In one week after deploying an Intelligent Virtual Assistant, powered by Conversational AI, our customer was able to handle 46.48% of its calls automatically.
The Challenge
Prior to noHold engagement, our customer’s call center was receiving more than 2.3M calls per month. The customer was using an intelligent IVR and through it, was able to handle 13% of its calls automatically. Due to the launch of a new product and service, the customer was expecting a 5% to 10% increase in support calls.
Our Approach
We divided the project in two phases. Phase I, the customer created a Pareto chart of the top call drivers affecting its call center. noHold and the customer identified the issues that could be addressed automatically and without any special integration. noHold created the first version of the Virtual Assistant to address those issues based on content already available. noHold created a phone number and linked it to the Virtual Assistant, so that any call to that number, would be answered by the Virtual Assistant. In Phase II, noHold created connectors to the appropriate back office systems and cores to personalize the customer experience. noHold linked the new version of the Virtual Assistant to the same phone number, so that no changes were needed to the customer phone system.
Lessons Learned
- The metrics collected by the Virtual Assistant are critical in identifying actionable customer behavior metrics.
- Strong Natural Language Understanding (NLU) algorithms are essential to compensate for the challenges of speech recognition through a phone line.
- An Inference Engine capable of rendering multi-turn conversations with the customer is also critical to completely understand customer intents.
- Enabling customers to interrupt a conversation, change subject, request a re-read are effective ways to guide callers to the right solution.
Conclusions
- Increased call center efficiency by 46.48%
- Increased Net Promoter Score by 2 points
- Deployed Phase I in one week and Phase II in three weeks, with minimal resources and risk