VentureBeat Research: Where enterprise AI agent governance hasn't caught up

2026-07-24 · Source: VentureBeat · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Cybersecurity & Data Privacy · Depth: Intermediate, short

Summary

Enterprises have deployed AI agents without establishing adequate governance controls, a finding from five parallel VentureBeat Research surveys conducted in June 2026 across 573 respondents. These surveys, spanning identity, evaluation, cost telemetry, context layers, and orchestration, reveal that 57% to 68% of enterprises plan to switch or add vendors for these controls within 12 months. A significant portion, around one-third, intends to act within the quarter. Specific issues include 71% of deployed "agents" being simple chatbots, not multi-step agents, and two-thirds of enterprises allowing agents to push code to production based on automated evaluations despite only 5% trust in these systems. Furthermore, 69% of companies permit agents to share credentials, correlating with a 63.5% security incident rate. Over 80% of GPU users report less than 50% utilization, and 57% attribute confident, incorrect agent responses to poor business context data.

Key takeaway

For AI Architects or MLOps Engineers deploying AI agents, recognize that your current governance likely lags. You must prioritize retrofitting controls for identity, evaluation, cost, context, and orchestration. Implement scoped identity for agents touching production systems to mitigate security incidents. Validate agent evaluations against real production outcomes, not just internal benchmarks, before removing human review. Focus on optimizing existing GPU utilization and governing business context data to improve agent accuracy and reduce costs.

Key insights

Enterprises knowingly deployed AI agents lacking critical governance, now retrofitting controls and budgeting for vendor changes.

Principles

• True multi-step agents demand full control layers.
• Validate agent evaluations against production outcomes.
• Scoped identity for agents reduces security risks.

In practice

• Test agent evaluations against production outcomes.
• Implement scoped identity for all production agents.
• Govern business context data for agent accuracy.

Topics

• AI Agent Governance
• Enterprise AI Deployment
• Agent Security
• AI Compute Optimization
• Agent Evaluation
• Context Layers

Best for: CTO, VP of Engineering/Data, Investor, Director of AI/ML, MLOps Engineer, AI Architect

Related on AIssential

• Every AI Agent Is a Non-Human Identity That Needs Governance — AI agents are new principals requiring existing identity governance, not novel AI-native security solutions.
• Executive Interview: Vigilant AI.ai — Agentic AI deployment requires provable governance to move from pilot to production in regulated sectors.
• Why Runtime Enforcement Is Replacing Pre-Deployment Approval in Enterprise AI — Agentic AI necessitates a shift from pre-deployment approval to continuous runtime enforcement for auditable, regulated enterprise systems.
• Rolling out AI agents? 4 ways to move fast and furious - but with extreme caution — Balancing rapid AI experimentation with human oversight and robust governance is crucial for enterprise AI agent deployment.
• AI confidence just dropped 17 points in six months. That’s actually great news. — The drop in AI confidence signals a realistic understanding of production challenges, not a retreat from adoption.
• The DevOps guide to governing and managing agentic AI at scale — Autonomous AI agents demand integrated governance across their lifecycle to ensure reliability, security, and compliance.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by VentureBeat.