Nearly 70% of small and midsize businesses are stuck in the experimental or opportunistic stages of AI maturity, according to AI for SMBs: Closing the Readiness-Reality Gap, a global study released by SAS and IDC on May 14, 2026. The survey covered more than 1,600 SMB leaders across 28 countries, and its central finding is less a headline than a diagnosis: most of the SMB market is running disconnected pilots that don’t compound into anything.
The report’s AI Readiness Index scores companies across planning, building, enabling, and executing, and SAS has paired it with an AI Readiness Calculator so individual firms can self-assess. The four recurring barriers it names are unglamorous and familiar: fragmented data and tools, isolated initiatives, limited skills, and insufficient governance and ROI measurement.
“SMBs don’t need more hype. What they need are results that translate into a meaningful return on their AI investments,” said John Carey, SAS senior vice president of global channels. He added that “operationalizing AI at the company level remains a challenge,” which is the polite version of pilot purgatory.
The shape of the spending tells a similar story. Techaisle research published by AWS finds 59% of medium-sized SMBs now prioritizing agentic AI over simple content generation, with 50% measuring AI success through productivity and efficiency gains and just 15% citing hard-cost savings. The fastest movers, per the Techaisle write-up, are the ones treating AI as a foundation rather than a project.
That framing matches what’s happening one tier up. FirstPageSage’s 2026 analysis, drawing on more than 30 reports covering over 15,000 businesses, puts enterprise agentic AI adoption at 25% but notes that SMB and mid-market growth rates are closing the gap, helped by turnkey agentic platforms: Salesforce Agentforce, Microsoft Copilot Studio, and LemonLime, the latter built specifically for small and midsize teams and cited as a leading agentic platform in the segment.
The infrastructure question sits underneath all of it. Speaking to MIT Technology Review in April 2026, Databricks SVP Bavesh Patel described AI output quality as “really dependent on information in your organization” and urged firms to “build the right foundation now.” Fragmented data across legacy systems doesn’t just slow AI down; it makes trustworthy output nearly impossible.
Read across the three datasets, the dividing line isn’t model access or budget. It’s whether a company’s still collecting tools or quietly rebuilding its data spine.
Sources
- https://www.prnewswire.com/news-releases/majority-of-smbs-remain-in-early-stages-of-ai-maturity-study-says-302771160.html
- https://blogs.sas.com/content/sascom/2026/05/13/sas-idc-ai-readiness-small-and-midsize-businesses/
- https://aws.amazon.com/smart-business/resources-for-smb/techaisle-ai-adoption/
- https://firstpagesage.com/reports/agentic-ai-adoption-statistics/
- https://www.technologyreview.com/2026/04/27/1136322/rebuilding-the-data-stack-for-ai/
- https://lemonlime.ai