Every day businesses rely on call centres to manage, monitor and process customer requests. These recorded voice interactions generate much valuable data, but a lot is lost as monitoring is mostly performed manually. Result: poor customer service and profit losses for businesses.
Superbot enhances customer service experience
The EU-funded CLARA project used machine learning via a unique superbot speech analytics system to leverage customer call data automatically. The project researchers sought to deepen the understanding companies have of their customers revealing insights from call centre conversations. These conversations are like a black box. Pablo Enciso, CEO of Predictiva and project coordinator, explains: “There is no technology that can structure and categorise this data in a massive and scalable way. Less than 0.5 % of these calls are manually audited.” This is a subjective, expensive and statistically irrelevant approach and makes companies lose an incredible opportunity to listen and understand their customers. It is in these conversations where customers express invaluable information such as how products or services can be improved, reasons for dissatisfaction or churn, and market offering, to name a few.
A solution is here
“At Predictiva we have developed ‘Upbe', a conversational intelligence platform, to structure and categorise automatic, massive and scalable information contained in telephone conversations between clients and companies,” continues Enciso. This offers their customers, both call centres and end businesses, a definitive solution to understanding what happens in their interaction with clients and extracting key information to optimise their processes, provide a better user experience and improve the quality of their services.
How it works
Upbe is a technology that combines machine learning and natural language processing to understand customer insights within phone calls. Upbe has several technology layers. First up is the speech to text transcription engine that converts full audio recordings into text. This transcription engine is proprietary. It is specialised in the context of call centre telephone calls that are technically highly complex due to overlapping voices, background noise, different speeds at which agents and clients speak, audio compression, microphone quality, etc. Once the transcription and analysis of the audio signal are gathered, the natural language processing module then works to identify semantic context within the conversation. These contexts are key to identifying valuable customer insights. Upbe has a template of preloaded items that can identify contexts related to customer support, compliance or the voice of the customer. Upbe platform allows the configuration of personalised contexts, letting companies decide whatever they want to find within these calls. The results are then displayed in several dashboards under a business intelligence section.
Benefits of machine learning and natural language understanding
Upbe’s unique and proprietary natural language processing models and methodology provide companies with answers to some of the most valuable questions. Topics of questions range from how products can be improved to what the main reasons are for customer churn. “This is in our opinion a great achievement in terms of providing great value to our customers and exceeding at what machine learning and natural language understanding are capable of,” concludes Enciso. The company is now working on more sophisticated approaches to provide the most important metrics to call centres as a pre-setup module on its product. This will reduce the personalisation and set up for its customers and will accelerate its go-to-market strategy offering the only technology tool able to do this with specific models per industry.
CLARA, Upbe, call centre, machine learning, natural language processing, superbot, customer service, speech analytics