What We Faced
Challenge
Calling a company’s customer service line and getting lost in the automated voice prompts can be a frustrating experience for a caller. It can also become costly for a company when calls are unnecessarily fielded by agents who could be focused on addressing more complex issues.
A company wanted to re-imagine their customer service IVR (interactive voice response) system to offer a top-notch experience on par with other customer care departments. The goal was to offer an improved customer experience which could deflect calls away from the call center.
The focus was on adding natural language processing capabilities so the IVR could interact more conversationally and allow the system to offer improved caller verification, routing, and automated servicing of calls.
Allowing the caller to voice their responses to more open-ended questions would provide a better experience for the caller than old school numerical input of dual tone multi frequency (DTMF).
If they could shift 5,000 calls monthly to self-service options, the company would save $15,000 based on an industry benchmark of $3 per customer service call.
Implemented Technologies
Amazon Connect, AWS Lambda Functions, Amazon DynamoDB, Amazon Lex, Amazon CloudWatch, Integration with Salesforce Call Center, Integration with MuleSoft APIs
What We Did
Solution
Our team worked with client business stakeholders to redesign the customer care caller experience.
Over 200 Amazon Connect flows were developed from scratch to deliver an enhanced and more efficient experience for callers. These flows primarily focus on automated caller identification using caller ID, phone number, and account number, as well as caller verification through street address and zip code. By integrating accurate and efficient caller identification and verification logic with Amazon Connect and Salesforce account data, we achieved a more than 20% increase in IVR caller identification and verification rates.
Amazon Lex was utilized to provide natural language processing capability in both English and Spanish. AWS Lambda functions were utilized to provide custom functionality and logic to the IVR. DynamoDB in conjunction with Lambda functions was utilized to provide dynamic content to the IVR.
For example, IVR prompts and messages were stored in DynamoDB. Allowing admins and power users to change and update the IVR messaging as needed, through a simple record update. Also, caller routing and IVR configuration setting changes were made simpler through DynamoDB config settings.
Integration with backend systems, like Salesforce, to provide additional customer data to the IVR system, allowing for automated detection of callers and caller intents. Examples of this include, allowing the IVR to automatically detect if a caller has an upcoming renewal, if a caller has a pending enrollment or if a caller has an active complaint. Then, providing content or automated routing to the caller based on these automatic detections. Other enhancements included offering the caller the ability save their payment method, for future use.
What We Delivered
Results
By improving the company’s IVR system with natural language processing, the Energy Provider was able to allow a caller to engage conversationally without numerical/dial tone inputs, providing for improved automatic customer servicing.
Also, the improved caller detection and routing capabilities based on caller ID, phone number and account number resulted in improved call times and an overall better experience for both callers and service agents.
The project resulted in:
Improved caller experience. Callers are now able to give voice responses when interacting with the IVR. Instead of having to use their keypad.
Improved caller verification. Callers were able to more easily verify their accounts when interacting with the IVR.
Improved caller detection and routing. New logic for identifying callers with active complaints, pending enrollments, upcoming renewals, or active disconnect notices was implemented in the IVR system. This enhancement enabled faster automatic routing of callers to the appropriate automated flow or a Service Agent, resulting in shorter call times.
- 87 – 90% of eligible Texas callers enter NLP experience, both English and Spanish.
- More than 65% of callers can verify their customer account – by dialed in phone number, phone number, account number, or service address.
- 89% reduction in customers bypassing verification. This has led to more unknown callers to identify their accounts through the enhanced IVR solution.
- True self-service and streamlined customer experience with NLP for natural conversations and an estimated $15k in monthly savings