AT&T plans to serve all of its home internet, phone and postpaid wireless customers through care staff located in the United States by the end of 2027, the company announced last week. It is also investing in better tools so its “Care experts spend less time navigating internal systems and more time helping customers.”
AT&T’s promise places it at the center of two major forces shaping customer service: regulation and AI.
In June, the Federal Communications Commission closed out the comment period for a proposed rule that would require U.S.-based companies handle a certain percentage of their customer service call volume with call centers based in the U.S.
That rulemaking is not final, but it’s casting a long shadow across the space. FCC Chairman Brendan Carr lauded AT&T for its announcement, calling it a “win for American jobs.”
The proposed rule is exerting real pressure, according to Maribel Lopez, founder and principal analyst of Lopez Research. “To ensure the best relationship with the FCC, providers like AT&T need to show not just effort but actual onshoring,” she told CX Dive in a text message.
Many companies outsource some, or all, of their customer service functions to business process outsourcers. Those BPOs might be based in the U.S. but they often employ agents who live in other countries because those agents are less expensive compared to the U.S.-based agents they might employ.
“The billable rates are $35-$40 an hour in the U.S. now for outsourcers, versus in the Philippines, where you're at $12-$14,” King White, CEO of Site Selection Group, told CX Dive.
Put another way, U.S.-based call center agents demand a different wage for essentially the same labor.
For a BPO, “70% of the P&L is always going to be the labor. It's always in that range,” Michael Replogle, SVP of contact center advisory services with Site Selection Group, told CX Dive. “If you cut your costs through labor arbitrage by 45% or 50%, and all of a sudden you have to bring it back on shore, you have to be competitive.”
Hiring agents in the U.S. has historically been difficult, Replogle said. “Staffing over the last 10-12 years just seems to get harder and harder to find good qualified people.”
AT&T will likely face such challenges as it shifts its workforce to the U.S.
“We use different workforce models across our business today, and we’re still determining the right workforce mix as we make this transition,” an AT&T spokesperson told CX Dive via email. “We expect that mix to include union employees and U.S.-based contractors.”
The company does not publish how many service representatives it directly employs — or indirectly through BPOs. However, at the end of last year AT&T, had over 133,000 employees, of whom approximately 43% were represented by unions, according to AT&T’s 10-K filing.
AI adds a new variable to the labor equation
AT&T’s announcement might herald a sea change for the U.S. contact center industry — a relatively sudden increase in demand for agents, which would cause U.S. wages for those agents to rise.
And yet, AI is a key variable since AI is already beginning to handle some easy-to-automate customer service inquiries.
“The first wave of AI is taking the issues that can be easily handled, but it’s increasing the complexity of what’s left,” Replogle said.
That’s raising the bar for the qualities companies seek in their agents. “It's somebody who really has good analytical skills, can think and provide that solid customer experience,” he said
Call centers use AI in multiple ways. One approach is self-service, where the automation tries to solve the customer’s issue without escalating to a human agent. This isn’t a new concept, since interactive voice response units perform the same function.
Another complementary approach uses AI during the call to assist the agent and take over tasks that require cognitive load, like live note taking.
“In the old days you had to make sure you captured everything. You don't do any of that today if you're using certain software,” Replogle said. “It's dispositioning the call. It's putting the proper follow-ups regardless of channel. It's helping agents find the answer quicker and better versus them digging through a knowledge-based system.”
Agent assist, which virtually every contact center as a service vendor offers, helps human agents become more efficient and productive. They handle each call more quickly and accurately, shortening their after call work.
These gains help the contact center operation save money, since the calls are shorter, but each agent can — ideally — handle more calls per hour.
“BPOs are typically lower cost, lower skilled, high-churn workers that are paid by the agent. If you can field routine requests with your own AI agents, the real value comes from having higher skilled workers supporting the complex queries,” Lopez said. “Many vendors are thinking about offering at least a portion of the software expense in an outcome-based pricing fashion. So there is a scenario where you could actually have fewer agents onshored working with AI to do higher value work at the same cost as a larger number of BPO agents.”
This approach could change the outsourcing arithmetic from simply cutting CX costs via labor arbitrage to figuring out how AI can be factored into that equation.
“If you’re outsourcing work to a BPO, their margins are double when they go offshore versus for domestic work,” White said. “But then if you look at the math, their revenues are twice as much. So, it might be a wash.”
The big question is how AI might enable an enterprise to cost-effectively reshore the CX work it had outsourced to lower cost regions.
The onshore versus offshore math might balance out from a labor cost perspective, but AI also comes with a price tag. "With all the software costs and token costs...the cost of an agentic AI agent is not cheap,” White said.
“I think what's going to be impacted the most is these offshore markets,” White said. “It's where the job erosion is going to happen. It's in India and the Philippines, in places [that handle] the low-end entry level calls. That's what's going to get hit really hard.”
