Gartner: 87% of Customers Demand Human Agent Access in GenAI Service

What happened

A new Gartner survey has found that an overwhelming majority of customers expect companies deploying generative AI in customer service to retain clear, accessible routes to human agents. Specifically, 87% of respondents stated that organisations using GenAI for service interactions must still provide access to a human representative — signalling that automation alone is insufficient to meet customer expectations.

The research underscores a growing tension in service design: whilst businesses accelerate the rollout of AI-powered service tools to reduce costs and improve response times, customers are drawing a firm line around human availability. Gartner's findings suggest that the appetite for AI-assisted service does not translate into an appetite for AI-only service.

Why it matters

For customer experience leaders and service designers, this data points to a critical design constraint that is easy to overlook when the business case for GenAI is framed primarily around efficiency. The behavioural economics principle at work here is well established: customers tolerate — and may even welcome — automated interactions when they perceive a safety net. Remove that safety net, and trust erodes rapidly, regardless of how capable the AI actually is. This is closely related to the concept of autonomy preservation: customers want to feel they can escalate, even if they rarely do.

For service operators in MENA and beyond, where relationship-based service cultures remain deeply embedded, the human-agent requirement is likely to carry even greater weight. Organisations that treat human escalation as a legacy cost to be eliminated, rather than as a structural trust signal, risk undermining the very confidence that makes AI adoption viable in the first place.

By the numbers

• 87% of customers surveyed by Gartner say companies using GenAI for customer service must provide access to a human agent.

The Renascence take

The instinct among many CX and operations teams will be to read this finding as a transitional problem — something that will resolve itself as customers become more comfortable with AI. That interpretation is likely mistaken, and acting on it could prove costly.

The demand for human access is not a symptom of AI scepticism that familiarity will cure — it is a permanent feature of how trust operates in service relationships. Behaviorally, the option to reach a human functions as a commitment device: it signals that the organisation remains accountable. What most operators will miss is that customers do not need to use the human channel frequently; they need to know it exists. The practical implication is straightforward — organisations should design human escalation as a visible, frictionless feature of any GenAI service layer, not as a fallback buried in a menu. Positioning it prominently may actually reduce the volume of escalations, because the reassurance itself does much of the work.