June 04, 2026
Autonomous AI Lost the Contact Centre to Humans and AI, Study Reveals
Across six customer-facing functions, enterprises have settled on the same answer to how they run service, and it is not what the autonomy narrative predicted. Human agents working with AI rank as the most-cited approach everywhere, from technical support to billing, and they outpaced fully autonomous agentic AI in every one.
The finding comes from Enterprise CX AI: 2026 Global Survey, a study of 815 enterprise CX decision-makers that Ryan Strategic Advisory conducted for TELUS Digital across 12 countries and 19 industry verticals. Respondents chose from four service approaches for each function and could select all that applied. The human-plus-AI model won outright in all six.
The Preferred Model
The numbers leave little room for interpretation. Human agents assisted by AI lead in technical support, customer retention and winback, onboarding, revenue generation, complaint management and billing. This approach outperformed human-only service, basic automation and fully autonomous agentic AI in every category.
The result runs against an industry that has spent two years promoting machines that resolve issues end to end without a person in the room. Gartner has predicted agentic AI will autonomously resolve 80% of routine customer service issues by 2029, and vendors have built their roadmaps around that destination. The enterprises buying and running these systems are telling a different story about what works today.
Gartner also expects more than 40% of agentic AI projects to be cancelled by the end of 2027, citing escalating costs, unproven business value and weak risk controls. The same firm estimates that only around 130 of the thousands of self-described agentic AI vendors offer the real thing, with the rest engaged in “agent washing.“
Researchers at Carnegie Mellon University and Salesforce have measured successful task completion for AI agents on multi-step tasks at roughly 30%. For a customer interaction that has to finish correctly, a coin-flip success rate is not a foundation enterprises can build service on. AI handles retrieval, drafting and routing at speed, and a person owns the judgement, the exceptions and the close.
That preference holds because AI handles routine queries at speed but still struggles with the emotional, complex conversations that need a person.
A Win Still Needs Proving
The same survey carries a warning for the model it endorses. Only a third of enterprises run AI-powered quality assurance and coaching tools, so two-thirds have no automated way to assess how their AI-assisted interactions actually perform. Companies have committed to a human-and-machine partnership without building the layer that tells them whether it works.
Over half of of surveyed organisations spend more than $10 million a year on CX delivery, and average handle time reduction registered as a priority for just 19% of them, a sign the market has moved past efficiency-only thinking toward quality and consistency.
The future of service is neither all-human nor all-machine. It is both, working together, and the autonomy that dominated the conversation did not lead a single function that enterprises measured.
