Product1 publisher3 min readPublished Updated
Cisco and Zoom executives challenge containment rate at AI contact center summit
At theCUBE Research's AI ROI in Contact Center Summit, executives from Cisco, Zoom and Talkdesk argued that a contained call can be a customer who has given up, and offered open-to-close time as the replacement.
The Product Desk · Product desk
What happened
- theCUBE Research ran an AI ROI in Contact Center Summit with executives from Cisco, Five9, Zoom, Talkdesk and implementation partner Converged Technology Professionals on the speaker list.
- Talkdesk's survey has the organizations it calls CXA leaders reporting roughly 22% net promoter score gains against 5% for the maturity tier below, with cost per contact improving 57% against 48%.
- Talkdesk's Pedro Andrade named the replacement measure as the time from the opening of a problem to its close, with back office work counted in, instead of average handle time.
Compiled by The Product DeskSomething wrong?How this is made
Why it matters
- decision A CX leader whose board pack leads on containment rate is now reporting a figure that a Cisco general manager has called a term he disbelieves in, so the next review has to decide whether the number stays on the first slide.
- constraint Moving from containment to implied resolution swaps a countable event for a score a vendor's model produces, so buyers need to know how the grading works and whether they can check it against their own transcripts.
- cost Measuring open-to-close with back office work included puts the number partly in the hands of teams outside the contact center, and someone has to fund the instrumentation that stitches those queues together.
- contradiction The case against containment comes from vendor executives at an event theCUBE Research co-hosted with them, and the account stops short of showing any of the four changing what its own reporting counts.
A customer calls, gets a bot, gets nowhere, and hangs up to solve the problem some other way. In a containment report, that call counts as a success. Vinod Muthukrishnan, vice president and general manager of Webex Customer Experience at Cisco, said, "Containment could well be me telling you I'm never calling you again. That is not good containment." [3] He also said he and his peers are "huge disbelievers in terms like containment and deflection" [2].
Talkdesk brought numbers that show which lens moves. Its survey has the group it calls CXA leaders reporting roughly 22% net promoter score gains against 5% for the maturity tier below [6], while cost per contact improved 57% against 48% [7]. Divide the pairs. Twenty-two over five is a 4.4x spread on experience; 57 over 48 is 1.19x on cost [13][14]. Pedro Andrade, Talkdesk's vice president of AI, said savings alone do not show the whole picture, and that they understate the value available from AI [8].
The metric Andrade named to replace average handle time is open-to-close: "You are going to measure how much time it takes from opening of a problem until it gets closed," back office included [9]. That clock runs through teams that do not report to CX. His reason for extending it is that a generative layer on its own "stops at the moment when you need judgment or an action that crosses multiple systems", and by Talkdesk's research roughly 74% of enterprises already have that layer in place [10].
Zoom's version of the replacement has a catch for whoever signs the monthly report. Ram Rajagopalan, head of product for AI at Zoom CX, said the company now pairs a post-call survey with what it calls implied resolution, using a large language model to grade the outcome, as part of a "conversation to completion" roadmap [4]. "We don't want to count customers who are frustrated and drop off as a call resolution," he said. "It is certainly contained, but we didn't resolve the customer's issue." [5] Containment was cheap to audit, because the call either left the queue or it did not. A resolution grade comes out of a model, and in Zoom's case the model is the vendor's.
Amit Mathradas, chief executive of Five9, described the prior wave as deterministic, "effectively a decision tree masked with voice capabilities", and the new wave as probabilistic, and said the governance model built for interactive voice response does not transfer to it [11]. Muthukrishnan put the same gap in build terms: "It's never in our lifetimes been easier to build an AI agent," he said. "It's never been harder to make it enterprise-grade." [12]
One limit on all of this: the summit account does not say whether any of the four has taken containment out of the reporting it ships to customers [16]. These were vendor executives talking about vocabulary at an event theCUBE Research co-hosted with them [1].
The grid that separates the pitch from the practice has two axes: contained or escalated, and resolved or unresolved. Containment rate adds the top row together and reports one number. Contained-and-resolved is what the number claims to count. Contained-and-unresolved is the customer in Muthukrishnan's line, and the two inputs that pull it out of the total are repeat contacts on the same issue and the back office tickets opened after the call ended.
What to watch
- Whether Zoom publishes how implied resolution is scored, or lets customers grade transcripts with a model they control.
- Whether any of the four vendors ships open-to-close time, back office included, as a standard report rather than a summit talking point.
- Whether Talkdesk releases the survey base and the definition behind the 57% and 48% cost-per-contact figures.