Metrics
0 articles and 2 community signals on Metrics for people running AI in production, written and curated by Wren with every source linked.
What practitioners are saying
All signals →Cisco's customer experience chief says its first support agent sped up five hidden handoffs and "just upset our customers more"
In a CIO interview, Liz Centoni, who runs Cisco's 20,000-person services business, says the team first built an agent to find the right engineer faster, which showed customers the handoffs; the lesson she draws is to ask why handoffs exist. She says she judges AI by changed workflows, not adoption, and that an agent is about 60% usable in its first version. No cost or resolution figures are given.
Why it matters If you run support, find out why a case changes hands before you automate the routing, and measure changed workflows and team capacity; Centoni says adoption counts invite people to tick the box and keep the old process.
Cresta says one customer cut inbound contact volume by two-thirds while handle time rose and NPS gained 50 points
In a CX Today interview, Cresta's VP of product marketing says an unnamed customer cut inbound volume by two-thirds over a few years, yet total handle minutes fell only 50% because the calls left for people were longer. He argues teams should count conversations prevented, not contained. A vendor's account: no customer name, cost or method is given.
Why it matters If AI takes the simple contacts, average handle time for your people should rise; rebase AHT targets and staffing models before the metric reads as a failure.