REPORTED DEVELOPMENT
Forrester predicts that in 2027 one in three large enterprises in Asia-Pacific will identify at least ten unapproved or informally deployed AI agents for every formally sanctioned agent. The warning concerns employee-built tools, vendor-embedded agents and systems adopted independently by business units.
The forecast highlights the difficulty of maintaining an inventory of agents, their owners, permissions and operating dependencies as organisations expand AI adoption. Forrester also expects some consumer-facing brands in the region to differentiate through visible human oversight of sensitive decisions.
These are forward-looking predictions for large APAC organisations, not an observed ratio across all companies or evidence of existing global adoption.
WHY IT MATTERS TO THE MARKET
As AI agents move from answering questions to accessing business systems and carrying out tasks, firms need a reliable inventory, named accountability, access controls and a defined way to retire each agent. Unmanaged agents can become a source of operational risk and technology debt even when the underlying AI model works as intended.