Skip to content

Product1 publisher3 min readPublished

Zoom pushes contact centers to score AI on completed tasks instead of contained calls

Zoom's Ram Rajagopalan says buyers who measured only containment a year ago now ask about end-to-end resolution. Neither he nor the implementation partner beside him gave a figure for either metric.

The Product Desk · Product desk

Illustration accompanying Zoom pushes contact centers to score AI on completed tasks instead of contained calls

What happened

  • Containment, the metric contact centers have used for years, counts how many calls never reached a human agent, which makes an unanswered self-service call look like a win.
  • About half the clients Rittenhouse's firm works with have a formal AI committee, and the rest are still deciding whether AI strategy is set top-down or coordinated across departments.
  • Both men spoke at "The AI ROI in Contact Center Summit," which theCUBE disclosed it covers as a paid media partner, with sponsors holding no editorial control.

Compiled by The Product DeskSomething wrong?How this is made

Why it matters

  • decision An operator who reported containment gains last year now decides whether to re-baseline on resolution, knowing the same call volume will produce a smaller number on the new definition.
  • constraint Completion rates from two vendors cannot be compared until someone publishes a definition of a completed task, and whoever writes that definition also chooses the denominator.
  • exposure Making completion true puts the burden on records systems and integrators: the variables collected by the bot have to arrive with the human agent, or the customer repeats the story.

A caller at nine in the evening asks about a charge on her bill, gets the virtual agent, and hangs up without the answer. That call scores as a success in the monthly report, because containment counts calls that never reached a human agent [1]. SiliconANGLE's writeup of the summit says the plain version: a call can be contained and still leave the customer without an actual answer [2].

Teams assume a contained call means the user self-served. Containment cannot tell a solved problem from an abandoned one, because the only event it records is the agent who never got involved [1].

Ram Rajagopalan, head of product for AI at Zoom CX, has spent the past year building end-to-end resolution into the company's virtual agent platform [4]. "Where I see conversations these days going is not just looking at containment, but end-to-end resolution," he said [6]. Zoom calls it Conversation to Completion, and Rajagopalan described the goal as "actually completing the task that the consumers are calling in for" [7]. He said customers "even about a year, 18 months ago, focused on purely measuring containment" [5], and that containment alone undersells what AI agents should be measured on [16].

The interview does not define what counts as a completed task, and neither man gave a containment or resolution figure [17]. Whoever writes the definition picks the denominator. An operator who wants to compare two vendors next quarter will need its own definition first.

Joe Rittenhouse, co-CEO of Zoom implementation partner Converged Technology Professionals, described the starting point in terms of the hours nobody staffs. "Just a simple question of what do you do after hours? We're usually staffed, we follow the sun from the East Coast to the West Coast, but then after hours we just have a general mailbox. And the answer is typically, I don't know," he said [9]. About half the clients his firm works with have a formal AI committee, and the rest are often still deciding whether strategy gets dictated top-down or coordinated across departments [10]. "These tools work. You just can't boil the ocean," Rittenhouse said [11].

The completion claim depends on the handoff. Rajagopalan said the handoff has to carry "the full context, the reason why they're calling, the collected variables that you need to pass on to a human being," so the agent can finish the job "without having to repeat themselves or look at 10 different places of record just to answer a simple question" [13].

The distance between the containment rates operators reported last year and what those customers actually got goes unmeasured here. The record is one vendor's product lead saying the questions buyers put to him have changed, in an interview at an event theCUBE was paid as a media partner to cover [14][15].

The usable version for Monday sorts last month's conversations, across an operator's top ten contact reasons, into four boxes: contained and completed, contained and abandoned, escalated and completed, escalated and abandoned. Reported containment is the first two boxes added together [1]. Contained and abandoned is the box an operator can count from transcripts this week, before any vendor hands over a definition.

What to watch

  • Whether Zoom publishes a definition of a completed task, including the denominator, that a buyer can audit against its own transcripts.
  • Whether completion targets show up in Zoom contracts and service credits, or stay in conference interviews.
  • Whether the share of Converged Technology Professionals clients with a formal AI committee moves above half.
Loading claim ledger
Loading source directory links
Loading share composer
Loading topic controls
Loading related stories