AI Labor Tracker

KT Says AI Will Shrink Its Customer-Service Staff 'a Lot.' Its Field Technicians Have Already Gone From About Seven Home Visits a Day to Eight or More.

When a company says AI is taking over customer service, it usually means three things at once: software answers the easy calls, the leftover calls go to cheaper or more flexible labor, and the people who remain do more. KT in South Korea and RevolutionRace in Sweden showed all three this month.

On Nov. 10, on KT’s third-quarter earnings call, an analyst asked what AI was doing to the Korean telecom company’s costs. Park Hyun-jin, who heads KT’s customer division, gave an unusually direct answer.

“As AI counselors are introduced, the share of customer service inquiries handled automatically by AI is much higher, so I think the number of counseling staff will also fall a lot,” Park said, according to Newsis. That’s my translation from the Korean. Inews24 reported the same line with a softer ending, closer to “I think the number of counseling staff will probably fall a lot.”

Then he talked about the people who go into customers’ homes. KT had automated the scheduling of field visits for wired internet installs and repairs with AI, he said. “Before, one person made about 6.5 to 7 visits a day, but it’s expanding to 8 or more,” Park said, per both outlets.

One week earlier, at a press briefing inside KT’s Gwanghwamun customer center in Seoul, executives from KTis, the KT Group subsidiary that runs call centers, had said something that sounds like the opposite. “There is no plan to cut staff,” a company official said, Sisa Journal-e reported.

Both statements can be true. Figuring out how is the point of this piece.

What KT actually said

Let me be careful about what’s on the record. Park didn’t give a number, a timeline or a department for any cut. The English-language 4th Journal summary of the call notes it’s unclear whether there’s a formal plan, or whether KT would shrink through natural attrition or by moving people to other work. KT said it expects to disclose the measured effects of its AI program next year.

So this is a forecast from an executive, not a layoff announcement. But it’s a forecast with a direction, made to investors, and investors heard it as a cost story.

The technician number is different. That’s a claim about something that has already happened. If a technician made 7 visits a day and now makes 8, that’s about 14% more visits per person. If the starting point was 6.5, it’s about 23%. KT didn’t say how many technicians that covers, or over what period. [Chart: what two companies said AI is doing to customer-service work.]

The three moves behind “AI is taking the calls”

Here’s my argument. When a company says AI is taking over customer service, it’s usually making three moves at once, and only one of them is the AI.

The first move is automation. Software resolves the simple cases: password resets, order status, a router reset. That’s real, and it’s the part that gets the headline.

The second move is rerouting what’s left. The cases AI can’t solve don’t disappear. They go somewhere, and the company gets to pick where: its own staff, a subsidiary, or an outside contractor that can scale up and down with demand.

The third move is intensification. With the easy work gone and the schedule tuned by software, each remaining person handles more, or handles harder cases, or both.

The KT call shows the first and third moves plainly. For the second, look at Sweden.

RevolutionRace sends the leftovers out

On Nov. 3, RVRC Holding, which owns the outdoor clothing brand RevolutionRace, announced it was reorganizing its customer service in a press release on Nasdaq Stockholm. About 25 full-time positions would be affected, subject to union negotiations.

The release explains the logic in three sentences. “AI-based automation has developed to the point where an increasing number of customer cases are resolved without an agent.” Case volumes “vary over the year.” And the company “has entered into an agreement with an external partner.” Cases AI doesn’t resolve “will primarily be handled by the external partner, while complex and escalated cases will continue to be handled by the Company’s own staff.”

That’s all three moves in one paragraph. AI takes the easy cases. A contractor takes the middle. A smaller in-house team takes the hardest ones.

“This is difficult news to deliver,” CEO Paul Fischbein said in the release. “The proposal affects colleagues who have contributed to building RevolutionRace over many years and who have met our customers every day with great dedication.”

The money is small and specific. RevolutionRace expects annual savings of about SEK 7 million with full effect from March 2027, and one-off costs of about SEK 8 million, mostly severance according to Techtidningen. By my arithmetic, that pays back in a little over a year.

Here’s the detail I’d underline. The savings work out to about SEK 280,000 a year per affected position, and the company says that figure is measured against its costs over the last twelve months. It’s a net number. Somewhere inside it is what the external partner charges. The work didn’t vanish. Part of it moved to people who don’t work for RevolutionRace.

Why the leftovers go to a contractor

I think there’s a first-principles reason the middle layer ends up outside the company, and it’s in RevolutionRace’s own release: volumes vary over the year.

A call center has to be staffed for its busy weeks. The release doesn’t say when RevolutionRace’s peaks come, only that volumes vary. When AI absorbs a steady share of contacts, what’s left for humans gets lumpier relative to the staff on hand. Some weeks you need everyone. Other weeks you don’t.

A permanent team is a fixed cost that can’t shrink for a slow month. A partner billing by volume can. So once AI takes the predictable base, the economics push the variable remainder toward labor you can rent. That’s the same logic hospitals follow with agency nurses, and the same move Qantas made when it sent work to Accenture first and promised AI later.

That’s why I’d be skeptical of any headline that says “AI replaced 25 workers.” In RevolutionRace’s case, by its own account, AI plus a contractor plus a smaller in-house team replaced them.

What KTis says happens to the people

KTis tells the other side, and it deserves a fair hearing. The subsidiary runs 169 customer centers for about 120 client companies and directly employs 4,690 people, including counselors and the staff who train and manage them, Sisa Journal-e reported.

At the briefing, KTis gave a concrete example. At the customer center it runs for Hansung Motors, AI now writes the summaries of each call. Of the 44 people on staff, 37 kept their old jobs and 7 moved to 24-hour customer response and outbound marketing, the company said. “We didn’t reduce staff by applying AI,” said Nam Sang-wook, a KTis solutions sales team leader. “We’re expanding into work we couldn’t do before because of cost or staffing.”

KTis was also honest about the limits. “AI is not at the stage where it actually solves every problem,” said Kwon Kyung-yoon, who leads a KTis wired and wireless customer service team. Those quotes are my translations of Sisa Journal-e’s Korean text.

KTis also described how the counselor’s job is changing as the software takes on more. “We started with assistant-type AI, and now we’ve developed to collaborative AI where the AI and the counselor divide roles,” Kwon said. “The final goal is an agentic AICC,” a contact center where the system understands the customer’s problem, finds the right system and handles it directly. In after-sales support calls, the paper reported, a voicebot can already run a remote reset, call the customer back later to check it worked, and book a technician visit if it didn’t.

Follow that sequence through and you can see what’s left for the human. The voicebot takes the reset and the callback. The technician gets the visit. The counselor gets the case that neither the bot nor the reset could fix, which is by definition the hardest one in the queue. Fewer calls, harder calls. That’s what the job becomes.

So which is it? I’d read it this way. KTis is a contractor that sells call-center services to other companies, and it’s growing that business. Its pitch to clients is AI plus people. KT is a customer of that kind of service, and its executive told investors the people part will shrink. A contractor saying “we won’t cut” and its parent saying “we’ll need fewer counselors” fit together fine if the contractor’s growth comes from other clients while the parent’s own need for counselors falls. I can’t confirm that from the public record, and nobody at KT has said it. It’s my inference, and I’m labeling it that way.

How typical is this

Neither company is a giant case. RevolutionRace’s 25 positions are tiny, and KT gave no number at all.

For a broader read, Gartner surveyed service and support leaders and reported on Nov. 5 that 21% had reduced human agent headcount because of AI. That’s a vendor survey of leaders, not a count of jobs, but it says these two companies are part of a real minority, not a fringe.

What makes these two cases useful isn’t their size. It’s that they said out loud where the residual work goes, which most companies don’t.

My read

In my opinion, the most important number in this story is the technicians’ eight visits.

Customer service cuts get the attention, and they matter. But the technician change is the one that shows what AI does to work that can’t be automated. Nobody’s router gets installed by a chatbot. What the software changed was the schedule: fewer gaps, shorter drives, more doors per day. The same person does more of the same physical job.

That can be good for everyone. Less windshield time is a real improvement, and customers get faster appointments. It can also mean a technician who used to finish at 4 now finishes at 5, for the same pay. KT didn’t say which, and the 4th Journal summary raised the same question about labor intensity. That’s the question I’d want answered before I called this a win.

The honest limit: every figure here is a company’s own claim, relayed by reporters. The KT quotes come from news reports of the call, not from a transcript or the audio, and they’re my translations.

Back to Gwanghwamun

KT has said it will publish the measured effects of its AI program next year. When it does, I’ll be looking for three numbers: how many counselors it employs directly and through subsidiaries, how many cases now go to contractors, and how many visits its technicians make in a day and how long that day is.

My recommendation, for anyone who reads “AI is taking over customer service” in a company announcement: ask where the calls AI can’t handle are going. If the answer is a partner, a subsidiary or a schedule with more stops in it, the AI is only part of the story. The rest is labor, moved somewhere you can’t see it.

Chart note: What two companies said AI is doing to customer-service work, November 2026. Measure: company-stated figures as reported (visits per person per day; people; full-time equivalents; SEK). Derived rows are our arithmetic. Sources: KT Q3 2026 earnings call, Nov 10, 2026 (Newsis; inews24); KTis briefing, Nov 3, 2026 (Sisa Journal-e); RVRC Holding AB release, Nov 3, 2026. CSV: ai-customer-service-three-moves-2026.csv.

Sources

1. Newsis, Shim Ji-hye, “[컨콜] KT ‘AI 상담사로 상담 업무 효율화…유선 방문 처리도 개선’,” Nov 10, 2026, 3:45 p.m. KST. https://www.newsis.com/view/NISX20261110_0003862422 (Park Hyun-jin quotes: AX cost efficiency; counseling staff “많이 줄어들 것으로 생각한다”; 6.5 to 7 visits to 8 or more; quantified effects next year).
2. inews24, Ahn Se-jun, “[컨콜] KT ‘AI 상담 확대에 인력 감소 예상⋯내년 AX 효과 공개’,” Nov 10, 2026, 3:52 p.m. KST. https://www.inews24.com/view/2024634 (same quotes; “많이 줄어들지 않을까 생각한다”).
3. 4th Journal (English edition), “The Aftermath of AI: KT Signals Significant Reduction in Customer Service Staff,” Nov 10, 2026. http://www.4th.kr/news/articleView.html?idxno=2121558 (no scale, timing or departments disclosed; unclear attrition vs reassignment; KTcs and KTis; labor-intensity question).
4. Sisa Journal-e, Kim Yong-su, “KTis ‘AI가 상담 넘어 고객요청 처리···상담사 인력감축은 없어’,” Nov 3, 2026. https://www.sisajournal-e.com/news/articleView.html?idxno=424771 (briefing at KT Gwanghwamun customer center; 169 centers, ~120 clients, 4,690 staff; Hansung Motors 44 = 37 + 7; Kwon and Nam quotes; “인력감축 계획은 없다”).
5. RVRC Holding AB, “RevolutionRace reorganises its customer service,” Nasdaq Stockholm, Nov 3, 2026, 11:15 CET. https://view.news.eu.nasdaq.com/view?id=b3a3fb1dadc7251f0b24a18821eb75572&lang=en&src=listed (~25 FTE; SEK 7M annual savings vs last twelve months’ costs, full effect March 2027; SEK 8M one-off; external partner; volumes vary; Fischbein quotes).
6. Techtidningen, Mats Stålbröst, “AI tar över i e-handlarens kontaktcenter: 25 jobb försvinner” (en dash in original title replaced with a colon), Nov 4, 2026. https://techtidningen.se/ai-tar-over-i-e-handlarens-kontaktcenter-25-jobb-forsvinner/ (one-off costs mainly severance; Breakit first reported).
7. Gartner, “Gartner Survey Finds 21% of Service and Support Leaders Have Reduced Human Agent Headcount Because of AI,” Nov 5, 2026. https://www.gartner.com/en/newsroom/press-releases/2026-11-05-gartner-survey-finds-21-percent-of-service-and-support-leaders-have-reduced-human-agent-headcount-because-of-ai

Derived values (our arithmetic): 8/7 = +14%; 8/6.5 = +23%; SEK 8M / SEK 7M per year = 1.14 years (about 14 months); SEK 7M / 25 = SEK 280,000 per position per year (net of partner costs, which are undisclosed).