Nearly 90% of global companies attempting to implement artificial intelligence (AI) agents have failed, according to a recent survey. The rush to accelerate AI transformation (AX) has led to governance gaps, resulting in AI solutions that do not meet the actual needs of employees.
On August 19, McKinsey released findings from its '2026 AI Trust Maturity Survey,' which involved over 500 executives overseeing AI governance, risk management, and investment. As of early this year, only 33% of companies successfully adopted AI agents, with just 14% reaching large-scale production. The remaining 86% either stalled at the pilot stage or failed to implement AI altogether.
The primary reason for these failures is the emphasis on speed. Companies have pursued AI agent adoption across departments without adequately customizing solutions to reflect the specific workflows of each area. As a result, expensive AI systems are often rejected by employees, who prefer to use familiar tools, leading to decreased efficiency and security issues.
A similar warning was issued in Deloitte's '2026 State of AI in the Enterprise' report, which predicts that 74% of global companies will adopt AI agents by 2027. However, it notes that only 21% of these companies will have the necessary systems in place to manage and operate them safely. While the pace of adoption is increasing, the lack of supporting governance remains a significant concern.
In response to these challenges, more companies are adopting 'agentic orchestration,' which involves implementing multiple AI solutions tailored to different job functions under a clear governance framework. Ernst & Young has made its AI platform available as open-source to over 300,000 global experts, continuously updating it to meet the diverse needs of various organizations in tax, audit, and consulting.
JP Morgan has also integrated agentic orchestration into its large language models (LLMs), connecting multiple agents under a unified command structure to allow for flexible modifications and applications across departments. Similarly, Salesforce has orchestrated thousands of AI agents in operational environments, consolidating previously scattered agents into a single management system, resulting in an 84% reduction in processing time.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.