Is AI really helping your SMB? Study finds a quarter of execs can't explain what their AI actually does

What happened

A new study covered by TechRadar finds that a significant share of small and medium-sized business leaders are adopting AI tools for critical functions without fully understanding how they work. According to the research, a quarter of executives at SMBs say they cannot explain the outputs their AI systems generate, even as a majority of these businesses now lean on AI to handle sensitive financial tasks such as processing expenses and approving payments.

The findings point to a widening gap between the pace of AI adoption in smaller businesses and the level of oversight, training or governance accompanying that rollout. Tools are being embedded into day-to-day financial operations faster than the understanding needed to scrutinise what they are actually doing.

Why it matters

For SMB leaders, the appeal of AI is speed and cost savings — automating expense checks, payment approvals and other back-office tasks that once required manual review. But when those decisions are delegated to systems whose outputs leadership cannot interpret, organisations are effectively trading operational efficiency for reduced visibility into their own financial controls. That is a meaningful risk in functions where errors, bias or hallucinated outputs can have direct monetary consequences.

This is a governance and trust problem as much as a technology one. Business leaders who cannot explain what an AI tool is doing are poorly positioned to catch mistakes, justify decisions to auditors or regulators, or build employee and customer confidence in AI-assisted processes. As AI moves deeper into finance and operations, the ability to explain, audit and override automated outputs becomes as important as the efficiency gains driving adoption in the first place.

The Renascence take

The instinct is to treat this as an AI literacy problem that better training will fix. It is really a service-design and accountability problem: tools were deployed into high-stakes workflows before anyone defined who owns the decision when the AI is wrong.

Unexplainable AI in expense and payment approval isn't a knowledge gap — it's an accountability gap dressed up as a technology story. If a leader can't explain an output, that process has effectively been handed over without a named owner, which is precisely the setup where small errors compound into real financial and reputational exposure. Customer- and employee-obsessed operators don't ban the tool; they build explainability, an audit trail and a clear human checkpoint into the workflow before scaling it, treating "can we explain this?" as a launch gate rather than an afterthought.