AI has become a routine part of finance operations, but widespread use is not necessarily translating into transformation. A new Rillion 2026 AI in Finance Report finds that while most U.S. finance teams are already using or evaluating AI, concerns around human oversight, manual processes and workforce skills continue to limit its impact.
AI adoption is becoming commonplace inside U.S. finance departments, but a new survey suggests that high usage may be masking a deeper problem: many organizations have not yet built the trust, skills or processes required to turn AI experimentation into meaningful operational change.
Rillion’s 2026 AI in Finance Report, based on a survey of 250 U.S. CFOs and finance leaders, identifies four areas where expectations are running ahead of reality: trust, skills, automation and transformation.
The headline number is encouraging. Sixty-eight percent of finance teams surveyed already use AI in their daily work, while another 28% are either piloting the technology or considering adoption.
But only 39% of CFOs say they are comfortable allowing AI to act independently without human review.
That gap matters. In finance, where decisions can affect cash flow, financial reporting, compliance and supplier relationships, organizations have less room for unchecked automation than many other business functions. AI can accelerate processes, but finance leaders still need confidence that outputs are accurate, explainable and appropriately controlled.
AI adoption is not the same as transformation
The findings challenge the assumption that simply deploying AI represents meaningful digital transformation.
Finance departments have spent years digitizing accounts payable, expense management, reconciliation and other back-office workflows. Generative AI and machine learning are now adding another layer of automation, but the technology does not automatically eliminate the operational constraints embedded in older processes.
Invoice processing illustrates the problem.
Nearly nine in 10 CFOs surveyed say their existing invoice capture and data-extraction solutions have shortcomings. Forty-five percent still require human review after invoices have been processed.
That suggests a familiar pattern in enterprise automation: technology handles the standard case, while employees remain responsible for exceptions.
The result can be an automation ceiling. A company may automate a large percentage of routine transactions but still depend heavily on people whenever an invoice format changes, data is incomplete or a transaction falls outside predefined rules.
For finance leaders, the next phase of AI adoption is therefore less about adding another AI feature and more about redesigning workflows around exceptions, controls and decision-making.
Trust remains a major barrier
The trust gap is particularly important as finance departments move from AI-assisted work toward AI agents and more autonomous systems.
Tools from Microsoft, Google, Salesforce and other enterprise technology providers are increasingly capable of generating recommendations, summarizing information and executing multi-step tasks. But financial processes require a clear distinction between an AI system suggesting an action and an AI system being authorized to take that action.
The Rillion findings suggest that most finance leaders have not yet crossed that threshold.
Only 39% are comfortable allowing AI to operate independently without human review. In practical terms, that means human-in-the-loop workflows are likely to remain the dominant model for many finance organizations even as AI usage increases.
This is not necessarily a sign that adoption has stalled. It could indicate that finance leaders are applying a higher standard of governance to autonomous technology.
The challenge is ensuring that human oversight adds genuine control rather than becoming another manual bottleneck.
The finance AI skills gap may be underestimated
The report also identifies a potentially important mismatch around workforce capabilities.
Only 21% of respondents currently identify a lack of internal AI expertise as a major adoption barrier. Yet 60% believe that understanding AI tools will become one of the most important skills for finance professionals.
That difference points to a forward-looking skills problem.
Finance teams may not need every employee to become an AI engineer. They do, however, need people who can evaluate AI outputs, understand where automation is appropriate, recognize errors and determine when human judgment should override a recommendation.
Those capabilities are becoming part of financial technology literacy.
The change could also alter the role of finance professionals. Routine data collection and processing are increasingly automatable, potentially giving teams more time for forecasting, scenario analysis, risk management and strategic decision support.
But that transition only works if organizations invest in training alongside technology.
Experience appears to change expectations
One of the report’s more interesting findings concerns the relationship between AI experience and expectations.
Among finance organizations already using AI extensively, 51% expect the technology to transform most finance processes. That compares with only 16% among organizations earlier in their AI journey.
The difference suggests that organizations with practical exposure to AI may have a clearer understanding of how far the technology could eventually reshape finance.
It also reveals a paradox: organizations with limited AI experience may underestimate the scale of change because they have not yet seen what advanced automation can accomplish.
For enterprise leaders, that makes experimentation strategically useful. Controlled deployments can provide finance teams with evidence about where AI delivers measurable value and where existing processes prevent automation from working effectively.
From AI tools to AI-native finance operations
The findings point toward a broader shift in enterprise financial technology.
The first wave of AI adoption focused heavily on individual productivity: generating text, summarizing documents, extracting data or assisting employees with routine tasks.
The next phase is likely to focus more on connected workflows.
In accounts payable, for example, an AI-enabled system could potentially extract invoice data, validate it against purchase orders, identify anomalies, route exceptions and prepare the transaction for approval. The objective is not simply to make one step faster but to reduce the number of points where manual intervention is required.
That distinction will become increasingly important as finance departments evaluate AI platforms.
The strongest systems may not necessarily be the ones with the most visible AI features. They may be the ones capable of operating reliably across messy real-world processes while maintaining auditability, human controls and clear accountability.
Rillion’s research ultimately presents AI adoption in finance as a maturity question rather than a technology race. Most finance teams are already moving. The harder challenge is building the trust, skills and process architecture needed to move from AI-assisted finance toward genuinely AI-enabled operations.
Market Landscape
The report reflects a broader transition from AI-assisted finance to AI-driven finance operations.
Enterprise platforms from Microsoft, Google, Salesforce, Oracle and SAP are increasingly incorporating AI into financial planning, accounts payable, analytics and workflow automation. At the same time, specialized fintech vendors are targeting specific processes such as invoice capture, reconciliation, expense management and financial close.
The competitive question is shifting from “Does the platform use AI?” to “How much of the workflow can AI reliably operate?”
That distinction is particularly relevant to accounts payable. If 45% of surveyed finance leaders still require human review after invoice processing, organizations may have automated individual steps without eliminating the underlying process friction.
The next generation of finance automation is likely to emphasize exception handling, workflow orchestration, auditability and controlled autonomy alongside generative AI capabilities.
Top Insights
- AI adoption is widespread in U.S. finance, with 68% of teams using AI daily, but limited autonomy shows trust remains a critical enterprise barrier.
- Invoice automation remains incomplete, as 45% of CFOs still require human review after processing, exposing persistent weaknesses in finance workflow infrastructure.
- AI skills are becoming strategic finance capabilities, with 60% expecting AI-tool knowledge to become one of the profession’s most important skills.
- Experienced AI adopters expect greater disruption, with 51% anticipating transformation across most finance processes compared with 16% of early-stage organizations.
- Enterprise finance is moving toward controlled autonomy, requiring AI systems that combine automation with auditability, exception handling and human oversight.
Get in touch with our fintech expert
