Contact Centre Post-call Speech Analysis Qorus-NTT DATA Innovation in Insurance Awards 2026

The contact center faced the challenge of understanding customer and agent sentiment on a large scale. The speech analysis process involved the supervisor sampling random recordings through a time-consuming, labourious process that took about five hours each week. Besides, this process often provided a limited view of our customers’ interactions and experiences while overlooking subtle nuances in customer feedback.

This solution is presented by Data Enablement Team, Old Mutual Africa Regions alongside our Executive sponsor -the Head of Digital and Data, Irene Akiy- and was inspired by Amazon Connect, but with the underlying LLM (Amazon Bedrock) having been craftfully fine-tuned by our skilled ML engineers for our own special internal use case.

Within 2025, the automation delivered unprecedented productivity and efficiency gains by analysing 66,044 calls, covering 127,444.42 minutes (≈ 89 days (just about 3 months) of customer/agent interactions for sentiment and insight generation using AI. Before automation, the speech analyst dedicating 5 hours/week would have taken 427 weeks, translating to 8.2 years of manual labour, meaning the voice of the customer would have remained unheard forever.