Asia practitioners report: AI agent deployment brings new management burden
SCMP reports on the real experiences of enterprises deploying AI agents across Shanghai, Singapore, and Hong Kong: agents locked in prolonged disputes over responsibility division, employees became the busiest people as projects multiplied, and companies began shifting from pursuing agent count to building coordination frameworks.
ImportanciaMaterialEvidenciaE3 inspeccionableAnálisisEstándar
Xiao Mengjie, an AI product marketing manager at a large technology company in Shanghai, recently assigned what she described as a simple task to two AI agents—one belonging to her and one belonging to her supervisor. She expected efficiency gains. Instead, she got bureaucratic gridlock: the two agents did not divide the work but fell into a prolonged discussion over responsibilities, debating who should verify details and who should take ownership of which tasks. 'It became very formal and cumbersome,' Xiao said. 'Agents can create communication overhead too.'
This is not an isolated case. SCMP recently interviewed practitioners at enterprises across Asia—from hotels and insurers in Singapore to the service sector in Hong Kong and tech giants in mainland China—and found a significant gap between how AI agents actually perform and the promise of 'digital workers freeing human labor.' Xiao described another layer of difficulty: 'I already have lots of ideas, and AI makes more of them feel achievable, so I start more projects. Then every project is waiting for my judgment, and I end up being the busiest person.' When agents can work continuously at high speed while humans do not gain extra hours in the day to supervise them, automation brings not leisure but new management pressure.
Facing this problem, some enterprises have begun adjusting their strategies. Raunak Mehta, co-founder and CEO of Singapore-based insurance technology company Igloo, said 'the market is over the phase of agent maxing.' Igloo operates across Southeast Asia with about 230 employees and uses software workers for customer service, development, claims processing, and financial reconciliation, but Mehta believes total agent count is no longer a meaningful measure of operational success. The focus has shifted toward 'agent harnesses'—control frameworks that coordinate agents, regulate their access to internal tools, and allow human oversight. Prabu Sangar, director of technology operations and integration at FCM Travel Asia, calls the state lacking such harnesses 'agent sprawl': autonomous software deployed by individual departments with conflicting permissions, overlapping data sources, and unpredictable interactions.