Deloitte’s newly released Tech Trends 2025 report reveals that AI agents did not achieve widespread adoption this year. Despite intense hype and rapidly growing investment in agentic AI, only 11% of organizations currently use AI agents in production. Deloitte’s findings highlight why adoption stalled and what enterprises can do to close the gap.
Although enterprises were eager to deploy autonomous AI systems capable of handling high-volume tasks, the report found a persistent disconnect between expectations and operational reality. Deloitte’s survey of 500 U.S. tech leaders showed that 30% of organizations are exploring agents and 38% are piloting them. Meanwhile, 42% lack a formal roadmap, and 35% have no agent strategy at all.
Several structural obstacles emerged:
- Legacy architecture remains the biggest barrier. Many organizations still rely on aging enterprise systems not designed for autonomous AI, causing bottlenecks that prevent agents from reliably executing tasks.
- Poor data architecture slows automation. Nearly half of organizations cite weak data searchability and reusability as major hurdles, limiting agents’ ability to retrieve and act on information.
- Governance models are not ready for autonomous decision-making. Traditional IT oversight does not address scenarios where AI systems operate independently across business functions.
Furthermore, Deloitte notes that the organizations seeing success took a process-first approach, redesigning workflows around AI rather than “dropping agents onto outdated processes.” Also, they invested in employee enablement—an area where most companies fall short. Currently, 93% of AI budgets go toward technology, while only 7% fund training and cultural readiness.
Finally, Deloitte warns that without rethinking systems, governance, and human roles, enterprises will continue to struggle. But with thoughtful implementation, agentic AI can still play a major role in future operations. As CTO Bill Briggs put it, the failure of 2025 is “a story as old as time”—another tech wave slowed not by potential, but by readiness.
Source:
https://www.zdnet.com/article/7-ways-to-be-a-data-superstar-in-the-ai-era-and-stay-ahead-of-agents/
