New research shows use of agent AI is increasing more than 5x as employees increasingly manage multiple agents and delegate day-long tasks
Companies are rapidly adding new types of employees to their organizations.
Access corporate data and business systems. You can perform complex assignments. It operates continuously and may perform actions on behalf of the employee.
But in many organizations, no one clearly defines what that job is, who controls it, how much authority they have, and how that access should be revoked.
That worker is an AI agent.
Companies never give their employees broad access to corporate systems, don’t give them clear responsibilities, and expect audit logs to explain subsequent outcomes, but they’re doing it with AI. ”
— Adam Harris
DoubleU.ai, Inc., which backs DBLU, says companies need to start treating AI agent management as an organizational discipline, not just a software implementation issue. This warning follows a new study, “The Shift to Agentic AI: Evidence from Codex,” which found that active users of OpenAI’s Codex increased by more than five times in the first half of 2026.
The study also found that more than 10% of users typically manage three or more agents simultaneously in a week. Since the beginning of 2026, the percentage of users assigning at least one task estimated to require 8 or more hours of experienced human effort has increased nearly 10x.
People no longer just ask AI questions.
They are delegating work there.
Also read: AiThority interview with Matej Bukovinski, Chief Technology Officer at Nutrient
“Companies are starting to manage their AI workforce before they build systems to manage AI,” said Adam Harris, founder and CEO of DoubleU. “Any meaningful AI agent must have the equivalent of a job description, an identity, a manager, a defined purpose, authority levels, authorization limits, and a record of what it is doing.”
Organizations have spent decades building systems for human authority. Employees have identities, roles, managers, permissions, policies, approval restrictions, and onboarding and offboarding processes.
AI agents are increasingly performing tasks, but many companies are unable to answer comparable questions.
– Who does the agent represent?
– What kind of work was assigned?
– What data and systems do you have access to?
– What decisions can you make?
– What needs human approval?
– Who is responsible if a mistake is made?
– How is that privilege revoked?
According to a study conducted by the IBM Institute for Business Value in collaboration with Oxford Economics, “While 87% of executives surveyed say they have a clear AI governance framework, fewer than 25% of companies have fully implemented and continually review tools to manage risks such as bias, transparency, and security.”
Supporting these findings, a recent DigiCert study found that while 90% of organizations are discussing AI governance at the executive or board level, only half have established a formal AI governance program. The same study found that only 53% can fully trace AI output back to the underlying model or source data.
“Companies will never hire employees, give them broad access to corporate systems, give them clear responsibilities, and then expect the results to be explained in audit logs,” Harris said. “Still, this is surprisingly close to the number of organizations that are beginning to implement AI.”
DBLU operates at runtime and builds a trust layer designed to help organizations manage AI identity, purpose, authority, delegated authority, human authorization, managed memory, and auditability.
The company claims that its enterprise software is built on the simple premise that there is a human behind every significant action. AI agents break that assumption.
“The next business challenge is not just how to leverage AI,” Harris says. “It’s about how you manage a workforce that is no longer fully human.”
Adam Harris joins us for an interview about the emergence of AI workers, the risks of unmanaged AI agents, and why companies need new management models for non-human workers.
Also read: AI Systems – Interoperable AI Systems: Connecting models across platforms
[To share your insights with us, please write to psen@itechseries.com]
