Workplace AI projects remain stuck at concept stage.

A Dynatrace report found about half of agentic AI initiatives are still in proof‑of‑concept or pilot stages, with governance, security and skills shortages cited as major barriers; three quarters of firms expect to increase agentic AI budgets next year.

Many organisations report that their workplace agentic AI projects are not moving beyond early stages. A Dynatrace report says about half of initiatives are still at proof‑of‑concept or pilot stages. The report attributes delays mainly to governance, safety and skills gaps rather than doubts about AI's value. It also notes that most firms plan to increase agentic AI spending next year. Key findings: - Around half of agentic AI initiatives remain in proof‑of‑concept or pilot phases. - Deployment is concentrated in IT operations and DevOps (72%), software engineering (56%) and customer support (51%). - Expected top ROI areas are IT operations and system monitoring (44%), cybersecurity (27%) and data processing and reporting (25%). - Main barriers include security, privacy and compliance concerns (52%), difficulty managing agents at scale (51%) and a shortage of skilled staff or training (44%). - Human oversight remains common: about 69% of agentic AI decisions are verified by humans, 87% of organisations build agents requiring human supervision, and 23% prefer solely human‑supervised agents. - Three quarters (74%) of respondents expect to increase agentic AI budgets next year. Summary: The report suggests that technical capability is not the primary obstacle; instead, organisational and governance issues are slowing full implementation. Dynatrace recommends redefining how return on investment is measured, setting clear guardrails for human‑machine collaboration and scaling deployments deliberately rather than funding rapid, large rollouts. These steps are presented as ways to help organisations move from experimentation to wider adoption in a controlled manner.