Companies are racing to adopt AI agents… but few have rules in place to govern them.
According to a new report from Code and Theory and WSJ Intelligence, just 18% of large US companies have governance, accountability and reporting measures in place for AI agents that are comparable to those for humans.
The problem is exacerbated by AI sprawl. Companies use an average of 183 AI tools and platforms, per the report, with 13% running at least 500 (!). Meanwhile, one in four respondents says there’s no single clear owner for AI orchestration across their organization.
The findings are based on a survey of 801 C-suite executives at US companies with at least $500 million in annual revenue.
Why This Matters:
AI agents can increasingly do a lot. But they can also make some serious mistakes.
An agent operating without proper accountability could delete important files or data, disclose sensitive information or make decisions with real financial or operational consequences. And as companies give more agents the ability to act independently, the stakes get higher.
The sheer number of AI tools is another problem. Two-thirds of executives surveyed say their organizations frequently discover redundant AI investments or overlapping pilots. In other words, different departments may be buying similar tools, testing the same capabilities or building AI workflows without knowing what other teams are doing.
That creates unnecessary spending, more systems to manage and potentially more opportunities for things to go wrong, especially when governance is limited.
Experts React:
The report argues that traditional software oversight isn’t necessarily sufficient for AI agents capable of making decisions and taking action.
“An organization can add hundreds of tools without extending the management structures that determine who answers for an agent’s decisions,” the report states.
It recommends giving every agent a defined role, a specific human to be accountable for its decisions, clear boundaries, and an auditable record of its actions.
Our Take:
These findings feel inevitable. The adtech industry has spent the past few years moving fast, shipping products and encouraging companies to adopt the technology as quickly as possible. That’s helped accelerate innovation, but it also means governance hasn’t necessarily kept up.
And that’s a problem as the industry transitions to agentic. There is a big difference between an AI tool that might produce a bad answer and an agent that might take a bad action. Companies don’t need to slow the pace of innovation, but they do need to strengthen reporting, accountability and policies to safeguard themselves.
