Why Excel Could Decide the Future of AI in Accounting

New research suggests that the biggest obstacle to artificial intelligence in accounting may not be distrust of AI itself. It may be the comfort, flexibility, and institutional weight of Microsoft Excel. That makes the spreadsheet both a barrier to disruption and a potentially powerful gateway for Microsoft.

A new study of 271 accounting professionals examined why some organizations adopt artificial intelligence while others remain attached to familiar analytical tools.¹ Its central finding is deceptively simple: using Excel for data analytics was negatively associated with AI adoption, while AI training and broader adoption of data-analytics tools were positively associated with it.

This does not mean that Excel is obsolete or that accountants are irrational. Excel remains flexible, widely understood, and deeply embedded in budgeting, reconciliation, forecasting, reporting, and audit preparation. The more important lesson is that technology adoption depends on workflow, habit, perceived risk, and training as much as technical capability.

For investors, that changes how the enterprise AI opportunity should be evaluated. The likely winners may not be the companies offering the most dramatic replacement for established work. They may be the companies capable of inserting AI into software that workers already open every morning.

Why Excel Can Slow AI Adoption In Accounting

The researchers approached AI adoption through a dual-factor model. Instead of treating adoption as a simple question of whether users see enough benefits, they considered both drivers and inhibitors. An employee can recognize that AI is useful while still resisting it because changing tools introduces uncertainty, training costs, integration work, and the possibility of errors.

That tension is especially important in accounting. An experimental marketing draft can be corrected. A mistake in a financial model, reconciliation, regulatory filing, or audit workpaper can carry material consequences. Accountants therefore have good reasons to favour systems whose formulas, controls, and limitations they understand.

Excel also benefits from accumulated organizational knowledge. Companies have spent decades building templates, macros, models, reporting procedures, and approval processes around spreadsheets. Employees know how to inspect formulas and trace calculations. Replacing that environment means giving up more than an application. It means disturbing an operating system for financial work.

The study interprets this through status quo bias. Familiarity produces confidence and a sense of control, while the potential losses associated with switching can appear more immediate than the prospective gains. This helps explain why merely demonstrating that an AI platform is more advanced may not be enough to drive adoption.

What The Accounting AI Study Found

The final sample included 271 accounting professionals who had been involved in AI, machine learning, or data-analytics initiatives. Approximately 92% worked in accounting or related service-consultancy organizations, and 62% worked at organizations with fewer than 500 employees.

Only 33% of respondents reported having received AI training. Meanwhile, just 36% of the respondents using Excel for data analytics had moved to more advanced AI solutions. Excel knowledge alone was not a statistically significant predictor once the full model was considered, but actual Excel-based analytical use had a significant negative association with AI adoption.

The distinction matters. Knowing Excel does not inherently prevent someone from adopting AI. The barrier emerges when established spreadsheet workflows already satisfy immediate needs, making the cost and uncertainty of changing systems difficult to justify.

AI Training Matters More Than An Additional Software License

Organizations often treat AI adoption as a procurement decision. They purchase licenses, activate features, and expect usage to follow. The study suggests that this approach confuses software availability with organizational capability.

Training had a positive association with AI adoption, as did prior adoption of broader data-analytics tools. Together, those results suggest that organizations benefit from building a progression of capabilities. Accountants who have already moved beyond basic spreadsheet analysis are more prepared to evaluate, supervise, and incorporate AI.

A practical adoption program should therefore focus on specific workflows:

• Reconciliation and anomaly detection
• Variance analysis and management reporting
• Forecasting and scenario development
• Document review and audit preparation

These bounded applications are easier to validate than a general-purpose mandate to “use AI.” They also create measurable outcomes, such as fewer manual steps, faster closing cycles, or more transactions reviewed. That is increasingly important as finance departments are asked to demonstrate a return on AI spending rather than reporting license counts or employee interest.

The finding also complements a broader Securities.io analysis arguing that proprietary data and institutional knowledge may become the most valuable AI moat. In accounting, the model is only one component. The harder advantage to reproduce is the combination of trusted financial data, established controls, company-specific procedures, and employees who understand what a valid output should look like.

Why Familiar Software May Become The Best AI Distribution Channel

The paper describes Excel as an inhibitor, but that does not mean the spreadsheet must disappear. The stronger commercial insight is that reducing switching costs may be more effective than asking users to abandon familiar tools.

Microsoft is already pursuing that route. Its Finance Agent connects Excel with existing financial systems and is designed to support activities such as reconciliation, variance analysis, and real-time financial insight. This is strategically important because it brings AI to the interface accountants already trust.

Embedding AI inside incumbent software can address several adoption barriers at once. It limits workflow disruption, retains familiar spreadsheets, allows outputs to be reviewed in an established format, and reduces the psychological cost of trying a new capability. The user is not being asked to replace Excel with AI. The user is being invited to extend Excel through AI.

This pattern is not limited to accounting. Enterprise AI increasingly competes on distribution, integration, governance, and access to context rather than model performance alone. Recent Securities.io coverage of UiPath and the shift toward agentic enterprise automation illustrates the same commercial reality: automation becomes more valuable when it can operate across the applications and processes a company already uses.

Adoption Does Not Automatically Produce Financial Returns

The study should not be read as evidence that deploying AI will necessarily improve profitability. It measured adoption, not the financial return generated by that adoption. Its survey was relatively small, concentrated heavily in accounting-related consultancies, and collected over a short period. The regression results identify associations rather than proving that training or analytics tools caused organizations to adopt AI.

There are also risks that the model did not fully capture, including data privacy, cybersecurity, trust in third-party providers, output reliability, and regulatory accountability. These issues are particularly important when AI is applied to financial reporting.

Current industry evidence nevertheless suggests adoption is expanding. KPMG’s 2026 Global AI in Finance Report, based on more than 1,000 senior finance leaders, indicates that active use has grown substantially, while exceptional performance remains less common than adoption itself. That gap reinforces the paper’s underlying message: acquiring technology is easier than changing the organization around it.

Investing In The Interface For Accounting AI

For investors seeking exposure to AI adoption in accounting, Microsoft offers a particularly relevant position. The company owns Excel, the incumbent tool identified in the study, while also developing the AI layer intended to modernize financial work.

That creates an unusual strategic advantage. Excel’s entrenchment can slow migration toward competing AI platforms, but it also gives Microsoft direct access to a vast installed base of finance professionals. Rather than overcoming status quo bias by removing the status quo, Microsoft can monetize it by placing Copilot and finance-focused agents inside the existing Microsoft 365 environment.

The investment case extends beyond one accounting feature. Microsoft can combine productivity software, cloud infrastructure, identity controls, enterprise data connections, and AI agents within a single commercial ecosystem. Securities.io’s broader analysis of Microsoft as an AI investment explores how Copilot fits within that larger platform.

There are meaningful risks. Customers may resist additional subscription costs, finance teams may limit AI access because of accuracy or privacy concerns, and competing platforms may offer stronger specialized accounting capabilities. Microsoft must also prove that embedded AI produces measurable value rather than merely adding another paid feature to an already complex software stack.

Still, the study points toward a durable commercial principle. Enterprise AI adoption is not determined solely by which system is most intelligent. It is also determined by which system asks users to change the least. In accounting, Excel may delay the transition to AI, but it may ultimately become the doorway through which that transition occurs.

References:

¹ Juma’h, A. H., & Li, Y. (2026). Unlocking AI’s potential in accounting: A study on adoption barriers and drivers. Advances in Accounting, 71, 100905. https://doi.org/10.1016/j.adiac.2026.100905