SMB leaders are moving AI agents into pilots faster than they can prove return on investment, according to Upwork Research Institute’s new Business Leader Landscape survey.
The survey covered 750 U.S.-based business leaders across business and professional services, healthcare, manufacturing, retail and consumer goods and software and technology. Upwork’s AI-agent analysis focused on a subset of 195 leaders at companies with 10 to 99 employees.
Within that SMB group, confidence is already showing up in deployment decisions. 62% of SMB leaders said they are very or extremely confident handing high-stakes tasks to AI agents, and 32% described agents as mission-critical to company strategy.
High conviction drives active pilots
The pilot data gives the clearest evidence of near-term adoption. Decision support had 41% of SMB leaders actively piloting AI agents, compared with 3% not considering the use case. Information retrieval stood at 36% versus 2%, workflow automation at 34% versus 2% and multistep planning across systems at 34% versus 3%.
Upwork described the pattern directly: SMB leaders are “operating on conviction while they wait for major productivity gains.” The underlying productivity data is more limited. 74% of SMB leaders reported productivity gains from AI, but Upwork found most improvements had not passed 25%.
The ongoing gap between adoption and integration
A separate Goldman Sachs 10,000 Small Businesses Voices survey shows a similar gap between AI adoption and full operational integration. The survey of 1,256 program participants found 76% were using AI and 93% of AI users reported a positive business impact, but only 14% said AI was fully embedded in core operations.
Goldman’s survey also pointed to the work still required after adoption begins. 73% of respondents said more training and implementation resources would help them adopt AI successfully.
“AI is already helping small businesses compete, save time, and better serve customers — but many of us are still figuring out how to use it effectively,” said Khari Parker, co-founder of Connie’s Chicken and Waffles in Baltimore, in Goldman’s release.
Contextualizing the numbers with national data
Broader small-business research gives the Upwork findings useful context, but also shows why the figures should be read carefully. The U.S. Chamber of Commerce’s 2025 Empowering Small Business report found 58% of small businesses said they used AI, up from 40% in 2024 and 23% in 2023. The report separately listed 44% using Gen AI chatbots.
Upwork’s survey goes further by focusing specifically on agents, not general AI tools or chatbots.
National data gives the Upwork findings another boundary. The Census Bureau’s Business Trends and Outlook Survey, collected from Dec. 14, 2025, to May 3, 2026, found overall AI use among U.S. businesses hovered between 17% and 20%, with 20% to 23% expecting to use AI in the next six months. Census also found adoption rose with firm size: 32% of firms with 100 to 249 employees and 37% of firms with at least 250 employees reported using AI in the period ending May 3.
The difference between those national AI-use figures and Upwork’s SMB agent findings comes partly from survey design. Census asks whether businesses used AI in any business function. Upwork surveyed business leaders and focused on AI agents, then separated active pilots from scaling and non-consideration.
Security hurdles and misaligned definitions
Upwork’s report also points to unresolved adoption barriers. It says data privacy and security topped the SMB barrier list at 49%, compared with 44% across the broader respondent pool. In its key takeaways, Upwork separately lists ROI uncertainty at 24% as the second-biggest barrier behind data security and compliance at 27%.
The definition problem sits alongside those barriers. Only 69% of SMB leaders said they were very or extremely confident their leadership team shared a common definition of what an AI agent is, compared with 75% across the broader Upwork sample.
Measurable outputs lead early scaling
For now, the clearest scaling areas are the functions with more measurable outputs. Upwork found data analytics led agent scaling at 27%, followed by content generation at 26% and inventory management at 24%. Those functions give SMBs a more direct way to compare agent output against cost, cycle time, volume or operational throughput.