News
Microsoft: AI Builders Are Growing More Confident in Enterprise Agents
Artificial intelligence agents are moving beyond experimentation and into production, according to Microsoft's inaugural 2026 Agent Confidence Index.
Published on June 29, the report surveyed 300 AI, data and cloud professionals to measure confidence in using AI agents across 101 workplace tasks.
Microsoft said the findings provide a snapshot of where organizations believe agents are ready for enterprise use and where caution remains.
"We surveyed 300 technical experts across AI, data and cloud domains, asking them to rank their confidence across 101 of the top tasks," Amanda Silver, corporate vice president of Microsoft 365 Core and Work IQ, said.
"What we got back is the 2026 Agent Confidence Index -- an honest map of where agents are delivering real value."
According to Microsoft, respondents reported an average confidence score of 64 out of 100 across all tasks. In contrast, 30 individual tasks scored above 70, suggesting that AI agents are already viewed as reliable for a range of business activities.
The highest confidence was associated with structured, repetitive work, including generating reports, creating boilerplate code and summarizing information.
"The highest scores cluster on work that is both predictable and draining: the late nights, the interruptions, the low-value repetition," the report noted. "These are exactly the kinds of tasks people want to hand off."
Confidence declined for tasks requiring greater judgment, creativity and complex decision-making, indicating that respondents still see clear limits to autonomous AI in the workplace.
The report also found that organizations continue to prioritize human oversight as they expand AI deployments. Nearly six in 10 respondents identified keeping humans in the loop as the most important way to build trust in AI agents, while more than half pointed to observability, monitoring and tracing as essential capabilities.
Microsoft said the findings reflect a broader shift in enterprise AI adoption.
"The future isn't agents replacing people," Silver wrote. "It's people and agents working together, each doing what they do best."
For Microsoft, the results suggest enterprise AI is entering a more practical phase.
Rather than asking whether AI agents can be used at work, organizations are increasingly identifying where they can deliver the greatest value while ensuring appropriate governance and human oversight remain in place.