OutSystems, in research conducted with CIO Dive and KPMG, has identified a gap between enterprise confidence in AI agent governance and the ability to manage autonomous systems in production. The report, Engineering Business Outcomes: The Shift to Agentic Systems, found that 82% of surveyed executives believe their organizations can keep pace with AI governance, but only 26% say they are well prepared to resolve conflicts across agents, applications, and automated workflows. The findings highlight the growing need for runtime controls as businesses move from AI experiments to interconnected agentic systems.
OutSystems Report Flags Gaps in Enterprise AI Agent Governance
Enterprise adoption of agentic AI is shifting from isolated experiments toward systems that can coordinate tasks across applications, business processes, and teams. That transition introduces a governance challenge: organizations must manage not only individual AI agents but also the interactions, permissions, and decisions that emerge when multiple agents operate across interconnected systems.
A new report from OutSystems, developed in collaboration with CIO Dive and KPMG, highlights the gap between executives’ confidence in AI governance and their operational readiness to manage these systems. The survey of more than 850 IT decision-makers found that 82% believe their organizations can keep pace with governing, integrating, and monitoring AI agents. Yet only 26% say they are well prepared to resolve conflicts across agents, systems, and automated workflows.
The findings suggest that enterprise AI governance needs to evolve beyond written policies and periodic reviews toward controls that operate while agents execute tasks.
Governance confidence exceeds operational readiness
According to the report, 83% of respondents rely on existing IT, security, or compliance processes, specialized reviews, centralized policies, or monitoring to oversee AI agents capable of autonomous action. However, just 5% report having automated controls, audit trails, and formal human review points in place.
This distinction matters as agents gain the ability to trigger actions across business applications. A conventional AI assistant may generate a recommendation for an employee to review. An agentic system can go further by calling tools, updating records, initiating workflows, or coordinating with other agents. Those capabilities create additional risks when permissions conflict, information becomes inconsistent, or one automated action affects another process.
The report points toward runtime governance, in which policies and safeguards are enforced as agents operate rather than relying exclusively on human review after an event. This can include controlling access to tools, validating proposed actions, recording decisions, and providing mechanisms to stop or reverse problematic workflows.
OutSystems CIO Tiago Azevedo argues that controls should be built into the platform rather than added after development. KPMG Managing Director Michael Harper similarly emphasizes policy-as-code and runtime guardrails as ways to maintain oversight while enabling enterprise automation.
Integration and orchestration become priorities
The report also indicates that enterprise agentic AI is likely to involve diverse models, applications, data sources, and third-party tools rather than a single consolidated platform.
Seventy-four percent of executives identify integration with existing applications and systems as a priority. Meanwhile, 63% prioritize agent orchestration across systems, and 62% value the ability to change models, vendors, and platforms over time.
These priorities reflect the complexity of deploying AI in established enterprise environments. Organizations need agents to work with existing business systems while preserving data access rules, application permissions, and operational requirements. They also want flexibility to adopt new models without rebuilding every workflow.
Interoperability, however, does not automatically ensure reliability. When multiple agents exchange information or execute interdependent tasks, organizations must account for conflicting instructions, duplicated actions, incomplete handoffs, and failures that propagate across workflows.
Enterprise context is another challenge. Thirty-five percent of respondents identify data consistency and context sharing as a major orchestration hurdle. Agents need access to relevant, governed business information and application states, not just the general knowledge encoded in a model.
Testing must cover complete workflows
The research suggests that organizations are increasingly looking beyond the performance of individual agents to assess the reliability of entire agentic systems. Eighty-one percent of executives consider testing, evaluating, and validating agent behavior very or extremely important when selecting AI platforms or tools.
For enterprise deployments, that means testing how agents interact with one another, handle exceptions, respond to conflicting actions, and recover when a process fails. A model that performs well on isolated tasks may still produce unreliable results when embedded in a longer workflow involving multiple applications and decision points.
Effective testing also needs to reflect changing conditions. Updates to models, prompts, data sources, permissions, or connected tools can alter system behavior, making ongoing evaluation important after deployment.
Business outcomes remain the measure of success
The report found that error reduction is the leading measure of agentic AI impact, cited by 64% of respondents. Compliance and risk reduction, employee productivity, and cost reduction are also important measures.
However, only 12% said agentic AI had accelerated the launch of new products, services, or customer experiences. The contrast suggests that organizations may be finding more immediate value in improving existing operations than in creating entirely new business capabilities.
OutSystems is positioning its Agentic Systems Platform around building, orchestrating, and governing enterprise agentic systems across complex technology environments. Whether platforms can close the readiness gap will depend on their ability to translate governance policies into enforceable controls, provide visibility across connected workflows, and demonstrate measurable business results.
The broader lesson is that scaling agentic AI requires more than capable models. Enterprises need reliable integration, governed context, workflow-level testing, and safeguards that remain active as autonomous systems operate. Without those foundations, confidence in AI governance may continue to outpace the ability to manage agents in production.
Market Landscape
Enterprise AI is moving from standalone copilots toward agentic systems that can coordinate work across applications and business processes. This shift increases the importance of orchestration, data governance, interoperability, runtime monitoring, and end-to-end testing.
OutSystems is positioning its Agentic Systems Platform around the development and governance of connected AI workflows. The report’s findings point to a market requirement that extends beyond model performance: enterprises need controls that can manage interactions among agents, connected software, and business rules.
The competitive opportunity lies in making agentic AI manageable at enterprise scale. Organizations will need to evaluate policy enforcement, auditability, integration flexibility, exception handling, and measurable operational outcomes when selecting platforms. The survey reflects reported executive perceptions rather than an independent technical assessment of individual vendors.
Top Insights
- OutSystems’ research found that 82% of executives feel confident about AI governance, but only 26% report readiness to resolve conflicts across connected agents and workflows.
- Only 5% of respondents report automated controls, audit trails, and formal human review points, highlighting gaps in operational AI governance.
- Integration with existing applications is a priority for 74% of executives, while 63% emphasize orchestration across systems.
- Eighty-one percent consider testing and validating agent behavior highly important, reinforcing the need to evaluate complete workflows rather than individual models.
- Error reduction leads reported agentic AI impact metrics at 64%, but only 12% say agentic AI has accelerated new product or service launches.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
