March 30, 2026
Consumers Are Already Letting AI Make Decisions for Them, EY Finds
Most conversations about AI still centre on whether people trust it. The more revealing question, according to new research from EY, is why that debate no longer seems to be slowing anyone down.
Sixteen percent of consumers across 23 markets have used AI that acts on their behalf without human involvement in the past six months alone, according to EY’s 2026 AI Sentiment Report, which surveyed more than 18,000 people across 23 markets.
In the same six-month window, 84% of respondents reported using AI in some form, suggesting that for many people the technology has become embedded in everyday life rather than used occasionally. But it is the autonomous dimension that carries the most weight for customer experience professionals. Ten percent of respondents have already used AI agents to purchase products on their behalf, 11% allow AI to automatically refill shopping carts or manage banking tasks, and 9% have used self-driving vehicles or autonomous taxis.
Familiarity Is Doing the Work That Trust Has Not
Adoption has not accelerated because consumers trust AI more than they did a year ago, but because low-stakes daily interactions, such as route planning, content recommendations, customer support, travel planning, have made people sufficiently comfortable to extend that comfort into higher-stakes territory. Familiarity is driving the transition from assistance to delegation.
Yet, 66% remain worried about AI systems being hacked or breached. A similar proportion (66%) says human oversight remains essential. Over 70% fear losing the ability to tell what is real from what is AI-generated. These are not niche anxieties but majority concerns held by people who are simultaneously adopting the very technology they mistrust.
Joe Depa, EY Global Chief Innovation Officer, said this is the defining challenge for organisations: “Trust will define the long-term winners in the AI economy, but today, adoption is moving faster than confidence. Organisations must earn trust through positive everyday experiences supported by reliable data, clear guardrails and accountability to close the gap between behaviour and sentiment.”
What Consumers Are Willing to Delegate
The report also maps consumer openness among those who have not yet used autonomous AI. More than a third (36%) say they would prefer AI to automatically apply discounts at checkout, and 34% would be comfortable having AI resolve customer service issues without their involvement. Thirty percent are open to AI managing home security, and 21% would delegate appointment scheduling.
Those consumers willing to let AI schedule appointments on their behalf are a direct signal to service teams and CX designers. The customer journey is no longer necessarily initiated by a human. As Gartner has previously forecast, agentic AI could autonomously resolve 80% of routine customer service issues by 2029, fundamentally changing who, or what, companies are actually serving.
Pioneer Markets Are Setting the Pace
The EY data shows adoption is not distributed evenly across the markets. India, the Chinese mainland, Brazil, Mexico, Saudi Arabia, the UAE, Hong Kong SAR, and South Korea are categorised as “Pioneer” markets where AI use is deeper, more frequent, and more autonomous than elsewhere. Ninety-four percent of people in these markets report using AI, and 24% have already used autonomous AI. These are the markets where the customer experience implications of agentic AI are playing out in real time.
The remaining markets in the study fall into “Transitional” and “Lagging” categories, trailing Pioneer markets by 12-15 percentage points on overall AI use and 11-13 points on agentic use, respectively. The pace gap between these groups is likely to narrow as low-stakes AI use continues to normalise globally.
How Service Design Has to Change
The data asks a question that goes beyond how to deploy AI in service delivery. It asks how to design for a world in which a growing number of customer interactions are initiated, mediated, and completed by AI agents acting on behalf of consumers, which requires more than operational efficiency. It requires clear escalation paths, transparent decision trails, and accountability structures that satisfy the two-thirds of consumers who still believe a human should be in the loop.
Voice-first agentic AI deployments from vendors including IBM and RingCentral point to one direction this design challenge is heading: agents that can handle complete customer interactions across natural spoken language, with the security controls and data residency standards that enterprise-grade deployment requires.
