Your contact center acts as the bridge between your business and your customers, where every call is a chance to better understand their needs and assess how effectively you’re meeting those needs. Each conversation holds valuable insights, but unlocking this potential has traditionally required agents to summarize calls after every interaction manually.
For many contact centers, call summarization is crucial for tracking performance, gaining insights, and ensuring seamless customer experiences. Well-crafted summaries not only help in performance analysis but also eliminate the need for customers to repeat themselves when transferred to another agent, improving the overall customer journey.
In this post, we’ll explore how Generative AI and call transcription can reduce the effort, improve the accuracy of call summaries, and enhance overall contact center operations.
The Challenges with Manual Call Summarization
As contact centers handle more interactions, the need for efficient and accurate call summarization becomes increasingly important. Unfortunately, manual summaries often fall short due to several challenges:
- Time-Consuming Process: Manually summarizing a call is tedious and can consume up to a third of the total call time. This adds pressure on agents and leads to longer Average Handle Times (AHT), negatively affecting agent productivity and customer satisfaction.
- Inconsistent or Inaccurate Summaries: Due to time constraints, agents may skip summarizing altogether or provide incomplete information, leading to poor documentation. This becomes especially problematic when customers are transferred between agents, requiring them to repeat their issues, thus creating frustration and negatively impacting the customer experience (CX).
- Negative Impact on Metrics: The longer agents spend on manual tasks like summarizing, the more it affects key contact center metrics like AHT and First Call Resolution (FCR). The inefficiency not only frustrates customers but also strains agent performance.
Generative AI: Improving Call Summarization
Generative AI is changing how customer calls are summarized by making the process faster and more accurate. It uses large machine learning models called foundation models (FMs) that are trained on a lot of data. A specific type of these models, known as large language models (LLMs), focuses on understanding and generating language. These LLMs can create summaries that sound natural and fit the context of the conversation.
The best LLMs can easily handle complex sentences and identify important details like topics, intents, next steps, and outcomes. By automating call summarization, LLMs save time compared to manual methods. This helps contact centers improve customer experience while reducing the amount of work agents have to do for documentation.
Solution
Generative AI needs something in the form of text to make a summary, for this purpose, transcription of the call is required. The accuracy and usefulness of the summary depend on the quality and accuracy of the transcription. For general purposes, amazon provides its transcription services through Amazon Transcribe which can generate real-time and accurate transcriptions of the call. Then this transcription is fed into an LLM module which in Amazon’s case can be Amazon Bedrock. This solution is applicable in the case of contact centers other than Amazon Connect. We will see how this process happens in the Amazon Connect contact center.
In Amazon Connect contact center, the contact lens for Amazon Connect provides call transcription and real–time analytics natively. Generative AI must be enabled for post–contact summaries. An agent can easily access post–contact summary and real–time transcription through his CRM. Generative AI also provides suggestions in real time. To save and for later use of transcription and summaries an architecture has to be developed, this not only helps in retrieving the data for later use either by another agent or supervisor to analyze agent’s performance.
The following architecture is for Amazon Connect contact center
Most of the heavy work is done by Amazon Connect itself, it provides transcription, real-time analytics, and summaries. A Lambda function processes all the data generated and stores it in Dynamo DB. An agent can easily access this data through API Gateway, either in real–time or after the call ends.
To enable transcription and summaries some steps have to be followed during IVR building
- Add a Set recording and Analytics block to your flow.
- Click on the block and there configure its properties:
- Turn on Call recording and select Agent and Customer
- Scroll down and turn Analytics on
- Turn on Enable speech analytics.
- Select Real-time and post-call analytics.
- Under Contact Lens Generative AI capabilities, choose Post-contact summary.
3. Then assign relative permissions to agent, supervisors and administrators.
Conclusion
In conclusion, real-time call summarization through Generative AI is transforming contact center operations by streamlining the way customer interactions are captured and analyzed. By automating this previously manual task, contact centers can significantly reduce Average Handle Time (AHT), improve First Call Resolution (FCR), and ensure that agents remain focused on delivering better customer experiences. Implementing AI-driven solutions like Amazon Connect enables contact centers to enhance employee productivity and performance while providing valuable insights for continuous improvement. As customer expectations evolve, embracing these advancements will be crucial for staying ahead in today’s competitive landscape
