Generative AI is becoming a routine part of financial modeling and client presentations, but financial-services firms are struggling to build the controls needed to verify AI-generated work. A new Macabacus report says 87% of surveyed users employ AI daily or weekly for models and presentations, while only 23% report having comprehensive guardrails in place.

Generative AI is changing how investment bankers, financial analysts and professional-services teams build models and client materials. The harder problem may be ensuring that faster work remains accurate when it reaches the people making financial decisions.

That tension is at the center of a new report from Macabacus, the Microsoft 365 productivity platform for finance and professional-services teams. Its report, GenAI for Financial Services: Velocity and Verification, examines how firms are adopting AI for financial models and client presentations while developing controls intended to catch errors.

Macabacus says the research draws on a survey of its 75,000 users as well as conversations with hundreds of clients and prospective customers. According to the company’s findings, 87% of respondents use AI daily or weekly for financial models and client presentations.

Yet only 23% say their organizations have comprehensive guardrails covering areas such as approved AI tools, accuracy checks, brand compliance and review workflows.

The gap creates a familiar enterprise-AI problem: deployment can move considerably faster than governance.

Financial models and client presentations are particularly sensitive applications. A small formula error, incorrect assumption, unsupported statement or outdated data point can change the interpretation of a transaction or investment opportunity. Unlike consumer-facing generative AI use, where an incorrect answer may be inconvenient, an error in deal content can carry financial, reputational and regulatory consequences.

Macabacus reports that 62% of respondents believe an AI-generated error reached a client or internal decision-maker during the previous 12 months. More concerning, 46% said an error had probably made it into a deliverable without being detected.

Those figures are company-reported survey results rather than independently audited industry statistics, but they point to a broader challenge facing financial technology: how to introduce AI without weakening existing verification processes.

The issue is becoming more relevant as AI moves from experimentation into everyday financial workflows. McKinsey’s 2025 Global Banking Annual Review found that generative AI could create substantial value for the banking industry, with use cases spanning software development, customer operations, marketing and risk management. The firm has also emphasized that realizing this potential requires organizations to redesign workflows rather than simply add AI tools to existing processes. (mckinsey.com)

For finance teams, verification is therefore becoming part of the technology stack.

The Macabacus findings also reveal a difference in how AI is perceived across seniority levels. Forty-three percent of analysts and associates surveyed said AI had made them more confident in their work. Among vice presidents, directors and managing directors, only 29% said the same.

That divide is notable because junior employees are often closer to the production layer where AI tools are being introduced, while senior professionals frequently carry responsibility for reviewing and signing off on the final product.

The result is a new handoff problem. AI can accelerate the creation of a financial model or presentation, but the person responsible for the final output still needs to determine whether the underlying numbers, assumptions and narrative are defensible.

This is where verification infrastructure could become a competitive category in enterprise financial technology.

Rather than treating AI governance as a separate compliance function, firms are increasingly integrating controls directly into workflows. These can include approved-model policies, source validation, permissions, audit trails, human review, data lineage and automated checks.

Macabacus is positioning its own platform around this concept. The company’s broader software suite operates within Microsoft 365, connecting finance workflows with applications such as Excel and PowerPoint. Its report argues that firms should not necessarily reduce AI usage because of accuracy concerns; instead, they need controls that allow AI-generated work to be reviewed before it becomes client-facing.

That philosophy mirrors an emerging principle in enterprise AI: the question is not simply whether a model can generate useful content, but whether an organization can establish enough context and controls to trust its output.

The technology challenge is complicated by the fragmented nature of financial data. Models may draw on spreadsheets, presentations, enterprise resource planning systems, customer relationship management platforms, market data and proprietary transaction information. A generative AI system can produce fluent output without necessarily understanding whether each source is current, authorized or appropriate.

For investment banking and professional-services firms, that makes data governance closely connected to AI adoption. The more financial work moves through AI systems, the more important it becomes to know where information originated, how it was transformed and who reviewed the final result.

Microsoft, Google, Salesforce and other enterprise technology providers are investing heavily in AI assistants and agents, increasing the availability of AI across office and business applications. The next competitive layer may be the controls surrounding those systems rather than the generation capabilities themselves.

The Macabacus report ultimately highlights a paradox: AI can increase the speed of financial work while simultaneously increasing the importance of human review.

For GlobalFinTechEdge, that puts AI verification at the intersection of financial technology, banking technology innovation and enterprise software. Financial firms are unlikely to abandon AI because of accuracy concerns. Instead, the market is moving toward systems designed to make AI-generated financial work faster to produce—and harder to get wrong.

Market Landscape

AI adoption across financial services is moving from experimentation toward embedded workflow automation, but governance remains uneven.

Gartner has reported that finance organizations continue to expand AI usage while struggling to translate experimentation into measurable business value. Its research increasingly emphasizes structured AI roadmaps, data foundations and governance alongside productivity improvements.

The market is consequently expanding beyond AI generation tools into verification, data lineage, model governance, workflow controls and auditability.

For investment banks, accounting firms, private-equity teams and corporate finance departments, the opportunity is to integrate these controls into familiar environments such as Microsoft 365 rather than force employees into entirely separate AI systems.

This creates a competitive field spanning Microsoft, enterprise software providers, specialized fintech startups and financial-workflow platforms such as Macabacus.

Top Insights

• Macabacus reports that 87% of surveyed users employ AI weekly or more for financial models and client presentations.
• Only 23% report comprehensive guardrails covering AI tools, accuracy, brand compliance and review workflows, according to the report.
• Sixty-two percent believe an AI-generated error reached a client or decision-maker during the past year.
• Junior finance professionals report greater confidence from AI than senior reviewers, highlighting a potential review-handoff problem.
• Verification, governance and data lineage are becoming strategic components of enterprise AI rather than separate compliance functions.

Get in touch with our fintech expert