Call Center Headcount Is Falling. Can You Prove AI Did It?

The call center is shrinking worldwide. Global call-center headcount is 5.3% below its December 2023 peak and has fallen year over year for eight straight quarters. The decline began in high-income countries and reached all markets, by income group, by Q4 2025, according to Revelio Labs, which published the data on Oct 6. a16z amplified the chart on X.

If you run revenue at an insurer, a bank or a hotel group, your contact center is one of your biggest teams. Your board will read this as "AI is replacing call centers." Revelio's own answer is careful: "The decline is surely multifaceted."

What Revelio found

All figures below are from Revelio Labs, Oct 6, 2026.

• Eight straight quarters down. "Of the 70 quarter-end readings since 2009, headcount fell from a year earlier only eight times, and all eight have come in the last eight quarters." The most recent quarter showed the steepest drop yet.
• It moved down the income ladder. The decline appeared first in higher-income countries, then spread to middle- and lower-income ones. By country, "certain countries, like China and Nigeria, still show growth."
• Offshoring stopped carrying the total. Revelio puts average call-center pay at $22.52 an hour in high-income countries, $5.44 in upper-middle-income countries and $2.32 in six lower-middle-income countries. Lower-income countries drove growth after 2020. By mid-2024 the global growth rate turned negative.
• Hiring slowed. Across 12 middle-income countries, the entry rate into call-center work fell from 25.1 to 17.2 per 100 workers since November 2022. The exit rate fell from 22.3 to 18.5. Entries dropped below exits.

The Proof Gap

Revelio finds the AI link "hard to wave off." It also says the decline is "surely multifaceted." Both lines belong in any board deck that cites this data.

Buyer question: is AI actually replacing customer-facing headcount, and can I prove it in my own org?

The Order Form

Before a CX agent or an SDR agent replaces any headcount, write these lines into the contract and the pilot plan:

1. Baseline. One full quarter of contacts per agent, handle time, first-contact resolution, transfers, repeat contacts and quality scores, measured before go-live.
2. Holdout queue. A share of contacts or leads stays with people only, so you can measure the difference.
3. Resolution metric. Count a contact as resolved only when the customer does not come back on the same issue within an agreed window.
4. Quality metric. QA scores, complaints and customer satisfaction, reported beside cost per contact.
5. Escalation path. Which requests always go to a person: a claim dispute at an insurer, a fraud report at a bank, a billing complaint at a hotel.
6. Reversal rule. What result sends contacts back to people, and who decides.

For operators: build your own headcount chart

Pull eight quarters of staffing for your contact center and your SDR team. On the same chart, plot contact volume, deflection to self-service, outsourced seats and attrition. Mark the date each AI tool went live. If headcount fell before go-live, or fell with volume, look past AI for the cause. Log any agent rollout you reverse in a Reversal Ledger.

An insurer can run this on first notice of loss calls. A bank can run it on card servicing. A hotel group can run it on reservations.

The Eight-Seat Read

• Sales: If an SDR agent takes first touch, track meetings held and pipeline per lead against the holdout.
• Marketing: Watch what happens to inbound when the first reply comes from an agent.
• RevOps: Own the baseline, the holdout split and the eight-quarter chart.
• Enablement: Train the people who take escalations. Their calls get harder when the easy ones go to an agent.
• Customer Success: Guard resolution and repeat-contact rates. Cost per contact can fall while customers come back more often.
• Partnerships: If an outsourcer runs your center, ask them for the same baseline and holdout.
• CRO, CEO and founder: Tell the board what your own data shows. Cite Revelio as market context, with its "surely multifaceted" caveat.
• Revenue Finance: Price cost per resolved contact, with the holdout as the baseline.

Other causes

Revelio names the rivals itself. The drop "could in part be a correction from the hiring boom that followed the COVID pandemic, or general corporate cost-cutting driven by higher interest rates." Revelio argues those causes should hit all office work, and in its 12 middle-income countries the rest of office work kept growing.

Offshoring is the other moving part. Revelio's pay data shows why companies moved call-center work to lower-cost countries, and its contribution data shows lower-income countries carried growth after 2020. A firm that moves seats offshore or to an outsourcer cuts its own headcount with no AI involved.

Vendor plans add a third factor. Revelio reports that call-center operator TP is targeting €150 to 170 million in annual savings from an AI and cost-reduction program, and that TaskUs's CEO has said agentic AI could cut customer support costs by 25 to 50%. Both are company statements about plans and potential.

Why we might be wrong

The timing match in Revelio's data may be AI after all, and the next few quarters could make that clearer. Your contact center may already run lean, so an agent could cut real cost fast. Or demand may be falling in your market, which would shrink headcount whatever tools you buy. The tests above work in every case.

Sources

• Revelio Labs (Caelan Wilkie-Rogers), After a Decade of Growth, Global Call Center Employment Is Shrinking, Oct 6, 2026.
• a16z on X, which amplified the Revelio chart. No a16z figures are used here.

Get the weekly read on where AI is moving revenue. Subscribe to the Report.