2026 ResearchMorning Consult362 U.S. IT decision-makers
How enterprises are scaling AI agents faster than they can see, secure, and control them
AI agent adoption is no longer the question. Enterprises are already deploying agents across engineering, sales, support, marketing, and other functions, and fleets are growing quickly. Most are already seeing real value from it.
The challenge is keeping visibility, ownership, security, and cost control in step with that growth. Organizations feel confident they know what agents they have and how they’re being managed. The underlying infrastructure tells a different story.
What the research found
Policy is ahead of practice. 96% of enterprises run agents in production. No function owns them. Only four in ten log what those agents actually do.
Security is the central risk. Nearly everyone feels confident in their oversight. Two out of three have had an incident anyway.
Closing the gap drives growth. Real-time visibility is what enterprises want most. Eight in ten plan to spend more to get it.
Visibility
Almost every organization believes its agent inventory is complete. Far fewer have the infrastructure to prove it.
of organizations feel confident their agent inventory is complete — but only 42.7% have a centralized dashboard to back that up.
Ownership
Agents span engineering, sales, support, and marketing, but ownership remains fragmented.
is the largest share any single function — engineering/IT — claims as primary owner. No function claims a majority.
Detection
Most leaders suspect employees are already deploying agents without formal approval. But suspicion isn't detection. Without registration and centralized visibility, organizations can't reliably know what's running inside their environment.
say employees are likely deploying agents without formal approval — but only 36% require agent registration before deployment.
Sprawl
As organizations add more agents, teams increasingly duplicate work that already exists elsewhere. Without a way to discover and reuse existing agents, agent growth creates duplicated engineering effort & unnecessary cost.
of engineering and IT teams lose time to duplicative agent work.
Security is the clearest example of the gap between perception and reality. It is both the top operational concern and a major source of resistance to broader agent adoption.
Organizations feel remarkably prepared to manage their agents.
feel prepared to own a long-term agent strategy.
But that confidence doesn't always translate into control.
have experienced an agent incident in the last 12 months.
The research points to a clear set of priorities for the next phase of agent adoption.
01
Real-time visibility
Know what agents are running, what they're doing, and how they're performing in production.
02
Agent registration
Know what is being deployed before it enters the environment, rather than discovering agents after the fact.
03
Permissions and controls
Define what agents can access and what actions they can take.
04
Cost controls
Understand where agent spend is going and prevent unnecessary duplication and runaway usage.
Five chapters covering how enterprises are deploying, managing, securing, and preparing for an increasingly autonomous agent fleet.
AI Agent Adoption and Landscape
Scale and maturity of adoption, including agent counts, deployment models, ownership, staffing, and use cases.
Visibility, Oversight and Governance
Inventory confidence, formal controls, kill-switch readiness, and the incidents that undercut self-reported confidence.
Unauthorized Agent Activity
How likely and how concerning shadow AI is, and why detection still lags well behind suspicion.
Organizational Readiness and Adoption Challenges
The biggest management gaps, employee resistance, and the real cost of duplicated agent work.
Future Needs and Priorities
What organizations want next: centralized visibility, permissions, audit logs, cost controls, and model-neutral infrastructure.