theCUBE Research's Laliberte ties contact center AI payback to knowledge management gaps

A contained call that solves nothing still scores well, because it is short and no human touched it. The customer calls back on Tuesday. "We've had such a focus on average handle time in this industry, but is that shorter call actually a good thing if the issue remains largely unsolved, right?" Kerravala said ^[4].

Move the score to resolution and the work moves with it. Laliberte's checklist for buyers sits almost entirely on the customer's side of the contract. "Organizations should be listening for some practical answers on the importance of high data quality, the integrations that need to be done," he said, adding that firms should be "open to redesigning their process" and that knowledge management raises issues of its own that need addressing ^[5][16]. Those are content and plumbing problems, and a voice agent answers out of whatever the company has already written down.

Laliberte's reason for treating contact centers as the clearest test is that the failures are public. "When something fails, the consequences are also pretty highly visible, and a poor AI interaction can increase customer effort, right? Damage trust and ultimately hurt the brand," he said ^[6].

Both men are analysts and the record here is an interview. Laliberte is principal analyst for networking and observability at theCUBE Research, and Kerravala is principal analyst and founder of ZK Research, a division of Kerravala Consulting ^[7]; they spoke during theCUBE's coverage of "The AI ROI in Contact Center Summit" ^[8]. The coverage does not report resolution rates or payback figures from any named deployment ^[15]. The one hard figure in it is Gartner's projection, and SiliconANGLE published it on September 11, 2026, with roughly three and a half months of the year left for the labor saving to appear ^[1][13][14].

Five9 has leaned on implementation playbooks and voice AI agents to shorten time to value, and the groundwork still sits with the customer, according to SiliconANGLE ^[11]. The playbook covers deployment. Reconciling the three versions of your refund policy is yours.

So sort your top 20 contact reasons on two axes before the pilot. One axis is whether a single canonical written answer exists that a new hire could follow without asking a colleague. The other is whether resolving it needs a write into another system: a refund, an address change, a plan swap. Automate first in the corner where the answer is written and no write is required, because it is the only corner where the vendor's demo matches your live queue. Spend the pilot budget writing answers for the corner where neither is true. Then score the thing on repeat contacts inside seven days and on the share of interactions that ended with a write to the system of record.

"It's important to understand that the brands that lead will not be the ones that simply automate the most, right?" Kerravala said ^[12].