Will AI Replace Customer Service Workers?

Customer service representatives score 55% on Oxford Martin School's 2013 automation scale — already high risk. But Anthropic's 2026 Economic Index reveals 70% observed AI exposure — meaning more than two thirds of the actual daily tasks performed by customer service workers are already being handled by AI in real professional settings. Our combined score is 64% — the highest observed-risk score of any profession we track. Check your specific role here →

No profession in this research is more directly and immediately affected by AI than customer service. The disruption is not coming. It is here.

What has already happened

Klarna announced in 2024 that its AI assistant was doing the work of 700 customer service agents. By early 2025, the company was quietly rehiring humans after customer satisfaction dropped — but the rehiring happened at lower wages and on more precarious contracts. The AI had not worked perfectly. It had worked well enough to justify the layoffs, and the standard for “well enough” had been set by comparing AI performance to the cost of human labour, not to the quality of human judgment.

This pattern — AI handles the volume, humans handle the escalations, at lower overall headcount — is now standard across the industry. Where companies once employed 100 agents, they now employ 15 to 20. The AI manages routine queries. The humans manage the situations the AI cannot.

Oracle’s April 2026 layoffs made the same logic visible at scale: technical and support roles eliminated to fund AI infrastructure, with the expectation that AI systems will absorb the work. The Oracle story is not unique. It is becoming the template.

What remains for humans in customer service

The tasks that AI struggles with in customer service are the ones that require genuine empathy, complex judgment, and the ability to navigate situations that fall outside any training data. A customer who has just been bereaved and is trying to cancel a service. A dispute involving genuine ambiguity about what was promised. A situation where the right answer involves bending a rule in a way that requires understanding why the rule exists.

These situations are real, and they still require humans. The question is how many humans — and the answer, across the industry, is: far fewer than before.

The honest picture for customer service workers in 2026

Of all the professions on this site, customer service is the one where the honest answer is hardest to deliver gently. The disruption is not marginal and it is not reversible.

The remaining human roles in customer service are concentrating at the escalation and complex-case end of the spectrum. They require more skill, more judgment, and more emotional intelligence than the routine roles they are replacing. The total number of positions is falling.

If you currently work in customer service, the most useful question is not whether AI will change your role. It is whether the skills you are developing — the ability to handle genuinely complex situations, to manage distressed customers, to exercise judgment in novel circumstances — are the skills that the remaining human roles will require.

Based on Oxford Martin School research (Frey & Osborne, 2013) and Anthropic Economic Index (March 2026).