Speech analytics – a set of tools and algorithms that converts speech to text, and mines this information for insights – can be invaluable in contact centres, given the overwhelming reliance on voice. Even in multi-channel, it can incorporate text inputs to provide more holistic insights.
Here are the five ways in which speech analytics can be in a contact centre:
1. Call compliance and sensitive data redaction
Contact centres typically have a sizable compliance burden because they deal with large volumes of customer information. There are local regulations pertaining to the need for consent before recording calls. There are also international laws controlling how sensitive/personal data is shared, stored, and utilised. Speech analytics monitors calls in real-time and also checks historical call records for sensitive data, performing the necessary masking and redaction. By coupling speech analytics with IVR, you could even automate call-based payments so that no human agent is privy to the customer’s financial information in the first place.
2. Automated call summarisation
Post-call summarisation is among the most popular use cases for speech analytics, actively leveraged by 35% of contact centres. The tool uses phonetic analysis, Large Vocabulary Continuous Speech Recognition (LVCSR), or a combination of the two to convert audio clippings or real-time calls into a comprehensible text format. This text transcription can be further analysed to extract important trends, organise/tag using keywords, and create succinct summaries by using analytics to extract the most vital and value-adding elements.
3. Sentiment analysis and script cues
Speech analytics can pick up on specific keywords and phrases that denote customer mood, opinion, and sentiment. It can even detect disfluencies like an awkward pause, or a stutter arising from customer anxiety/frustration. There are two ways to leverage sentiment analysis – one is through real-time script cues, where the agent is nudged towards an action based on the customer’s mood at that moment. The second is analysing historical call records and transcriptions based on sentiment tags to find correlations. For example, is the sense of urgency linked with satisfaction or frustration?
