Customer support and contact centres
A faster answer can help a customer and the person serving them. Whether AI also improves jobs depends on what happens to pay, workload, recruitment and the quality of the service.
A resolved enquiry is more useful than a quickly closed conversation. That distinction matters when assessing AI in customer support: the worker must still establish what the customer needs, recognise exceptions and know when to seek help. The sector’s scale provides the starting point; its working conditions determine what an improvement means.
The sector in context
European Union · 27 countries · 2024
Activities of call centres
Eurostat · NACE Rev. 2, group 82.2
- Persons employed
- 500,000 estimated
- Enterprises
- 11,014
- Net turnover
- €30 billion estimated
Annual structural business statistics, including employees and self-employed persons. Employment and turnover carry Eurostat’s estimate flag. These are industry totals, not counts of jobs exposed to AI. [2]
United States · Private sector · 2025
Telephone call centers
BLS · NAICS 56142 · Annual averages
- Covered payroll jobs
- 327,577
- Establishments
- 14,961
- Average annual pay
- US$58,411
QCEW administrative records. Pay is the mean across covered jobs in this industry, including managerial and other occupations; it is not a median agent salary. [3]
Read these as separate baselines. The years, classifications and statistical units differ. An enterprise may contain several establishments; payroll jobs are not a count of unique people. The US category includes answering services and other contact centres, including units serving other establishments within the same enterprise. No EU–US employment ratio or wage comparison is inferred. [2][3][4]
The broader US occupation of customer service representative illustrates the difference: the BLS records 2,666,000 jobs across industries in 2025. Its 2025–2035 projection anticipates a 5% decline, alongside about 289,500 annual openings from occupational transfers and labour-force exits. These projected openings are not net job creation, and the decline is not a measured effect of generative AI. [1]
What assistance has demonstrably changed
Generative AI at Work examines 5,172 agents serving US business customers through a software company and its subcontractors. Of those agents, 89% worked outside the United States, mainly in the Philippines. The main rollout occurred in autumn 2020 and winter 2021; publication followed in 2025. Agents could accept or reject the suggestions. [5]
The quasi-experimental analysis of staggered, non-random access estimates approximately 15% more issues resolved per hour. Benefits were greater among less experienced and initially lower-performing workers; some highly skilled agents experienced small quality reductions. The study lacked individual pay data and did not estimate the sector’s long-term employment response. [5]
This provides evidence of a contextual productivity benefit, subject to the study’s comparison and assumptions. It does not show what happens when an autonomous system replaces agents. Nor can that percentage be converted into a corresponding staffing reduction: service demand, quality requirements, checking work and the organisation of shifts can all change.
Employment signals need an explanation
Spain supplies a reason to investigate employment rather than assume its stability. CEX’s public account of its 2025 market study reports 65,859 workers in Spain, down 6.7%, and 12,375 offshore workers serving the Spanish market, up 12.5%. Participating companies represent around 75% of sector revenue, not necessarily employment. [6]
Participation is voluntary, the sample changed, and overseas activity was redefined. The full report was not examined here. These figures cannot isolate an AI effect or be joined mechanically to older editions. [6][7] A stronger assessment would follow comparable workplaces, technology adoption, service demand and contract movements, distinguishing fewer recruits from redundancies, reduced hours and relocation.
The ILO’s warning concerns this work
The joint UN–ILO report Mind the AI Divide explicitly discusses contact and call-centre work. It identifies business-process outsourcing as a source of formal, relatively well-paid employment in countries including India and the Philippines, particularly for women, while warning that technological advances may threaten some jobs. Exposure is not itself displacement. [8]
Its illustrative framework places routine call-centre operations among activities susceptible to substitution and associates that category with reskilling and social protection. The framework assumes smooth technological integration and acknowledges real infrastructure and cost constraints. This is an institutional assessment of risks and transition needs, not an estimate of redundancies in the EU or United States. [8]
Who should receive the gain?
Additional capacity could shorten queues, improve difficult cases, reduce working time or support fewer staff. Those are possible organisational responses, not measured shares. For customers, the test is dependable resolution and access to effective escalation. For workers, it includes pay, breaks, intensity, discretion and time to check an answer. For new entrants, it includes recruitment and supervised learning.
The Bible of AI (BAI) considers those outcomes part of the evidence required to describe technological change as progress in working life. A commercial benefit is worth measuring; it does not establish that benefits are shared. Equally, a reduction in repetitive work deserves recognition when it demonstrably improves people’s time, security or ability to help customers.
The present evidence supports a measured judgement: assistance can improve performance, while the distribution of benefits and the employment consequences remain to be established in each setting. A credible claim of success must follow what happens after the faster answer.
Scope and method
Focused narrative analysis; sources checked on 18 November 2026. Industry statistics, occupational projections, workplace estimates and institutional assessments are kept distinct. The workplace results have not been independently reproduced, and no interviews were conducted. The workplace study’s commercial setting and acknowledged Stanford Digital Economy Lab support should be considered when assessing the evidence. [5]
References
- U.S. Bureau of Labor Statistics. Customer Service Representatives. Occupational Outlook Handbook [online]. Updated 29 October 2026. Occupation SOC 43-4051; employment 2025, projections 2025–2035. Available from: BLS occupational profile.
- Eurostat. Enterprises by detailed NACE Rev. 2 activity and special aggregates [dataset]. Updated 30 September 2026. EU27_2020; N822; 2024; indicators ENT_NR, EMP_NR and NETTUR_MEUR. DOI: 10.2908/SBS_OVW_ACT. Exact data selection; SBS methodology. Flag e: estimated.
- U.S. Bureau of Labor Statistics. Quarterly Census of Employment and Wages: 2025 annual averages, Telephone call centers [dataset]. NAICS 56142; US000; private ownership; all establishment sizes. Available from: industry data. Definitions: annual-average fields, ownership codes and coverage and statistical units. Employment and establishment counts are annual averages.
- Executive Office of the President, Office of Management and Budget. North American Industry Classification System: United States, 2022 [online]. 2022. Section 56142, including 561421 and 561422. Available from: U.S. Census Bureau, official manual.
- Generative AI at Work. The Quarterly Journal of Economics [online]. 2025;140(2):889–942. Published online 4 February 2025. DOI: 10.1093/qje/qjae044. Setting: sections III.B–IV.C. Table II, column 3: 0.301 additional resolutions per hour, standard error 0.0329. Interpretation and limits: results and conclusion.
- Asociación CEX. Tecnología y deslocalización redefinen el mapa del sector según el Estudio de Mercado 2025 de CEX [online]. 27 July 2026. Available from: CEX’s public account. Methodology and employment sections; reported year-on-year changes.
- Asociación CEX. Estudio de Mercado CEX 2025 [online]. Undated public catalogue entry. Available from: publication and participation information. The complete report was not examined.
- United Nations, Office of the Secretary-General’s Envoy on Technology; International Labour Organization. Mind the AI Divide: Shaping a Global Perspective on the Future of Work [report]. UN/ILO; 2024. ISBN 9789211066524. DOI: 10.18356/9789211066524. Call-centre employment: p. 7; illustrative framework and transition measures: pp. 15–16. Official publication record.
Liked this post? Follow this blog to get more.