Genpact’s new research, released on November 12, points to a gap between deploying AI agents and turning them into financial results. The survey of more than 2,000 enterprise executives found that 79% have AI agents in production. Yet only 15% have fully connected AI spending to profit-and-loss outcomes.
Production use is not the same as broad use
Only 20% of surveyed businesses use AI agents broadly and consistently across functions. That is a sharp difference from the 79% production figure. The study covered 17 industries and 14 business functions, with executives at companies earning $1 billion to more than $50 billion. Genpact calls the missed opportunity the autonomy dividend: gains that show up in business performance, not just productivity.
Investment returns lag deployment
Just 49% of AI initiatives funded in the prior 24 months delivered expected value, on average. And only 9% reported completing substantial operating model changes. Sanjeev Vohra, Genpact’s Chief Technology and Innovation Officer, said enterprises have become effective at deploying agents, but running them is the hard part. That sounds right. Installation is one task; daily management, governance, and accountability are another.
AI also takes a larger decision role. It leads or fully handles 40% of decisions, more than double the share a year earlier. People remain involved, though: at each decision stage, humans make or audit the call in at least 87% of cases. Among respondents with agents in production, 92% said human headcount was unchanged or higher.
Leaders link operations, governance, and financial ownership
Genpact identifies 17% of respondents as AI Leaders, based on its AI Maturity Index. Those companies have CEO-level mandates and are further along in blending people and agents into one operating model. They lead across five building blocks the research assessed, though most enterprises have not connected all five.
The report’s practical point is less about buying more tools and more about redesigning work. Define where agents act, where people decide and validate, and which outcomes the combined workforce owns. Without that, AI adoption can look strong in production dashboards while staying weak in the P&L. Perhaps the real test is not how many agents are live, but who is accountable when their decisions hit the bottom line.
