Customer service
Call centers are in demand nowadays to reduce costs and centralize operations/customer support/service. Due to this, is important to guarantee that customers are happy and that what led to initial contact was solved. Due to the number of calls received daily, it’s impossible to listen and analyze customer feedback, sentiment analysis, call reason and result of call. Additional insights like call duration, waiting time , overall interactions could be added to improve customer service
Challenge
There are many challenges that make this an interest use case:
- Volume of Data
- Complexity of Analysis
- Integration and Real-Time Processing
- Accuracy and Reliability
- Privacy and Data Security
- Scalability
- Customization and Learning
- Multilingual Support
Approach
Harvest information from transactional systems against data lake:
- 1. After each call, a new audio is generated and recorded in call center app (Talkdesk)
- 2. Using an API, the new call is pulled into a storage
- 3. Using Speech Recognition Service, audio is translated into text
- 4. Using ChatGPT and given context, data is anonymized and features are extracted (eg: call context, customer feedback, customer sentiment analysis)
- 5. Data is stored along side with call identifiers
- 6. Data is analyzed using PowerBI
Value Delivered
Insights and call quality control, that was measure by call complaints given on specific channels, now can be automated and analyses in real-time. For this business challenge the value isn’t quantitative but qualitative, since customer service is expected to improve and problems to be detected earlier. As an example:
- • Guided scripts for operators improvements
- • Operator feedback
