The task
HAVAL has a dealership network across Russia and a huge flow of sales calls, over 300,000 minutes of conversation a month even on a sample. Checking that volume by hand is impossible, spot checks covered a fraction of a percent, and off the shelf tools could not cope with such volumes and 40+ communication criteria. Script deviations and lost sales went unnoticed.
What we did
We built speech analytics as a fully custom solution: it processes calls across the whole network and scores every conversation against 40+ criteria based on a checklist and scripts, handling over 300,000 minutes a month. AI agents flag weak calls: if a conversation drops below 70%, it is automatically escalated to a senior manager so the sale is not lost. We connected BI dashboards with deep analytics on how calls influence sales and tracking of deals after a call, NPS and feedback collection, and we give management ready summaries. The speech analytics, the agents and the dashboards are all custom development. On top of that, agents book customers into the service center, remind about scheduled maintenance and upsells, and act as account managers.
The result
Now HAVAL sees how the whole network talks to customers, at a scale beyond manual review or off the shelf tools. Weak calls are caught and escalated in time, script compliance is measurable, and the link between a call and a sale is visible in dashboards rather than guessed.
speech analytics + agents + BI: 40+ criteria, auto escalation below 70%, dashboards.
