SAN FRANCISCO, Sept. 22, 2026 (GLOBE NEWSWIRE) -- New research from Guild.ai reveals a striking disconnect. 96.4% of IT decision-makers are confident their organization has a complete and accurate inventory of its AI agents. However, 66.7% of organizations with agents experienced an agent-related operational consequence in the past 12 months.
The gap becomes much clearer when you look at what organizations actually have in place. Only 42.7% have a centralized dashboard or monitoring tool, 39.8% have logging or audit trails, and just 31% can immediately stop a malfunctioning agent with an automated kill switch (76.5% can act within minutes when automated and manual responses are combined).
The AI Agent Management Gap highlights the widening disconnect between how fast organizations deploy AI agents and their ability to effectively govern, secure, and operate them in production, impacting operational efficiency.
Today, Guild.ai published The AI Agent Management Gap, a comprehensive survey and report of hundreds of U.S. IT decision-makers at mid- to large-sized companies, conducted by Morning Consult. The research shows that the challenge isn't whether enterprises will adopt agents (they already have). The challenge is building the infrastructure to manage what comes next.
AI agents are already everywhere
96% of organizations report having AI agents in production, and 47% of organizations with agents are running dozens or more.
The benefits are real, and organizations report improvements in productivity, quality, decision-making, response times, costs, and employee experience. As agents move from experiments into real-world production environments, traditional approaches to application management don't always translate.
Different teams can create agents, connect to systems, use models, and make decisions at runtime. Still, as their numbers grow, companies need visibility into not just which agents exist, but who owns them, what they can access, and what they're doing.
Confidence runs ahead of capability
The research found that 96.4% of IT decision-makers are confident their organization has a complete and accurate agent inventory. Yet only:
- 42.7% have a centralized dashboard or monitoring tool.
- 39.8% have logging or audit trails.
Even organizations that can respond quickly aren't necessarily catching problems early. 76.5% of organizations say they can stop a malfunctioning agent within minutes, whether through an automated kill switch (31%) or a manual process. Yet the speed hasn’t prevented the incidents: two-thirds of organizations with agents still experienced one in the past year, suggesting the harder problem isn't response speed once an issue is flagged. It's knowing something is wrong in the first place.
Shadow agents are an enterprise problem
The management challenge extends beyond the agents IT teams know about. 60.6% of IT decision-makers say it is very or somewhat likely that employees are deploying agents without formal approval. Only 35.9% of organizations require agents to be registered before deployment, and about half have a formal policy governing unauthorized agent use.
That combination, widespread suspicion paired with limited detection, is what makes unauthorized agent activity hard to manage. Most organizations believe it is happening. Few have the registration or monitoring in place to know for certain, or to catch it before an unapproved agent reaches production.
Agent sprawl creates an efficiency problem
The management gap isn't only about security. 61.2% of engineering and IT teams say they lose at least some time to duplicative agent work, rebuilding or maintaining agents that already exist elsewhere in the organization.
The problem becomes distinct as organizations scale. Companies with 20 to 199 engineers report agent-related incidents at nearly twice the rate of the smallest engineering teams – 78% versus 43%. More agents can create more value, but they also create more systems, integrations, permissions, and workflows to monitor and coordinate.
Enterprises recognize what is missing
The research shows organizations clearly see the gaps in their current agent infrastructure, even as many remain confident in their broader ability to manage AI agents long-term.
- 83.9% say real-time visibility into agent activity is extremely or very valuable.
- 81% consider a centralized hub for managing and deploying agents extremely or very important.
- 80% want security and compliance built directly into runtime tooling.
- 79.7% expect their organization's investment in agent-management infrastructure to increase over the next 12 months.
Recognizing a gap and acting on it are different. The same survey found that 93.2% of organizations feel prepared to own a long-term agent strategy, even though two-thirds have already experienced an agent-related incident. As AI agents become production infrastructure, the next 12 months will show whether that confidence turns into the operating layer enterprises say they value.
The next phase of enterprise AI
The first phase of enterprise AI was about proving what agents could do. The next phase is about operating them. Organizations need to know what agents are running, who owns them, what they can access, what they're doing, what they cost, and what happens when something goes wrong. Those are the problems Guild solves.
Read the complete findings and methodology in The AI Agent Management Gap Report.
This survey was fielded between August 4 – August 9, 2026 among 362 IT decision makers with 100+ employees. The interviews were conducted online, and the data was weighted to approximate a target sample of IT decision makers based on age, company size, industry and role. Results from the full survey have a margin of error of plus or minus 5 percentage points.
Media Contact
media@guild.ai
A photo accompanying this announcement is available at https://www.globenewswire.com/NewsRoom/AttachmentNg/c979978b-fb04-416d-a815-6545d5c4790b
