Enterprises have plunged headlong into deploying AI agents without the necessary governance structures in place, according to a new report from VentureBeat Research. The study highlights a growing disparity between the deployment of AI technologies and the establishment of essential control mechanisms. In a landscape where AI is becoming increasingly integral to business operations, this oversight could have significant ramifications for both efficiency and security.

### What Enterprise AI Agents Actually Do

AI agents are designed to automate various business processes, ranging from customer service chatbots to complex decision-making systems. However, the term “agent” is often misapplied. VentureBeat’s research indicates that 71% of enterprises admit their so-called agents are merely chatbots, incapable of executing multi-step processes autonomously. Only a small fraction—10%—of enterprises report that true agents form the majority of their deployments. This mischaracterization suggests that many organizations are not fully capitalizing on AI’s potential, mistaking basic automation for advanced intelligence.

The report identifies five key control layers required for trustworthy AI agent deployment: identity, evaluation, cost telemetry, context layer, and orchestration. These controls govern what an agent can do, assess the quality of their output, track costs, provide necessary business context, and coordinate complex tasks. Yet, many enterprises are only now beginning to integrate these controls, with a significant portion planning on vendor changes or additions within the next year to address these gaps.

### Competitive Context and Industry Challenges

The rush to deploy AI agents without proper governance is not an isolated issue. The competitive pressure to innovate and automate is leading companies to prioritize deployment over due diligence. This phenomenon is underscored by the fact that two-thirds of enterprises are moving towards allowing AI agents to make production changes based solely on automated evaluations. Despite this, only 5% of these companies fully trust the evaluations that underpin such significant decisions.

The risks of this approach are evident. Half of the enterprises surveyed experienced customer-facing failures due to poorly evaluated AI agents. This highlights a critical misalignment between AI ambitions and operational readiness, which could potentially erode consumer trust and lead to financial repercussions.

### Real Implications for Founders, Engineers, and the Industry

For founders and engineers, the findings of this report serve as a stark reminder of the importance of robust governance in AI deployment. The data suggest that many companies are treating AI as a checkbox exercise rather than a strategic initiative requiring careful planning and execution. Engineers, in particular, should advocate for comprehensive testing of AI evaluations against real-world outcomes, rather than relying solely on internal benchmarks.

Security concerns also loom large, with 69% of companies permitting credential sharing among AI agents. This practice significantly increases the risk of security breaches, as evidenced by a higher incident rate in such organizations. Implementing scoped identities for each agent is a critical step that engineers and IT leaders must prioritize to safeguard sensitive information and maintain operational integrity.

The underutilization of costly hardware resources, such as GPUs, further underscores the inefficiencies plaguing current AI deployments. With over 80% of enterprises reporting suboptimal GPU usage, there’s a clear opportunity for IT departments to optimize resource allocation and reduce unnecessary expenditure.

### What Happens Next

As enterprises scramble to align their AI deployments with necessary governance frameworks, there’s a clear opportunity for solution providers that specialize in AI control layers to capture market share. For founders and investors, this gap represents a chance to innovate and offer comprehensive governance solutions that address these critical needs.

For engineers, the focus should shift towards developing scalable, reliable evaluation systems and advocating for stringent security measures. The industry must recognize that the effective deployment of AI agents is not simply about speed but about ensuring that these systems are trustworthy and aligned with business objectives.

In a rapidly evolving AI landscape, the message is clear: rushing to deploy without proper governance could lead to costly missteps. For those in the trenches, the challenge is to balance the allure of rapid deployment with the necessity of robust, reliable systems.