OutSystems: 26% Ready to Resolve Conflicts Between AI Agents as Agentic AI Moves into Wider Production Use
TECHWORLD ·
✦ AI Summary
OutSystems released research analyzing the gap between enterprise AI agent governance and real-world operational readiness.
In the report, 82% of corporate executives said they could maintain management, integration, and monitoring capabilities for AI agents, but only 26% said they were sufficiently prepared to resolve conflicts among agents, systems, and automated workflows.
The survey found that 83% of respondents use existing IT, security, and compliance processes, separate review processes, and centralized policy and monitoring frameworks, while only 5% of companies had all three: automated management functions, audit logs, and formal human review procedures.
OutSystems released research analyzing the gap between enterprise AI agent governance assessments and real-world operational readiness. The report is titled "Shifting to Agentic Systems for Business Outcomes," and OutSystems, CIO Dive, and KPMG collaborated on its publication.
The report found that companies are relatively confident in their overall ability to manage AI agents. 82% of corporate executives said they could maintain management, integration, and monitoring capabilities as AI agents are introduced into live operating environments.
At the same time, the report showed that readiness to respond to conflicts between AI agents and other systems remains far lower. Only 26% said they were sufficiently prepared to resolve conflicts among AI agents, systems, and automated workflows.
Against that backdrop, enterprise adoption of agentic AI has reached an important turning point. The report said companies are shifting from experimental use to company-wide, systematic adoption of agentic AI as development tools become more accessible and economic pressures intensify.
71% of corporate executives said they expect applications to take on a larger role over the next 5 years, with applications coordinating work across multiple systems, teams, and business processes or supporting more complex, multi-step workflows.
The report was designed to address these changes, analyzing the core capabilities companies need to build, manage, and orchestrate agentic systems and generate tangible business outcomes, while also offering insights.
Tiago Azevedo, CIO of OutSystems, said that while 82% of corporate executives said they have AI agent management capabilities, only 26% said they could resolve conflicts between agents. He said closing the gap between 82% and 26% is the key challenge, and explained that to ensure safety through experience building and operating AI agents across the enterprise, it is important to establish management controls at the platform level rather than relying on post-hoc checks.
He said that when predefined policies and standards can be applied during AI agent operations, repeated interruptions to verify the reliability of corporate systems become unnecessary and operating speed can improve.
The survey of more than 850 IT decision-makers also found a gap between perceptions of enterprise AI agent governance capabilities and actual operational readiness.
It also found that companies are already using a range of management frameworks. 83% of respondents said they use separate review processes and centralized policy and monitoring frameworks, in addition to existing IT, security, and compliance processes, to manage autonomous AI agents across applications and workflows.
However, the report found that these management methods still rely on people and manual procedures. Only 5% of companies had all three: automated management functions, audit logs, and formal human review procedures.
The report said the expansion of agentic AI is changing the main challenges companies face. As a result, converting existing governance policies into management frameworks that can be applied consistently across AI agent operating environments has emerged as a major task.
Companies' main challenges are shifting from developing individual agents to managing complex environments made up of various models, applications, data, and external tools. Rather than relying on a specific vendor solution, companies showed a tendency to prioritize flexibility and interoperability.
In addition, 74% of respondents identified integration with existing applications and systems as a major challenge. 63% emphasized orchestration of AI agents across systems, while 62% prioritized the flexibility to switch AI models, vendors, or platforms when needed.
The scope of enterprise validation is also expanding. Validation now ranges from individual agents to inter-agent interactions, exception handling, inter-task conflicts, recovery functions, and system-wide operations.
In this context, 81% of corporate executives said the ability to test, evaluate, and validate agent behavior is important when selecting AI platforms and tools.
The survey found that 35% of respondents identified maintaining data consistency and sharing context as the main challenge in orchestration. In response, OutSystems said that agent operation requires enterprise-governed data, application states, and business processes as the foundation. OutSystems said this points to the limits of AI agents' dependence on general model training information.
Companies said they value actual business improvements as the benchmark for evaluating the performance of agentic AI. Error reduction ranked first among key performance indicators, at 64%. The ranking was followed by compliance and risk reduction, employee productivity gains, and cost savings.
The share using employee productivity as a performance indicator was 73% in APAC survey markets Australia, India, and Japan. The share was 52% in Europe and 45% in North America.
By contrast, only 12% of respondents worldwide said agentic AI had improved the speed of launching new products and services or delivering new customer experiences. Business transformation across companies still appears to be at an early stage compared with operational improvements.
Among APAC respondents, 39% said they were sufficiently prepared to resolve conflicts between AI agents. The share saying they were sufficiently prepared to resolve conflicts between AI agents was 39% in APAC, 29% in Europe, and 18% in North America. The share saying they had all three—automated management functions, audit logs, and human review procedures—was also higher in APAC at 14%, compared with 3% in Europe and 2% in North America.
Michael Harper, Managing Director at KPMG, said that when it comes to these operational readiness issues, the point at which reliability and stability of agentic systems are secured is the early stage of development. He said post-implementation checks alone are not enough, and that existing log-based and procedure-centered management frameworks were designed for static software. He added that those frameworks have limitations in responding to a changing AI environment.
Michael Harper, Managing Director at KPMG, also said the gap between high confidence and actual operational readiness needs to be closed. To do that, he said policies need to be implemented in code, and safety standards must be applicable even during system operations. He added that this would make it possible to drive innovation centered on business outcomes while maintaining security and reliability.
Source: TECHWORLD · Lee Gwang-jae
Original: https://www.epnc.co.kr/news/articleView.html?idxno=408512
References
This article was produced with the help of an automated content generation algorithm.
Source: TECHWORLD
View originalThis article was summarized and organized by BizCrush based on the original article from TECHWORLD. For exact quotations and full details, please refer to the original article.