NEW YORK — North American retailers have moved artificial intelligence from the lab into the core of their operations, but their ability to govern it is struggling to keep pace, according to a new Coresight Research survey.
Released this week and sponsored by T-Mobile for Business, The Retail AI Playbook found that 85% of retail decision-makers say their data is growing at least somewhat faster than their ability to govern it. The report attributes this gap less to a lack of policies than to organizations' limited ability to put them into practice.
AI is no longer a side project. Surveyed retailers have AI in production across an average of 4.8 of their seven core business functions. Customer service leads at 83%, followed by fraud and loss prevention at 77% and supply chain optimization at 72%.
AI agents are spreading rapidly, too: 59% of respondents use them in production, while another 31% are piloting them. Drawing on data from spending platform Ramp, the report notes that 49% of US retailers had paid AI subscriptions in August 2026, up from 11% two years earlier. Coresight projects that the US agentic commerce market will grow from $13.3 billion this year to $175.8 billion by 2030, a 91% compound annual growth rate.
Data risks are a top concern for executives. Sensitive data leakage ranked among the three leading AI risks for 63% of respondents, while data privacy was the most frequently cited top concern, at 28%. Excessive autonomy among AI agents ranked third, cited by 42%.
Shadow AI—employees using unsanctioned tools—may indicate a blind spot. Only 26% ranked it among their top three risks, yet 59% said they had already found it in their organizations. Most retailers keep a hand on the wheel when using agents: 72% require human sign-off for consequential actions above set thresholds.
The stakes are high. According to IBM figures cited in the report, the average data breach in the US costs $11.5 million.
Coresight's central argument is that governance accelerates AI rather than slowing it down. Among retailers with formal programs, 67% reported faster deployment approvals, while 63% said customer trust had improved. Building these programs takes time, however: 92% said it took six months or more.
The main obstacle is technical, not political. Complexity around AI-specific controls was the leading barrier, cited by 39%, while only 1% pointed to a lack of executive buy-in. Just a third of retailers manage AI through a single enterprise-wide control plane; 61% use hybrid setups, where rules may be applied unevenly.
Coresight urges retailers to scale AI gradually, starting with high-value, manageable-risk applications and tightening oversight of agent permissions as autonomy increases. "We believe the next competitive advantage will come from operationalizing trust," the report concludes.
The survey polled 100 North American retail executives at companies with at least $500 million in annual revenue, Sept. 2–4, 2026. Its margin of error is ±8 percentage points.
