Businesses are rapidly expanding their use of AI agents. However, safety and governance frameworks are failing to keep pace, according to a new report from Deloitte. The State of AI in the Enterprise study, based on a survey of more than 3,200 business leaders across 24 countries, highlights a widening gap between adoption and oversight as agentic AI moves from pilots into production environments.
Deloitte’s data shows that 23% of companies are already using AI agents at least moderately, a figure projected to surge to 74% within the next two years. At the same time, the share of organizations not using agents at all is expected to drop from 25% to just 5%. Despite this rapid uptake, only 21% of respondents said their companies currently have robust safety and oversight mechanisms in place to manage agent-related risks. Deloitte cautioned that as agentic AI scales, weak governance could limit value creation. Furthermore, there is an increasing exposure to operational, security, and compliance failures.
Unlike traditional chatbots, including OpenAI, Microsoft, Google, Amazon, and Salesforce, have promoted agents as productivity accelerators. Yet, Deloitte notes that autonomy without adequate controls can amplify potential introduces new risks. From unexpected behavior to prompt-injection attacks, especially when agents operate without clearly defined boundaries. Major technology providers have promoted agents as productivity accelerators, but Deloitte notes that autonomy without adequate controls can amplify harm.
Other research reinforces the concern. Recent studies found that while 84% of IT professionals report AI agent usage in their organizations, fewer than half have policies governing those systems. Surveys also show many employees use AI tools without formal safety training. And they lack visibility into how their employers deploy AI at scale.
Key takeaways for business leaders:
- AI agent adoption is accelerating faster than safety and governance frameworks
- Only a small minority of organizations have robust oversight in place
- Autonomous agents increase risk exposure compared with traditional chatbots
- Clear boundaries, monitoring, and audit trails are essential as agents scale
Deloitte concludes that while early-stage adoption rarely comes with perfect guardrails, oversight must become a priority. The firm recommends defining clear limits on agent autonomy, implementing real-time monitoring, and maintaining audit trails to ensure accountability as agentic AI becomes a core part of enterprise operations.
Source:
