[package] Automatic CSAT from conversation analysis

Context: Understanding customer satisfaction is important for our business, however, today's approaches are too imperfect:

• Survey: asking users directly.

  • Response rate low (1-3%) and polarized (super happy or frustrated)

  • Not representative of all messages or conversations. This can be disrupting and

  • Time-consuming for the receiver. Can be delicate to ask

• Manual reporting: support agent or RM logging the sentiment of their customers takes time, is hard to scale and hard to get accurate.

• Low volume: difficult to get enough customer survey responses for meaningful CSAT reporting

Request: Get real-time satisfaction insights (~CSAT) across every conversation, without introducing a change in behavior or new software.

We would like to know customer satisfaction :

• at different levels

  • Message

  • Conversation

  • Contact

  • Account: "Account Health" type scoring --- ex, if someone writes in 6 times over the year, what's the sentiment of all those conversations (whether they are by the same person or from different people within the same company)

• at different time

  • Real time → to adapt workflows (SLA, escalation, tone …)

  • Specific time (First inbound, at the end (resolution))

  • Over-time

Actions:

Making the (AI) customer satisfaction ratings visible in analytics would be another way of getting insights into quality of communications.

• be able to drill down into conversations where the satisfaction is low.

• reporting on satisfactions

• create a workflow based on satisfaction