Agentic AI hype surges but enterprises remain stuck in pilot mode, Forrester finds

Enterprise leaders are racing to adopt agentic AI, but the vast majority remain trapped between flashy demos and real-world deployment, according to a new Forrester report. The analyst firm found that 75% of enterprise leaders claim they are adopting agentic AI, yet only a small minority have moved beyond pilots into meaningful production.

Forrester's assessment, published Friday, warns that vendors are slapping the "agentic" label on products while most organizations struggle to scale. The firm notes that long-horizon agents capable of operating for days or weeks are now technically feasible, but the realities of enterprise environments quickly derail those demos.

The gap between hype and deployment

Forrester identifies several key barriers keeping agentic AI in proof-of-concept purgatory:

• Governance remains immature, with more than half of enterprises experiencing "agentic sprawl" despite adopting formal policies.
• Platform strategies are fuzzy, and many organizations build agents in isolation without a clear way to track or coordinate them.
• Return on investment is difficult to demonstrate, making broader deployment hard to justify.
• Autonomous systems need automated guardrails that can enforce rules in real time, but most companies rely on static policies.

The findings echo a separate report from The Hacker News, which found that only 10% of security operations centers say they are getting excellent value from AI. Eighteen months ago, the AI SOC was a marketing line; today it is a budget item, but value remains elusive for most.

What comes next

Forrester argues that companies are not struggling because they have too many AI agents, but because managing them gets messy fast. As projects grow, overlapping systems, duplicated work, and unpredictable agent behavior become common. The firm concludes that writing governance rules is one thing; enforcing them is another.

Meanwhile, consumer AI companies face their own monetization challenges. Light Reading reports that ads will not fix consumer AI's money problem, and companies must convert heavy free users into paid subscribers at a higher rate. Anthropic, which filed for an IPO this week, urged the industry to slow down AI sprints, saying it "would be good for the world." Canada also announced plans to develop its own AI to reduce reliance on US technology.

For enterprises, the path forward requires tying agent autonomy to measurable changes in how work gets done. Until then, Forrester warns, agentic AI will remain stuck in proof-of-concept purgatory.