AI Agents: Enterprises Deploying Despite Known Risks, Study Reveals

A new report from Economist Enterprise, titled "Power without control: Rethinking cybersecurity for the age of agentic AI," uncovers a critical paradox in the corporate world: enterprises are aggressively deploying AI agents despite a clear understanding of the inherent and often unmanaged risks. The study, which surveyed over 800 business decision-makers across nine countries, found that a staggering 98% of organizations have already encountered a disruptive incident directly attributable to AI agents. The findings indicate that competitive pressures are forcing organizations to accelerate the adoption of these autonomous AI systems, even as 90% of leaders admit they are deploying agents at a pace exceeding their security teams' ability to evaluate or govern them. This rapid deployment without adequate oversight is leading to a landscape where disruption is not merely a risk to be avoided, but an inevitable reality that enterprises must prepare to manage. AI agents introduce fundamentally new categories of risk that existing cybersecurity frameworks were not designed to handle. These risks extend beyond traditional IT concerns, potentially leading to severe business consequences such as regulatory fines, supply chain disruptions, significant revenue loss, and reputational damage. The report underscores that agentic risk is primarily a business problem, not solely a technological one. Despite widespread awareness of these dangers, organizational preparedness remains low. Two-thirds of surveyed organizations lack full visibility into the actions of their AI agents, making it difficult to track decisions, data access, and actions. Furthermore, only 30% possess robust, fully tested rollback capabilities, and 43% report that their recovery processes do not comprehensively cover all agent types or incident scenarios. This lack of visibility and recovery capability extends to the boardroom, where only one in four organizations regularly reports on cyber recovery performance, leaving senior leadership ill-equipped to make informed decisions about AI investments. The study concludes that in the age of agentic AI, the focus must shift from preventing all incidents to effectively managing and recovering from inevitable disruptions.