What is Agent Chaos?
Agent Chaos is the condition in which AI agents proliferate inside an organization faster than its ability to attribute their value, assign accountable owners, control their cost and retire the ones that do not pay back.
It is not a technology failure. The agents may work flawlessly. It is capital misallocation at scale.
The term is defined in the Agent Chaos Foundational Statement, written by Douglas Siqueira (2026). Foundational Statement →
How to recognize it
- Several teams run agents that do the same job.
- Usage costs rise every month, but nobody can say what they buy.
- Expensive automation is applied to cheap work, with no payback analysis.
- Agents run in production with no person accountable for their cost.
- No agent has ever been retired.
Why it matters now
Creating agents has become cheap; operating them has not. In Dataiku's September 2026 survey, 84% of CIOs said employees build agents faster than IT can govern them, and 72% cannot consistently confirm that all their agents deliver business outcomes.
Only 21% have near-real-time visibility of AI costs by business unit or use case. Meanwhile, Gartner expects more than 40% of agentic AI projects to be canceled by the end of 2027 because of escalating costs, unclear value or inadequate risk controls.
The five capabilities that prevent it
- 1
Portfolio-level value attribution
Visibility into the collective cost, function and estimated contribution of the agents in production.
- 2
Explicit executive ownership
Every agent has a named economic owner, accountable for its continued existence.
- 3
Differentiated thresholds
Efficiency agents and strategic agents are judged by different economic criteria.
- 4
Retirement capability
The organization can systematically switch off agents that no longer justify their cost.
- 5
Periodic portfolio review
The agent portfolio is reviewed with the same rigor as capital budgeting decisions.
Not to be confused with
The words "agent" and "chaos" appear together in a few other contexts. They are related, but they are not the same thing.
- Chaos testing for AI agents
- An engineering practice: deliberately injecting failures into agent workflows to test their resilience before production. It asks whether an agent survives failure. Agent Chaos asks whether the agent should exist at all. Sources ↗
- "Agents of Chaos" (Northeastern University, 2026)
- A study in which researchers manipulated autonomous agents into leaking information and wiping an email server. It shows what happens when agents hold broad access with nobody accountable — one of the symptoms of Agent Chaos. Sources ↗
- "Agent of chaos" (expression)
- An everyday expression for a person who disrupts order, popularized by the Joker in the film The Dark Knight (2008). Unrelated to AI.
Frequently asked questions
What is Agent Chaos?
Agent Chaos is the condition in which AI agents proliferate faster than an organization's ability to attribute their value, assign accountable owners, control their cost and retire the ones that do not pay back.
What causes Agent Chaos?
Agents are cheap to create and expensive to operate. They are built inside business units, SaaS tools and personal accounts, without a shared model for ownership, cost and review.
Is Agent Chaos a security problem?
Security failures are one symptom. The underlying problem is economic: agents with no owner, no known cost, no contracted outcome and no retirement criterion.
How should companies measure the ROI of AI agents?
At the portfolio level. Measuring ROI agent by agent creates false precision because agents work in interdependent workflows. Track total cost, contribution by workflow and different thresholds for efficiency agents and strategic agents.
Who should own AI agents in a company?
Each agent needs a named economic owner in the business — a person accountable for its continued existence — and the portfolio should be reviewed at executive level, like capital budgeting.
Where should an organization start?
With an inventory. Answer five questions: how many agents are in production, who owns each one, what each costs per month, what outcome each is meant to move, and when it was last reviewed.
Five questions for your next board meeting
If your team cannot answer them immediately, the organization is not scaling AI. It is accumulating exposure.
Ask your team these 5 questionsExecutive Briefings
One brief per month: a pattern, the data behind it and three questions for your team. You also get The Board Questions as a printable one-pager.