AI Agents Reshape Customer Service and Finance

Zendesk CEO Tom Eggemeier says customer-service interactions are about to surge, and he has a receipt to prove it. He recently asked a personal AI agent to renegotiate his cable, streaming, internet and family mobile plans. Over four days, the agent contacted providers 37 times and saved him $3,150.

That’s a striking number for a man who, by his own account, hadn’t called those companies in three or four years. Eggemeier now plans to have the agent check in with them monthly. “What’s changing is the amount of interactions is just going to explode,” he told Newsweek.

The Customer Side

Eggemeier said Zendesk sees interaction volume climbing fastest among customers who have adopted AI most heavily. Automated service can run around the clock and in languages businesses previously supported only during limited hours. That’s a real advantage for companies willing to build for it.

But the technology doesn’t decide everything. Companies need to sort out which issues should stay self-service, which AI can handle and where judgment or company policy requires a person. Eggemeier drew the line plainly. “I think there’s certain situations where you want more empathy and you’re not looking for the quickest decision,” he said.

He was just as direct about the frustrations. “On the other hand, if I’m getting in a kind of circle loop with an AI agent and I’m not getting it solved, I want the ability to go to a human being.” He urged executives to measure whether automation actually improves service, rather than counting how much work the machines absorb. “If you’re getting higher resolution rates, but lower satisfaction, that is gonna impact your business in the future,” he warned.

Measures of success vary by interaction. An issue might count as resolved, a payment collected or a sale closed, and Eggemeier said chief executives should weigh those outcomes alongside customer satisfaction rather than trusting any single number.

Competing Gateways

Eggemeier also expects customers to increasingly reach companies through personal agents or general-purpose interfaces such as ChatGPT and Claude, rather than a firm’s own website or support tools. That could mean fewer visits to company sites and less human-to-human contact. It also means businesses must protect their information and customer experience when another AI system stands between them and the buyer. “I don’t think closed ecosystems are gonna work that well,” he said.

A separate Newsweek analysis made a related case: businesses should prepare for customers who talk rather than type. It cited a June 2026 YouTube video in which Boris Cherny, creator of Claude Code at Anthropic, said roughly half his engineering work now happened on his phone. He could manage AI agents remotely and direct them by voice while walking in the woods, getting coffee or sitting on a couch.

The numbers behind that shift are hard to ignore. OpenAI reported that more than 150 million people use ChatGPT through Voice and Dictation each week. Its July 2026 launch of GPT-Live introduced voice functions that listen and respond at the same time, handle interruptions and delegate complex work mid-conversation.

The analysis also pointed to the limits of talking. A 2026 study of more than 10,000 U.S. adults published in Scientific Reports found median typing speeds, adjusted for errors, of 37.8 words per minute on computers and 29.4 on smartphones. Standard American English conversational speech runs at about 150 words per minute. Voice can fill gaps where typing isn’t practical, but it can stumble in noisy rooms, misread instructions or expose sensitive information when spoken aloud. The analysis argued that typing remains better for detailed writing and precise editing, and that users should be given the choice.

Enterprise Returns and Financial Guardrails

The business case depends on how deeply a company commits. A Wharton and GBK Collective study found that 67 percent of enterprise leaders whose departments had fully embedded AI, adopted agents and significantly redesigned workflows reported strongly positive return on investment. Among those that had not, the figure was 19 percent. Tools bolted onto old processes, in other words, rarely pay off on their own.

Money movement raises the stakes. Charles Schwab Corp. Chief Executive Officer Rick Wurster said his brokerage, wealth-management and banking company actively blocks unauthorized AI agents attempting to move clients’ money, including Meta Platforms Inc.’s Muse. Bank executives, he said, welcome consumers using AI to support their financial lives, so long as the agents follow the rules. That’s a sensible line. Customers should get the convenience of software that works for them, but the firms holding their assets should decide who gets access to them.

Why This Matters: Customer service is becoming a contest between companies and the software their customers carry in their pockets. Eggemeier’s own experience shows the upside for households, with $3,150 saved in four days by an agent that never tired of hold music. Businesses face a harder question: whether to measure automation by the work it removes or by the satisfaction it leaves behind. The Wharton findings suggest that firms which rebuild their workflows, rather than layering new tools onto old ones, are the ones seeing strong returns. Schwab’s stance on unauthorized agents shows that the rules governing money will matter as much as the technology itself. Individual consumers and private firms are driving this change, and the market will reward the companies that adapt to it on their own terms, while those that wall themselves off may find customers simply walking around them.