Generative AI spending among small and mid-sized businesses is projected to grow more than twice as fast as revenue, but data, security and cost controls are slowing the move from experimentation to production, according to an AWS-published Techaisle study of global SMB leaders.
The study says generative AI spend is set to rise 15.4% to 18.7%, compared with 6.3% revenue growth.
Data and integration complexities block deployment
According to the findings, 96% of SMBs are acting on AI, while 54% are held back by data. The top barriers listed in the survey are cybersecurity at 53%, data quality at 47% and integration complexity at 47%, placing the adoption problem inside the operating stack rather than at the level of executive interest.
AWS’ accompanying analysis adds another constraint: 37% of SMBs cite AI skills gaps as a barrier, with small businesses more affected by skills shortages and medium-sized businesses more affected by integration complexity.
NIST’s AI Risk Management Framework describes AI systems as riskier to manage than traditional software in part because training data can change over time, affecting functionality and trustworthiness in ways that may be difficult to detect.
That data problem is already shaping cloud priorities. The Techaisle summary says strengthening data governance for AI is the No. 1 cloud initiative for SMBs, ahead of AI integration itself.
It also found that 84% of SMBs see integrated stacks as critical, but only 29% have operationalized them.
Token shock and the push for private AI architecture
Pricing is the second pressure point, according to AWS. 42% of SMBs cite inflexible pricing models and unexpected true-up costs as their top frustration with AI vendors, a problem it labels “token shock.”
The analysis shows 78% of SMBs prefer private or hybrid AI architectures because they offer more predictable fixed infrastructure costs and reduce dependency on volatile API pricing.
That preference is visible in the architecture split. Only 17% of SMBs accept fully public cloud AI architecture, while 41% favor virtual private cloud and 38% prefer hybrid deployments, according to the Techaisle summary. The same page says 43% have ended up multicloud “by accident,” making platform consolidation part of the AI-readiness problem for many buyers.
CISO veto power and shifting budget structures
The study also divides the SMB market around a $30 million threshold. Below that level, Techaisle describes AI decisions as CEO-led and cost-driven; above it, the study says decisions become more CISO-influenced and productivity-focused.
AI budgets are also moving outside central IT, with 38% of AI budget sitting with functional department heads, while the CEO signs off and the CISO holds veto power.
Mid-sized businesses pivot to agentic workflows
Agentic AI is moving into that budget process. AWS’ Techaisle analysis says 59% of medium-sized businesses are already prioritizing agentic AI over simpler content-generation tools, defining agentic systems as autonomous tools that execute complex workflows independently.
The same analysis says SMBs measure AI success primarily through productivity and efficiency at 50%, compared with hard-cost savings at 15%.
FinOps emerges as a strategic adoption requirement
IDC’s 2026 SMB outlook points in the same direction, saying SMBs are moving from experimentation to strategic adoption while focusing on use cases that are easy to deploy and produce measurable ROI.
IDC also expects FinOps to become more important as SMBs manage AI and cloud costs, and says security, compliance and risk management will increasingly influence which vendors SMBs trust and adopt.