A company specializing in software for distribution automation has integrated digital employees into its support department. The implementation process involves three parallel workflows: a consultant widget in the CRM, autonomous ticket handling by an AI agent, and automated suggestions for human engineers.
Data analysis shows that the median time to reach a substantive resolution dropped from 21 hours and 52 minutes to 22 minutes. The 90th percentile for actual resolution decreased from 211 hours and 41 minutes to 76 hours and 36 minutes. These figures are based on a sample of 1,876 tickets before the launch and 3,078 tickets after.
Between July 20 and August 30, the digital consultant received 1,934 inquiries, the AI agent processed 3,558 tickets independently, and engineers received 1,208 automated suggestions. Of the 3,078 closed tickets analyzed post-launch, the agent resolved 1,647, while 743 were escalated to human engineers.
The company highlights that faster resolution times do not always equate to higher quality. In a regression test of ten previously solved cases, two updated versions of the agent provided worse answers than the original, despite completing the tasks faster. Consequently, the team implemented stricter validation rules for data searches and external service calls.
