Customer-side AI agents make service volume your problem, and the evidence is one CEO's phone bill
Newsweek's AI Impact newsletter, published in October 2026, reported that Zendesk chief executive Tom Eggemeier had a personal AI agent renegotiate his cable, streaming, internet and family mobile plans. Within days it had contacted the providers repeatedly and saved him $3,150. He had not dealt with those companies for years and now intends monthly check-ins. He told Newsweek that interactions are "just going to explode", and said Zendesk already sees volume rising among its most AI-heavy customers.
Will personal AI agents really push up my contact volume?
Probably, for recurring, price-driven contacts such as renewals and plan changes. But the only evidence offered is an anecdote and an unquantified Zendesk observation about its heaviest AI adopters.
The mechanism is plausible. When contacting a provider costs the customer almost no effort, frequency rises, and Eggemeier went from silence to a monthly routine. But one executive's household bills are not a cohort. Zendesk gave no baseline and no definition of what counts as an interaction. Nor did it separate agent-initiated contact from growth caused by customers' own always-on bots and wider language coverage. The heaviest adopters may simply have made themselves easier to reach.
How should I split work between self-service, AI and people?
Audit existing contact reasons and sort each into self-service, AI-handled or human-judgment, as Eggemeier suggests, then give customers a way out to a person when an AI conversation loops.
He tied the sorting to customer preference. Some people want empathy rather than the quickest decision, and anyone stuck going round in circles with a bot wants a human. Both assume a person on the other end. If the counterpart is a customer's agent, there is no empathy to protect, and a human escape hatch means little. Retention offers designed for an occasional caller could be harvested every month. That is our inference, not his, but your routing rules should address it.
Which success metric should I trust when automation looks good?
Treat resolution rate as incomplete. Pair it with satisfaction over time and a per-interaction outcome such as money collected or a sale, which is the framing Eggemeier offered.
He warned that higher resolution paired with lower satisfaction will hurt a business later. Ask who defines resolution and who measured it. Ask whether the count includes conversations the customer simply abandoned. Satisfaction surveys also assume a human respondent. His own anecdote cuts the other way from his advice: the outcome was $3,150 the providers gave up, not money collected. A customer's agent measures success in the opposite direction from your collections team.
What should I ask my vendor after this?
Ask for measured volume change at comparable customers, how interactions are counted and billed, and whether you can identify and route traffic from customers' own agents.
Capacity plans built on historical human contact rates are the first thing to break if agents change the mix. Our earlier piece on AI in workforce management is a reminder that forecasting is where this lands first. Eggemeier's remarks also sit alongside our recent Freshworks CEO interview, where the theme was software and AI working together. If any part of your contract scales with conversations or resolutions, find out what a surge would do to your bill. The newsletter text we read breaks off as it turns to the front door of customer service moving, so we have not assessed that claim.
Frequently asked questions
Did Zendesk publish data showing volume growth?
Not in the coverage we read. The CEO said volume is rising among the heaviest AI adopters but gave no figure. The only number was his own $3,150 saving.
Is $3,150 a typical result?
Nothing suggests it is. It is one executive's report of one agent working across several household plans. Results will depend on how overpriced the plans were and how much each provider is willing to concede.
Do customers want to deal with AI?
Eggemeier said it depends. Some want empathy, some want speed, and anyone stuck in a loop wants a human.
Sources: Newsweek, "Zendesk CEO: AI agents could reshape customer service" (AI Impact newsletter).