Artificial intelligence (AI) agents are rapidly moving from experimental technology to essential business tool, with adoption rates and investment levels that suggest a fundamental shift in how companies approach AI – refuting a controversial MIT study on the same topic, according to a new report released Thursday by G2.
The software marketplace’s 2025 AI Agents Insights Report, based on a survey of more than 1,000 B2B decision-makers, found 57% of companies already have AI agents in production — a remarkably fast transition from testing to full-scale deployment. And unlike the first wave of generative AI tools designed primarily to advise and assist, these agents autonomously execute tasks, coordinate across systems, and make decisions with minimal human intervention.
Financial investment backing the trend is substantial. Forty percent of companies surveyed have allocated more than $1 million for AI agent initiatives, while one in four large enterprises are prepared to spend $5 million or more. More than half of organizations say they’re highly likely to expand their agent budgets over the next 12 months.
“Agents are the great leap in the application of AI technology that we’ve all been waiting for,” Tim Sanders, chief innovation officer at G2, said in a statement. The shift represents a transition “from measuring individual productivity to predicting organizational velocity,” he added.
Despite initial skepticism as famously outlined in an MIT AI study earlier this year that showed glacial enterprise adoption, AI agents are delivering returns quickly. The median time to first meaningful outcome is six months or less, matching or exceeding previous AI investments.
Survey respondents reported a median 23% improvement in speed-to-market, while G2’s review data shows velocity gains reaching up to 50% in marketing and sales applications.
The most common enterprise use cases were in customer service, business intelligence, and software development, where 83% of buyers reported satisfaction with agent performance.
The report emphasized that trust, particularly through human oversight, remains critical to successful agent deployment. Programs that kept a human in the loop were twice as likely to achieve cost savings of 75% or more compared to fully autonomous strategies.
Still, confidence is growing. About a third of participants employ a so-called “let it rip” approach, deploying agents with plans to make reactive corrections as needed. Nearly half of buyers would grant full autonomy to agents in low-risk workflows, they said.
Surprisingly, agentic AI seems to be boosting rather than threatening employee satisfaction. Nearly 90% of participants reported higher employee satisfaction in departments where agents were deployed, and 45% of leaders predict a net increase in jobs by 2028 as talent is redeployed to higher-value work.
The rise of AI agents is forcing a reckoning for the traditional software-as-a-service industry. More than one in three companies would consider switching from their current SaaS vendor to acquire agent functionality, the report found.
However, most leaders don’t see agents as a complete replacement for traditional software. Two-thirds believe the future will involve agents augmenting SaaS in a complementary relationship rather than replacing it entirely.
The survey also revealed the emergence of an “agent-to-agent” era, with 50% of companies reporting that their agents already hand off work across different vendors and platforms, suggesting a future where AI systems coordinate with each other across organizational boundaries.
Sanders argues that software vendors will need to adapt their messaging, targets, and pricing strategies to thrive in what he calls an “agent-first future.”
