62% of Finance Firms Sent AI Errors to Clients in Past Year

More than six in ten financial services professionals believe an AI-generated error has reached a client or internal decision-maker within the past year, according to new research that exposes a significant gap between AI adoption and verification infrastructure.

The findings come from Macabacus, a New York-based Microsoft 365 productivity platform serving finance and professional services teams. The company's 2026 GenAI in Financial Services: Velocity and Verification report, released September 10, surveyed 75,000 users across investment banking, private equity, corporate finance, and advisory firms.

Of the 62 percent who believe an error escaped review, 46 percent described it as a "probably, but no one noticed" scenario—suggesting mistakes may be circulating in client presentations and financial models undetected. Only 38 percent expressed confidence that no such error had occurred.

Why it matters

Financial services firms face a structural risk as AI tools become embedded in client-facing work. The gap between daily AI usage and verification systems creates liability exposure and threatens the client trust that underpins advisory relationships. Firms that build verification infrastructure now can accelerate AI adoption without compromising accuracy or credibility.

Widespread adoption, minimal oversight

The research found that 87 percent of respondents use AI daily or weekly to generate financial models and client presentations. Yet only 23 percent work at firms with comprehensive guardrails—defined as approved tools, accuracy checks, brand compliance, and structured review workflows. Another 36 percent of daily or weekly AI users operate at firms with no guardrails at all.

"Deal teams should not slow down their use of AI," said Paul Ross, chief marketing officer at Macabacus. "They need guardrails that let them move faster while maintaining accuracy and their clients' trust."

A confidence divide between junior and senior staff

The report surfaces a meaningful split in how different levels assess AI reliability. Among analysts and associates, 43 percent said AI has made them more confident in their models and presentations. Among vice presidents, directors, and managing directors, that figure drops to 29 percent.

Senior reviewers are also six percentage points more likely than junior staff to say AI has made them less confident in the work. This divergence matters because junior employees often generate AI-assisted deliverables while senior professionals provide oversight—creating conditions where errors can pass through undetected.

One portfolio management firm quoted in the report compared AI to "a glorified intern," noting that output still requires proofing and challenge regardless of the tool used.

What the industry wants

When asked what would increase confidence in AI-generated content, respondents aligned around three priorities: 27 percent wanted a full audit trail showing AI changes, 27 percent called for mandatory human review at handoff, and 25 percent requested automated verification built into existing tools. Only 5 percent said firm-approved tools alone would suffice.

The time cost is substantial. Eighty-five percent of respondents spend 30 minutes or more checking AI-generated models and content before client delivery—overhead that better guardrails could redirect toward higher-value work.

The details were first reported by Macabacus in its 2026 GenAI in Financial Services report.