Enterprises winning with AI agents are limiting how much the agents can do alone

Insight summary

• •Enterprises deploying AI agents are finding that limiting agent autonomy leads to better outcomes than maximizing it.
• •Over 40% of current agentic AI projects are predicted to fail by 2028 due to rising costs, unclear value, and risk management issues.
• •Many organizations lack mature governance, with only 30% achieving a higher level of responsibility and control in AI deployment.
• •High autonomy complicates traceability and accountability, causing legal and compliance challenges, especially in regulated industries.
• •Integration complexity with existing workflows is a common cause of project delays and cancellations.
• •Security and risk issues are now cited as the biggest hurdles to scaling agentic AI, surpassing regulatory and technical concerns.
• •Successful enterprises use narrow-scope agents, human checkpoints before high-stakes actions, decision traceability by design, and active data governance aligned with emerging regulations.