Updated Your correspondent once met the manager of a contact center dedicated to processing parking and speeding fines. This chap’s “customers” were understandably nearly always unhappy, and the folks who answered the phones therefore copped more than their fair share of surliness and abuse.
Ninety percent therefore quit their jobs within a year.
Other call center operators don’t have to deal with 90 percent attrition, but 50 percent staff annual turnover is not uncommon.
Customer service operations do their best to automate routine chores, so that their best people can be applied to the sort of tasks that require humans to solve problems. That kind of job is more fulfilling, so those workers stick around for longer and, over time, learn to offer superior service. In theory, automated systems therefore improve service and reduce the enormous recruitment and training costs contact centres face.
We offer this quick primer on the customer service caper in the context of a Monday outage* at Google’s Dialogflow CX service, which the search and ads giant describes as an AI-infused offering that creates automated agents that are “similar to a human call center agent. You train them both to handle expected conversation scenarios, and your training does not need to be overly explicit.”
The tech behind Google’s Assistant powers the service, which offers both chatbots and voicebots. Google says whichever you choose they “Deliver more natural customer experiences” and “support multi-turn conversations with supplemental questions.”
When they’re running.
Which they weren’t between 10:19 and 12:40, Pacific time (1719 to 1940 UTC), on Monday September 19.
During that time, Google’s incident summary labelled the incident “Agents unable to receive phone calls” and said its impact meant “Phone calls may not be reachable to Dialogflow agents, and customers may also expect longer wait times on the calls.”
No explanation has been offered for the outage, which Google stated impacted both Dialogflow and its Cloud Machine Learning services.
Google outages are often as not caused by misconfigured software going into production. Whatever the cause, outages at top tier clouds are embarrassing because the major hyperscalers bill themselves as offering superior resilience to even top tier SaaS players thanks to the massive scale of their operations.
In this case, the outage is particularly irksome given contact centers have been encouraged to adopt automation to address their other problems.
The Register has asked Google to provide more info about the cause and impact of the outage and will update this story, or pen another, if we receive a substantive response. And we’ll try to divine if that response comes from a human or an AI, too. ®
* Updated on 21 September to add:
Google has since been in touch to correct its outage comms. It says that the Dialogflow CX and Cloud Machine Learning outage was “misdiagnosed” and its engineers’ investigations later showed the impact was limited to Google internal systems only and didn’t “impact any Google Cloud products or our customers ability to reach Google Cloud support.” It added: “We apologize to our customers for any confusion.”