Sarkor Telecom has automated much of its customer support line using an AI voice agent supplied by Apifonica. Both are Boxwood portfolio companies, and this is the first published deployment in which one is the other’s customer.
The constraint was peak load rather than total volume. According to Apifonica, Sarkor’s contact centre — 40 staff across three shifts — handled up to 1,200 calls a day in normal conditions, but outages and technical failures could multiply that seventeenfold, pushing average waiting times past two minutes. A shorter summary Apifonica published a year later puts the same spike at 1,100 calls rising to 17,000, with about half of callers going unanswered. The two accounts do not quite agree, and Sarkor has published nothing of its own on the project.
The agent diverts outage-related and repetitive calls into a separate queue and answers them from a defined set of intents, in Uzbek, Russian and English. Apifonica reports that it runs without a hitch in 95% of cases, that average waiting time fell to between 40 and 60 seconds, and that customer satisfaction rose from 78% to 91%. It also reports that cleaning up subscriber records lifted sales conversion from 30% to 65%, and that no one was made redundant. Planning took roughly six months.