The 19th Coffee with Callbi webinar, held on 13 October 2026, explored how speech analytics and AI can turn customer conversations into practical improvements in quality, efficiency, revenue and customer experience. Hosted by Rod Jones, the session featured Glenn Davids, Quality Assurance Manager for South Africa and Botswana at iSON Xperiences, alongside Callbi’s Corey Springett, Henriette Potgieter and Nosihle Mbatha.

Glenn described iSON’s journey from an initial deployment to using Callbi across several campaigns. Moving from small samples of manually assessed calls towards comprehensive analysis gave teams a clearer view of performance, compliance and coaching needs. The panel explained why quality scores can initially fall: broader coverage exposes issues that limited sampling previously missed.

Commercial value was a strong theme. Glenn shared how analysis of unsuccessful sales conversations revealed customer objections to pricing and additional charges. Taking this evidence back to the client supported revised offers and opportunities to reconnect with customers. Henriette demonstrated why managers should also question disposition codes. A reported right-party contact or unsuccessful sale may not accurately reflect the conversation.

Operational efficiency received equally practical attention. Silence, extended hold time and unanswered customer enquiries can reveal process weaknesses, training gaps or call avoidance. Understanding the cause matters before deciding on corrective action.

The discussion then examined Callbi’s AI capabilities, including contextual questions, automated summaries and coaching insights. Traditional text queries remain valuable for specific wording and scripted requirements, while AI helps interpret meaning, sequence and customer concerns. The approaches complement one another.

Audience questions covered South African languages, transcription accuracy, empathy and data security. Corey distinguished transcription model improvement from the use of large language models, explaining that customer data is not used to train those LLMs.

Key Points

• Broader call coverage reveals quality and compliance gaps that small manual samples can overlook.
• Customer objections can inform pricing, product decisions and renewed sales opportunities.
• Incorrect disposition codes distort performance reporting and conceal the real reasons for poor conversion.
• Silence and hold time require investigation into both processes and contact centre agent behaviour.
• Text queries and contextual AI questions work together, supported by reliable transcription and human review.

Professional Learnings and Take-Aways

1. Establish a credible performance baseline.
   An initial decline in quality scores may reflect better visibility rather than deteriorating performance. Explain this to teams and use the findings to focus improvement.
2. Validate what your reports actually measure.
   Check disposition codes against conversation evidence before drawing conclusions about conversion. Reliable decisions depend on knowing whether the reported opportunities existed in the first place.
3. Investigate the cause before assigning blame.
   High silence or hold time may arise from inaccessible information, unclear processes or inadequate support. Use analytics to identify the cause and select the appropriate intervention.
4. Connect conversation insights to commercial action.
   Recurring objections provide evidence for reviewing offers, pricing and customer communication. Share these findings with decision-makers and track the results of subsequent changes.
5. Apply AI with purpose and professional judgement.
   Choose text queries for precise requirements and AI for contextual interpretation. Validate findings, monitor transcription quality and retain human oversight where ambiguity or risk demands it.