesis Firehouse
Step Functions
Outcomes
- Achieved 90% accuracy of the generated responses
- Created a workflow so responses can be sent to any relevant channel (email, Slack, etc.)
- Enabled responses to be generated in different languages, even though the knowledge base is in Dutch-only
- Delivered responses in 5 seconds or less, greatly reducing previous response times
