Will AI replace contact centre agents? What the evidence says, and how to plan

AI will replace some contacts, and in many contact centres a growing share of them. It is much less likely to replace the contact centre, or the need for skilled people in it. Simple, predictable requests such as checking a status, resetting a password or booking a slot are increasingly resolved by bots and AI agents, and that will reduce the number of people needed for that work. What remains for people is the contact that needs judgement, discretion, persuasion or empathy, and the work of designing, supervising and improving the AI itself.

That is not reassurance for its own sake. Contact centre roles have changed with every wave of technology since the automated call distributor, and each wave has moved simpler work away from people while raising the skill needed for what remains. AI is a bigger wave than most, and it is the first that can hold a conversation and act on it, so the change will be faster and deeper. But the evidence so far describes jobs being reshaped, and the organisations that plan for that will keep the people and the knowledge they depend on.

The most cited field study of AI in customer support, by Brynjolfsson, Li and Raymond, followed 5,179 support agents as a generative AI assistant was introduced. Access to the tool increased issues resolved per hour by 14 per cent on average, with a 34 per cent improvement for novice and low-skilled workers and minimal impact on the most experienced. The authors also found that AI assistance improved customer sentiment and increased employee retention, and suggest it spreads the practices of the best agents to newer ones. That is a picture of AI making agents better, not making them unnecessary.

The International Labour Organization's 2025 refined global index of exposure to generative AI reaches a similar conclusion at the level of the whole economy. Clerical occupations have the highest exposure, and customer service is named among the exposed roles, but the ILO finds that because most occupations consist of tasks that need human input, transformation of jobs is the most likely impact, and few jobs consist of tasks that are fully automatable with current technology.

Customers have a say too. In a SurveyMonkey study of 2,017 US adults in December 2025, only 8 per cent preferred AI over humans in customer service, and its 2026 report says 89 per cent believe companies should always offer the option to speak with a human. Preferences will shift as AI improves, and they differ by country and by type of contact, but no published evidence we have seen suggests customers are ready to give up the option of a person.

Thinking in tasks rather than jobs makes the change easier to plan. Every agent's day is a mix of tasks: understanding the request, finding information, updating systems, explaining, reassuring, negotiating and writing up. AI is already good at some of these and poor at others, and the mix varies by contact reason. The table shows how we see the balance today, journey by journey rather than role by role.

When AI resolves the simple contacts, the contacts that reach people change in character. They are longer, more complicated and more often emotional, because the customer has a problem the AI could not solve or a situation no customer wants to explain to a machine. Average handle time rises even as total workload falls, and the emotional load on each agent can increase. Agents need deeper product knowledge, more authority to resolve, and more support for the difficult conversations, and contact centre leaders need to measure them on outcomes rather than speed.

Copilots help with exactly this. Genesys describes Agent Copilot as determining customer intent, providing next best actions during the interaction, and producing a summary with suggested wrap-up codes afterwards, which removes much of the searching and writing that fills an agent's day. The research above suggests the gains are largest for newer agents, which matters when the contacts that remain are harder to learn.

• Conversation designers who shape how bots and AI agents talk and when they hand over
• Knowledge owners who keep the content AI depends on accurate and current
• AI quality analysts who review AI conversations as carefully as agents' calls
• Integration and tool owners who decide what AI agents may do in each system
• Specialist agents for complex, sensitive and high-value contacts

The organisations that handle this well plan it in the open. They forecast the contacts that will remain for people, not just the contacts AI will take, and recognise that those remaining contacts will take longer. They phase automation journey by journey, measure real containment before adjusting staffing, and use natural attrition, redeployment and retraining wherever they can. They involve agents early, because agents know better than anyone which contacts are simple and which only look simple, and they often become the best conversation designers and AI testers.

Workforce management needs to keep pace. Genesys Cloud workforce management forecasts and schedules from interaction history, and as AI resolves more contacts the history changes shape: volumes fall, handle times rise and the mix of skills shifts. Forecasts, schedules, service levels and the targets agents are measured against all need revisiting as each journey changes, not once a year.

AI that monitors or manages staff needs particular care. Under the EU AI Act, using AI to recognise agents' emotions from their voice is prohibited, and AI used to monitor and evaluate workers' performance becomes high-risk from 2 December 2027, with duties for the organisations that deploy it. Our guide to the EU AI Act and UK GDPR for contact centres sets out what applies and when. Whatever the law requires, agents should know how AI is used to assess their work and be able to question it.

1. Map contact reasons and decide, journey by journey, what AI will do
2. Forecast the remaining work for people, including longer handle times
3. Phase automation and measure real containment before changing staffing
4. Retrain and redeploy into specialist and AI-related roles
5. Revisit measures, schedules and coaching for harder contacts
6. Govern any AI that monitors or evaluates staff, and tell agents how it is used

QVCCS brings leadership in contact centre technology since 1984, through every wave from the automated call distributor to agentic AI, and our view is consistent: technology should take away the work people should not have to do, so they can do the work only people can. When we design AI into a Genesys Cloud CX contact centre, we start from contact reasons and outcomes, choose the simplest automation that resolves each journey, design the hand-offs so customers can always reach a person, and test with real calls. We help clients model the workforce effect alongside the cost, so the decision is made with both in view, and we involve the agents who will work alongside the AI from the start.