Your Call Center Job Is Changing. Here's What's Actually Being Cut.

Ylonda Sherrod drove to her AT&T call center in Ocean Springs, Mississippi every day for 17 years. She earned $21.87 an hour, won the company's top performance award two years running, bought a home. Then AI arrived — generating transcripts, routing calls, suggesting responses — and Ylonda started asking her managers a question she'd never had reason to ask before: "Will I be jobless?"

Three thousand miles away, Armen Kirakosian was asking a different question about the same technology. The 29-year-old TTEC agent in Athens used to spend calls frantically clicking through menus and scribbling notes. Now he enters every conversation with the customer's full history already in front of him, often knowing the problem before the caller says hello. "AI has taken the robot out of us," he said.

Same technology. Same job category. The difference between their experiences isn't luck — it's which part of the job each one is being asked to protect.

What AI Is Already Doing to Your Queue

The honest answer isn't that AI is coming for call center jobs. It's already there, and it's being specific about what it takes.

The clearest evidence comes from companies that stopped hedging. Brink's Home Security used AI to cut its inbound call volume by roughly two-thirds — then reduced its call center workforce from about 800 to 400. Microsoft trimmed its customer service headcount from approximately 50,000 to 40,000 as it expanded automation, with an executive saying AI was saving the company around $750 million a year in customer service costs. These aren't projections. They're reported outcomes, already in the rearview.

The task category getting automated has a consistent profile: routine, high-volume, and rule-governed. What's your account balance? Can I change my reservation? Where's my order? These contacts share one trait — the answer is already in a database, and retrieving it doesn't require a person to make a judgment call. Jobs focused on this tier-one work are the ones facing real pressure.

AI is supposed to make our lives easier, but I just see it as my boss.
— Renso Bajala, call center agent, Concentrix Philippines

What's left for humans is different in kind, not just difficulty. A field study of 5,179 customer support agents found that an AI assistant raised average productivity by 14% — and by 34% for newer, lower-skilled workers. The mechanism matters: the system surfaced what experienced agents already knew, compressing the learning curve. The AI didn't replace judgment. It delivered knowledge faster so judgment could happen sooner.

This is the actual picture. The tasks being automated are predictable. The contacts that remain for humans — diagnosing an ambiguous problem, reading a frustrated caller, applying a policy to a situation the script didn't anticipate — are the ones no bot handles well. If most of your current calls are balance-check-level predictable, your exposure is real. If your calls regularly require you to read a situation and decide something a chatbot couldn't, you are not in the same position.

The Redesign Is Already Happening — and So Are the Risks

Here's what makes this moment complicated: most employers aren't mass-firing call center workers. They're redesigning the job.

In a survey of 321 customer service and support leaders conducted in late 2025, Gartner found that 85% were expanding human agent responsibilities as AI reduced contact volume. Three in four were shifting agents into entirely new roles within their organizations. Only 31% had implemented or planned frontline layoffs through early 2027; most expected to reduce headcount gradually through natural attrition while reallocating the people who stayed toward harder, higher-stakes work.

That sounds like good news. In many cases it is. But the transition carries risks that employer surveys don't fully capture — and Ylonda's story makes one of them concrete.

AT&T's AI transcript tool struggled with Ylonda's Mississippi drawl. The transcripts came out full of errors. During the pilot phase, she could correct them. Once the system went live, she wouldn't be able to. That's not a minor inconvenience — it means an inaccurate machine-generated record of a call she handled correctly can become a performance document used against her. She pushed back. She asked her union to establish an AI task force. She testified at a White House listening session on automated technology in the workplace. She was in the top 3% of her company nationally, and she still had to fight for basic accuracy in the tool supposed to help her.

Call center work — it's life-changing. Look at my life. Will all that be taken away from me?
— Ylonda Sherrod, AT&T customer service representative

A separate survey of 1,000 agents reinforces why this matters at scale. Ninety-four percent expected AI to change their roles within three years. Sixty-one percent expected to handle more complex and technical work as a result. In 45% of calls, agents currently spend about three minutes searching for answers — the exact friction that AI is designed to remove.

The employers moving well are stripping out that friction while preserving the human's authority to correct, override, and escalate. The ones moving badly are adding AI monitoring without adding support — raising complexity, raising accountability, and leaving workers unable to fix the record when the machine gets it wrong. Ylonda's experience isn't a cautionary tale about technology being evil. It's a precise warning: an AI tool can fail a high performer in a way she can't fix if she doesn't have a mechanism to push back. Every agent should know this week whether they can correct an AI-generated summary of their own work.

The Agents Who Are Hardest to Replace

Return to Armen for a moment — not as contrast to Ylonda, but as a behavioral model.

Before AI, Armen physically wrote notes between calls and searched menus while customers waited. Those weren't skills. They were friction. When the friction disappeared, what remained was the part of the job that required a person: interpreting what the customer actually needed, adapting when the situation didn't fit the script, staying present instead of hunting for information mid-conversation. He didn't resist the new tools or wait for training. He engaged early and became the person who understood how they worked.

Forrester's analysis of where customer service roles are heading describes this dynamic in useful terms: lower-tier representatives will increasingly "manage teams of AI agents, unblock them when they encounter issues that require a human judgment call, and give feedback to AI to optimize their outcomes." That's not a software engineering job. It's a judgment job that requires knowing the work from the inside — which agents already do.

Critically, working alongside AI may build durable skills, not just temporary output. The same NBER study that found the 14% productivity gain also found that agents who had used the AI assistant performed somewhat better even when the tool was unexpectedly disabled. The system wasn't just doing their job for them. It was teaching.

Three concrete actions follow from this. First, tell your supervisor this week that you're willing to be part of any AI pilot or testing group — being early gives you influence over how it's deployed, not just compliance with how it lands. Second, in your next team meeting, ask one question: "Can agents correct AI-generated call summaries if they're inaccurate?" The answer tells you whether the system is designed to help you or just monitor you. Third, ask yourself: of your last ten calls, which ones would a bot have failed? The list you generate is professional leverage — evidence of what makes you specifically useful.

None of these steps require technical skills. They require the domain knowledge you already have and the willingness to make that knowledge visible to the people making deployment decisions.

What the Numbers Actually Say About Your Future

The Bureau of Labor Statistics projects the customer service representative occupation will decline about 5% over the next decade — roughly 153,700 fewer jobs on net — while still projecting about 341,700 openings per year because people leave and careers move. The industry isn't vanishing. It's sorting.

The sort is between contacts that are routine enough to automate safely and the ones that still require a person to be accountable, empathetic, and right. Gartner's August 2026 survey of more than 3,500 customers found that 87% said access to a human agent was essential even when companies used AI. Customers haven't stopped needing people. They've stopped tolerating bad service at any price — human or automated.

Ylonda's question — "Will all that be taken away from me?" — was never really about AI. It was about whether 17 years of skill, relationship, and performance could be overwritten by a system that misread her drawl and couldn't correct its own transcripts. The question is worth asking. But it's more useful when it's specific: which calls require what you know? Which problems would go wrong without you?

Before you do anything else: write down the last five customer problems you solved that a chatbot would have gotten wrong. Not in abstract terms — specifically: what did the customer say, what did you have to figure out, and what would a scripted response have missed? That list isn't a résumé exercise. It's a map of your current irreplaceability.

The workers who will be hardest to replace are the ones who already know exactly which problems need a person. You probably already know. Start writing it down.

Recommended Tools & Resources