The Customer Service Paradox: More AI, More Value for People
Quality, costs and the role of partners in tomorrow’s customer service
When a company outsources its customer service, the discussion usually starts with two things: quality and cost. It wants customers to receive a good service at a sustainable price. The problem begins when both are measured too narrowly: quality through a handful of response metrics, and cost through an hourly rate or the price of a call.
I believe artificial intelligence will force us to rethink this approach. More work can be automated, and that will change what companies should pay a customer service partner to do. To understand where value will come from, we first need to look at what companies and their customers are actually buying.
What determines service quality
The first element is availability: customers must be able to reach the service at useful times, it must absorb peaks in demand and meet agreed service levels (SLAs). Extended coverage matters when customers need it. A number that is active around the clock but only records messages is not the same as a service that can take action.
The second element is the ability to resolve issues. Product knowledge, accurate information and the ability to carry out the necessary steps all matter. An agent may answer quickly and politely, but if the customer has to call three times to get a solution, overall quality is still low.
The third is the relationship. Customers need to explain the problem, feel understood and know what will happen next. Language, clarity and emotional awareness are especially important when someone is worried, has already lost time or has received conflicting answers. The commissioning company must also consider the partner’s ability to adapt to volumes, integrate with its systems, keep information up to date and take clear responsibility for operations. Data protection and continuity when a system fails matter too.
Deloitte’s 2024 Global Outsourcing Survey of more than 500 executives places access to skills and agility alongside cost reduction. The survey covers outsourcing as a whole, not customer service alone.[1] There is no universal percentage to assign to each quality factor. A customer seeking an invoice has different needs from someone whose equipment has stopped working. Some requirements are also a baseline: if the service cannot be reached or give a correct answer, the other benefits lose value.
There are useful indicators, though. Qualtrics’ Contact Centre Trends 2025 study, based on more than 23,000 consumers, found that first-contact resolution was associated with a 2.1 times greater likelihood of recommending a company. Satisfactory wait times were associated with a 2.6 times greater likelihood of increasing purchases. These are survey associations, not guaranteed revenue multipliers.[2] Companies should therefore connect quality to outcomes: issues resolved, repeat contacts, rework and customer retention. Speed is only part of the result.
The economic model that built the sector
For many years, outsourcing created savings by sharing resources. Centralized operations spread technology investment and overhead across several clients, while shift planning, peak management and existing processes added further efficiencies. Yet in a labour-intensive activity, people remain a major cost. ICMI has cited staff as roughly 70% of contact-centre costs, based on an industry reference published in 2019 and updated in 2022.[3] TP’s 2025 accounts provide a more recent example: personnel costs of €6.95 billion, around 75.8% of the operating-cost items shown in its income statement. That is a calculation for the consolidated group, not an industry average.[4]
The exact share varies, but the economic point is clear: when labour carries this much weight, differences in labour costs can significantly affect an offer’s competitiveness. Eurostat data published by Destatis put average hourly labour costs in 2025 at €45.00 in Germany, €44.30 in France, €32.00 in Italy, €26.40 in Spain, €19.10 in Poland, €13.60 in Romania and €12.00 in Bulgaria. These are employer costs across market-economy sectors B–N, not contact-centre wages or outsourcing rates.[5]
Outside Europe, the US Bureau of Labor Statistics reported median hourly pay of $21.53 for customer service representatives in May 2025.[6] The Philippine Statistics Authority reported average monthly pay of 24,979 pesos for voice-call-centre information clerks in August 2024.[7] These figures use different periods and definitions, so they cannot be compared directly with European employer costs. They help explain interest in offshoring, but geography alone does not measure quality. The problem arises when the lowest rate comes at the expense of training, team stability, language coverage or operational autonomy.
The price of a contact and the value of a solution
Billing practices reflect this economic structure. TTEC’s 2025 annual report says most of its inbound and outbound outsourcing contracts are based on minutes, hours, full-time-equivalent resources, transactions or calls.[8] The customer buys capacity or a quantity of work. The price covers people, management, technology, service upkeep and the provider’s margin. The client’s value comes from what the work achieves: resolved problems, less rework, continuity and preserved customer relationships.
Consider a simple example. If a contact costs €3 and three contacts are needed to resolve a case, that case costs €9. A service charging €5 per contact is cheaper if it resolves the same case on the first attempt. Looking only at the price list could lead to the more expensive choice. Cost per contact should therefore be considered alongside total cost per resolved case, including follow-ups and corrections. The service can also show a company recurring problems, unclear instructions and process defects, helping remove the reasons customers need to get in touch.
AI and the cost of service
AI changes the relationship between the work required and the service delivered. When information is available, rules are clear and actions can be checked, more of the process can be automated. In these cases, the advantage of access to lower-cost labour may shrink. Klarna reported in its 2025 accounts that its AI assistant handled 80% of customer-service chats. This is a company-reported figure for one channel; it does not mean that 80% of problems were fully resolved without human help.[9]
AI can also support employees. A 2025 study, “Generative AI at Work,” analysed 5,172 agents and found that an AI assistant increased the average number of issues resolved per hour by 15%. Agents remained responsible for their conversations: the study measured support for human work, not its replacement.[10] My expectation is that prices for highly standardized work will face strong pressure. Providers whose offer depends mainly on wage differences will need to rethink their service. The same technology can also improve productivity in offshore markets, so the ability to adapt will matter more than location.
AI costs include integration, maintenance, quality checks and exception handling as well as model usage. Savings are real only when the total cost falls while the result stays the same. A system that gives cheap answers but creates errors or repeat contacts may raise the cost of service. Even a significant reduction in working hours could put pressure on a model built mainly around selling those hours.
The partner who designs and maintains the service
An advanced partner’s first contribution will be understanding the client’s business and turning that knowledge into a service that works. This means studying products and services, tracing why customers make contact and identifying the information needed to respond. The partner can then define processes and decide what to automate.
The partner must integrate the service with company systems, prepare people and check automation in real situations. On the phone, that includes recognizing names, responding at the right pace and handling interruptions. If a system says it has opened a request, a control must confirm that the request was actually recorded. The work continues after launch: products, procedures and expectations change; knowledge must be updated, errors analysed and handovers made with the information already collected. This ongoing operational responsibility should be visible in shared results and checks.
The 2025 UK Contact Centre Outsourcing Report describes growing demand for services combining design, operations and technology, and considers how automation might free resources for selective investment in local support.[11] A company may build these capabilities internally if it has the necessary skills, volumes and systems. A fair comparison with outsourcing should include launch time, development and maintenance costs, accountability and the ability to keep improving. A provider should show a concrete advantage in running the whole service.
The value of people when it matters
If automation handles some requests effectively, more human expertise can be assigned to situations that need it. A single human interaction may cost more because the agent handles harder cases; sustainability must be judged across the whole service. In a 2026 Gartner survey of 3,566 B2B and B2C customers, 87% said it was essential to be able to reach a human agent when a company uses generative AI. In the same study, 50% said AI makes interactions easier. Human access can therefore coexist with a positive view of technology.[12]
The agent in this model needs to understand the client’s products and processes, use an up-to-date knowledge base and have the authority to act. They must spot exceptions, know when to involve a technician and take responsibility for the next step. Language is also practical: in a 2020 CSA Research survey of 8,709 consumers across 29 countries, 75% said they were more likely to buy again from the same brand if support were available in their own language. This concerns the language of support, not the agent’s location.[13]
Native-language teams who know the market can better understand customer habits, retailers, technical support and expectations about timing. Technical preparation must be matched by listening, clear communication and credible follow-through. Empathy is useful when it comes with a solution or a practical route to one. Continuous training, stable teams and good working conditions are necessary, especially when people are left with the more complex cases.
The contract must recognize this value
If a provider is paid only for hours worked, an improvement that reduces those hours may also reduce its revenue. The client and partner need an agreement that makes better service worthwhile. It may help to distinguish design and integration, ongoing service management and the operational capacity required. Part of the fee can be linked to agreed outcomes if they are measurable and the partner can genuinely influence them.
Counting closed tickets is not enough. The parties need to define what “resolved” means, how reopened cases are handled and what quality levels must be maintained. AI-handled requests should be considered alongside errors, repeat contacts and satisfaction. The economic goal is to make it worthwhile to resolve the customer’s whole problem effectively.
The model we offer
This outlook is close to the model we are developing at VOICE&WEB: a service built around the client’s needs, knowledge of its products and care for the people who interact with its brand. Our European organization, with hubs in Milan, Klagenfurt, Lyon and Barcelona, combines language and market expertise with technology integrated into clients’ processes.[14]
AI can make this approach more sustainable. Automating suitable work well can free resources for agent training and cases that need more attention. We want to keep working on this balance, measuring efficiency and the quality of the outcome. We believe this evolution can make a model built on people’s expertise and customer knowledge more competitive: understand the company we represent, shape the service around its needs and care for its customers with prepared people and reliable technology.
Sources and data notes
Sources consulted on 25 October 2026. Bracketed numbers in the text refer to the sources below. Statements about the sector’s future are the author’s views.
1. Deloitte | Global Outsourcing Survey 2024 — More than 500 global executives; outsourcing across several functions.
2. Qualtrics | Contact Centre Trends 2025 — Published 2 December 2024; over 23,000 consumers and 3,072 employees. Reported associations do not establish causation.
3. ICMI | The Metrics of Contact Center Productivity — 26 February 2019, updated 15 June 2022; industry indication, not a universal 2026 estimate.
4. TP | Consolidated Financial Statements 2025 — Income statement, PDF p. 4; calculation: 6,950 / 9,166 = 75.8%.
5. Destatis / Eurostat | Labour cost comparison across EU countries — 2025 data, updated 29 April 2026; employer cost per hour across sectors B–N.
6. BLS | Customer Service Representatives — Median hourly pay, May 2025; not total employer cost.
7. Philippine Statistics Authority | Occupational Wages Survey 2024 — PDF p. 12; August 2024, establishments with at least 10 employees.
8. TTEC | Annual Report 2025 — PDF p. 49, Revenue Recognition.
9. Klarna | Annual Report 2025 — Company-reported data on chats handled by its assistant.
10. Brynjolfsson, Li and Raymond | Generative AI at Work — Quarterly Journal of Economics, 2025, 140(2), 889–942.
11. CCMA and The Knowledge Group | UK Contact Centre Outsourcing Report 2025 — pp. 4–5, with Everest Group contribution.
12. Gartner | Access to a human agent in GenAI customer service — 4 August 2026; 3,566 B2B and B2C customers surveyed.
13. CSA Research | Consumers Prefer their Own Language — 7 July 2020; 8,709 consumers in 29 countries.
14. VOICE&WEB | Service model and European organization — Company source for hubs and operating model.
Article by Andrea Macario.
Voice & Web works across Customer Care, Customer Satisfaction, CRM and e-commerce processes to improve the customer experience.
Did you like what you read?
Hi, I'm Simona. Contact me to find out how Voice&Web can help your company.