© 2024 Created by Anuradha Samaranayake
Extract and analyze call center data with Azure Open AI Services🤖
Solution Idea : Extract and analyze call center data with Azure Open AI
Azure OpenAI Service allows us to extract rich insights from customer conversation in the contact center. To achieve this, calls between an agent and a customer are recorded and stored in Azure Storage. From here, Speech API with Speech to Text converts the audio into written text. For this, Batch Ingestion client can be used: Ingestion Client – Speech service – Azure Cognitive Services | Microsoft Learn . As a next step, the output is fed into Azure OpenAI Service, where summarization and information extraction happens. Typically, agents take multiple minutes to manually type down this data, but here, it is fully automated, which saves the agents a few minutes per call. This output can then be fed into PowerBI for near real-time insights, so operators can understand what is happening in the contact center. Furthermore, this information can be stored in the CRM so agents have a rich view of why a customer called in the past and are able to solve problems quicker.
Dataflow
When an agent talks to a customer on the phone, the call is recorded and saved in Azure Blob Storage. The audio files can be uploaded using tools like Azure Storage Explorer or a Storage SDK or API.
Azure AI Speech transcribes the audio files in Batch mode, identifying different speakers. The transcriptions are then saved back in Blob Storage.
Azure AI Language checks the transcriptions to find and hide personal data.
To handle batch transcriptions and personal data redaction, you can use the AI services Ingestion Client tool, which doesn’t require coding and is designed for call centers.
Azure OpenAI processes the transcriptions to extract key information, summarize the conversation, and analyze sentiments. The results are stored in Blob Storage and can be further analyzed and visualized using other services. The output can also be stored in a database for tracking and reporting purposes.
Power BI or a custom web app hosted on App Service can be used to visualize the processed data, providing near real-time insights. This data can be integrated into a customer relationship management (CRM) system to give agents context about the call and help them resolve issues quickly. This automation saves time and effort for the agents.
Components
📁 Blob Storage: This is used for storing raw files. It supports various programming languages like .NET, Node.js, and Python, and files can be accessed via HTTP or HTTPS. Blob Storage offers different access tiers (hot, cool, archive) to optimize cost.
🤖 Azure OpenAI: This service provides access to language models like GPT-3, Codex, and others for tasks such as content generation, summarization, semantic search, and natural language-to-code translation. It can be accessed through REST APIs, Python SDK, or the Azure OpenAI Studio web interface.
🗣️ Azure AI Speech: This AI-based API offers speech capabilities like speech-to-text, text-to-speech, speech translation, and speaker recognition. This scenario uses the batch transcription functionality of Azure AI Speech.
🧠 Azure AI Language: This service consolidates natural-language processing features. It includes both prebuilt and customizable options for various language tasks.
🖥️ Language Studio: A UI tool for exploring and analyzing language features offered by AI services. It allows for building, tagging, training, and deploying custom models.
📊 Power BI: This SaaS tool provides visual and interactive insights for business analytics. It offers data transformation capabilities and can connect to various data sources.
Summery
This solution uses Azure AI Speech to convert audio into written text. Azure AI Language redacts sensitive information in the conversation transcription. Azure OpenAI extracts insights from customer conversation to improve call center efficiency and customer satisfaction. Use this solution to process transcribed text, recognize and remove sensitive information, and perform sentiment analysis. Scale the services and the pipeline to accommodate any volume of recorded data.
