New analysis reveals agentic AI use rising greater than fivefold as staff more and more handle a number of brokers and delegate day-long assignments
Corporations are quickly including a brand new form of employee to their organizations.
It may entry firm knowledge and enterprise methods. It may carry out complicated assignments. It could work constantly and take actions on behalf of staff.
However in lots of organizations, nobody has clearly outlined its job, who manages it, how a lot authority it has, or how its entry ought to be revoked.
That employee is an AI agent.
Corporations would by no means give an worker broad entry to company methods, present no clear duties, and hope an audit log explains the results afterward, however they’re doing that with AI.”
— Adam Harriss
DoubleU.ai, Inc., the corporate behind DBLU, says companies want to start treating AI-agent administration as an organizational self-discipline, not merely a software program deployment downside. The warning follows new analysis, The Shift to Agentic AI: Proof from Codex, which discovered that energetic customers of OpenAI’s Codex elevated greater than fivefold through the first half of 2026.
The research additionally discovered that greater than 10% of customers handle three or extra concurrent brokers throughout a typical week. For the reason that starting of 2026, the share of customers assigning no less than one job estimated to require greater than eight hours of skilled human work elevated almost tenfold.
Individuals are now not merely asking AI questions.
They’re delegating work to it.
Additionally Learn: AiThority Interview with Matej Bukovinski, Chief Know-how Officer at Nutrient
“Corporations are starting to handle AI staff earlier than they’ve constructed a administration system for AI,” mentioned Adam Harriss, Founder and CEO of DoubleU. “Each significant AI agent ought to have the equal of a job description, an identification, a supervisor, an outlined objective, an authority stage, approval limits, and a report of what it did.”
Organizations have spent many years constructing methods for human authority. Workers have identities, roles, managers, permissions, insurance policies, approval limits, and onboarding and offboarding processes.
AI brokers more and more carry out work, however many companies can’t reply equal questions:
– Who’s the agent appearing for?
– What job has it been assigned?
– What knowledge and methods can it entry?
– What choices can it make?
– What requires human approval?
– Who’s accountable when it makes a mistake?
– How is its authority revoked?
In line with IBM Institute for Enterprise Worth analysis performed with Oxford Economics, “87% of surveyed executives say they’ve clear AI governance frameworks, however fewer than 25% have absolutely applied and constantly assessment instruments to handle dangers corresponding to bias, transparency, and safety.”
Corroborating these findings, latest DigiCert analysis discovered that whereas 90% of organizations have mentioned AI governance on the government or board stage, solely half have established formal AI governance packages. The identical analysis discovered that solely 53% can absolutely hint AI outputs again to the underlying fashions and supply knowledge.
“Corporations would by no means rent an worker, give that individual broad entry to company methods, present no clear duties, and hope an audit log explains the results afterward,” Harriss mentioned. “But that’s remarkably near what number of organizations are starting to deploy AI.”
DBLU is constructing a Belief Layer designed to function at runtime and assist organizations handle AI identification, objective, permissions, delegated authority, human approvals, ruled reminiscence, and auditability.
The corporate argues that enterprise software program was constructed round a easy assumption: a human was behind each essential motion. AI brokers break that assumption.
“The following administration problem is just not merely how one can use AI,” Harriss mentioned. “It’s how one can handle a workforce that’s now not solely human.”
Adam Harriss is on the market for interviews on the emergence of AI staff, the dangers of unmanaged AI brokers, and why firms might have a brand new administration mannequin for non-human labor.
Additionally Learn: AI methods – Interoperable AI methods: Connecting fashions throughout platforms
[To share your insights with us, please write to psen@itechseries.com]
