Executive Briefs · · 5 min read

Shadow AI: The Agents You Pay For and Cannot See

AI agents are now built inside business units, SaaS tools and personal accounts — faster than IT can govern them. Shadow AI is not a policy failure. It is an inventory failure, and it has a cost nobody is tracking.

Douglas Siqueira · Agent Chaos

How big the gap is

In Dataiku's September 2026 survey, 84% of CIOs said employees build agents and applications faster than IT can govern them, and 81% lack complete oversight of agents built outside approved systems.

EY's survey of senior leaders points the same way: shadow AI was among the most common AI risks encountered in the past year (46%), and it surfaced in 39% of formal AI assurance reviews.

Why prohibition does not work

Creating an agent now costs almost nothing, and the capability ships inside tools the company already pays for. Banning it pushes it further out of sight.

Even formal processes bend under pressure: 47% of the leaders EY surveyed admit they have skipped their AI governance process for urgent deployments. A rule that is easier to skip than to follow will be skipped.

The economic cost of what you cannot see

Hidden agents duplicate each other across teams. Their costs scatter across departmental budgets, corporate cards and SaaS add-ons. Their data access goes unreviewed. And nobody can retire an agent that nobody registered.

Only 21% of the CIOs in the Dataiku survey have full, near-real-time visibility into AI costs by business unit, team or use case. For everyone else, shadow AI is a line item that does not exist.

Two portfolios: the one on the register and the one outside it

On the registerAgents on approved platforms.Cost appears in: the technology budget.+you pay for bothOutside approved systemsAgents inside SaaS tools, business unitsand personal accounts.Cost scatters across: department budgets,corporate cards, SaaS add-ons.

81% of CIOs lack complete oversight of agents built outside approved systems. Only 21% see AI costs by business unit or use case in near real time.

Sources: Dataiku / The Harris Poll — Global AI Confessions Report: CIO Edition (Sep 2026, 685 CIOs)

What works: make registering cheaper than hiding

The goal is visibility, not punishment.

  • A lightweight agent register: name, owner, purpose, data access, monthly cost, review date.
  • A short amnesty period in which any existing agent can be registered without consequences.
  • Budget tied to registration: unregistered agents do not get paid for.
  • A quarterly review of the register at leadership level.

Executive 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.

One email per month. Unsubscribe anytime. No sales sequence. Privacy policy

Sources

Figures come from public research. Several studies are sponsored by vendors that sell related products; each source is linked so you can judge it yourself.

Keep reading

  • Agent Sprawl: When More Agents Means Less Control

    AI agent sprawl is not about how many agents you run. It is about running them without a portfolio. The signs, the cost and four decisions that contain it.

  • Who Owns an AI Agent?

    When an AI agent fails, accountability splits across four teams. Why "IT owns it" is the wrong answer, and how to name a single economic owner for every agent.