The latest Medius Financial Census reveals a quiet epidemic permeating corporate finance departments: the normalization of minor, seemingly harmless financial indiscretions that collectively erode profitability. Dubbed ‘shallowfakes’ by researchers, these low‑value manipulations—such as inflating mileage claims, adding a few extra dollars to a lunch receipt, or approving a questionable reimbursement—may appear trivial in isolation. Yet when repeated across teams and over months, they siphon away resources that could fund innovation, staff development, or strategic investments. The study, which surveyed 2,386 finance leaders in the United States, United Kingdom, Sweden, and France, shows that 87 % of respondents have at least once turned a blind eye to an expense they suspected was not legitimate. This tolerance is not merely an isolated lapse; it reflects a broader cultural shift where small rule‑bending is rationalized as harmless or even expected. Understanding why these behaviors persist is the first step toward designing controls that catch them before they accumulate into material losses.
The survey goes further, exposing the psychological triggers that make finance professionals complicit in these micro‑frauds. Two‑thirds (67 %) admitted they would be more likely to submit a slightly inflated claim if they observed peers doing the same, indicating that social proof plays a powerful role in shaping ethical boundaries. Additionally, 64 % said they would feel justified in bending the rules when they perceive themselves as underpaid or undervalued, turning personal grievances into a rationale for financial misconduct. Over half (57 %) confessed they would round up an expense or mileage figure if they believed the act would go unnoticed, revealing a classic opportunity‑driven motive. These findings suggest that traditional compliance training, which often focuses on high‑profile scandals, misses the subtle pressures that drive everyday dishonesty. Leaders must therefore address both the perceived fairness of compensation and the visibility of controls, ensuring that employees feel respected while also knowing that even small deviations are monitored and discouraged.
The cumulative impact of these seemingly minor infractions can be staggering. When each instance adds only a few dollars, the aggregate loss across a large organization can reach hundreds of thousands of dollars annually—what Chris Wilmot, Medius CFO, aptly described as ‘death by a thousand cuts.’ Unlike a single large‑scale fraud that triggers immediate alarms, shallowfakes operate beneath the radar of conventional audit sampling and exception‑based reporting. They exploit the assumption that low‑value transactions are immaterial, allowing them to slip through automated matching rules that focus on duplicates, mismatched amounts, or vendor mismatches. Over time, the constant drain reduces cash flow, inflates expense budgets, and distorts key performance indicators such as cost‑to‑collect or operating margin. Moreover, the erosion of financial integrity can undermine stakeholder confidence, making it harder to secure favorable credit terms or attract investors who demand transparent stewardship of corporate assets.
While shallowfakes represent an internal, culture‑driven threat, finance teams are simultaneously bracing for an external surge in AI‑generated deception. An overwhelming 93 % of respondents expressed concern that deepfake technologies—synthetic audio, video, or documents crafted by generative models—will be used to perpetrate fraud within the next year. This apprehension is warranted: as AI tools become more accessible, bad actors can forge convincing invoices, mimic executive voices for payment authorization, or produce falsified contracts that evade superficial inspection. The tension lies in the fact that the same AI capabilities that enable sophisticated scams are also being harnessed by finance departments to improve efficiency, detect anomalies, and automate decision‑making. Consequently, organizations must adopt a dual‑focus strategy: fortify defenses against AI‑enabled external attacks while ensuring internal AI applications are governed, transparent, and resistant to manipulation.
The financial toll of traditional invoice fraud remains substantial, underscoring that the threat landscape is multifaceted. U.S.‑based organizations participating in the census reported average annual losses of $168,000 attributable to fraudulent invoices, with respondents experiencing roughly one successful attempt each month. These figures highlight that even with growing investments in AP automation, criminal tactics continue to find gaps—whether through social engineering that bypasses validation rules, compromised vendor portals, or sophisticated spoofing of payment instructions. The monthly frequency suggests that fraudsters are adopting a volume‑based approach, relying on the law of large numbers to eventually slip a false invoice through. Finance leaders should therefore treat invoice fraud as a persistent operational risk, continuously updating rule sets, integrating real‑time vendor verification, and employing anomaly‑detection models that adapt to evolving patterns.
Automation has undoubtedly transformed the accounts payable landscape, yet the census reveals a stark gap between technology deployment and operational outcome. Although 85 % of finance teams claim to have implemented some level of AP automation, nearly half of all invoices still require manual intervention before they can be cleared for payment. This dependence on human touchpoints translates into delayed settlements, with organizations estimating that 28 % of invoices are paid later than agreed terms in a typical month. The persistence of manual steps points to several root causes: incomplete integration between ERP systems and supplier portals, exceptions triggered by non‑standard purchase orders, and a lack of straight‑through processing capabilities for complex invoices that include multiple line items, taxes, or discounts. Moreover, the manual burden contributes heavily to workplace stress, with 96 % of respondents linking late‑payment management to burnout among AP staff—a clear signal that technology alone does not guarantee relief unless it is configured for end‑to‑end efficiency.
The repercussions of delayed payments extend far beyond the finance office, rippling through the supply chain and affecting overall business health. Nearly half (45 %) of survey participants said their suppliers have responded to chronic lateness by imposing stricter upfront payment conditions, such as demanding deposits or shortening credit windows. An additional 43 % reported that suppliers have terminated relationships altogether after repeated late payments, eliminating valuable sources of goods or services. Moreover, 42 % noted that vendors have reduced the speed or quality of their deliveries as a punitive measure, while an equal proportion indicated that disputes have escalated to formal legal proceedings. These outcomes illustrate that payment timeliness is not merely an internal efficiency metric; it is a critical component of supplier trust and partnership stability. When payment reliability deteriorates, companies risk higher procurement costs, supply disruptions, and potential litigation expenses that far outweigh the short‑term cash‑flow benefits of holding onto funds longer.
Industry analysts echo the sentiment that automation must deliver measurable results, not just box‑checking compliance. Kevin Permenter, Research Director for Financial Applications at IDC, observed that the original promise of AP automation was to make late payments the rare exception rather than the routine outcome. Yet the data show that despite sustained technology investment, many organizations remain stuck in a cycle of manual overrides and delayed settlements. Permenter argues that the next phase of digital transformation must focus on outcome‑driven metrics: tracking the percentage of invoices achieving straight‑through processing, measuring reduction in exception rates, and linking automation improvements to working‑capital gains. Only when technology investments are explicitly tied to financial performance indicators can finance leaders justify further spending and ensure that automation evolves from a tactical tool into a strategic lever for liquidity optimization.
Artificial intelligence is rapidly becoming a fixture in modern finance workflows, yet trust in its recommendations remains incomplete. The census indicates that 38 % of finance executives have already deployed agentic AI—systems capable of autonomous decision‑making—in at least some of their processes, with another 50 % planning to adopt such capabilities within the next year. Despite this enthusiasm, a striking 90 % of professionals insisted that a human must always retain the final say for any transaction that crosses a predefined financial or compliance threshold, regardless of the AI’s historical accuracy. This dichotomy reveals a pragmatic acknowledgment that while AI can process vast volumes of data and identify patterns faster than humans, accountability for erroneous outputs cannot be fully delegated to algorithms. Finance teams are therefore navigating a transitional state where AI augments judgment but does not replace the need for oversight, especially in areas involving regulatory risk, fraud detection, or strategic resource allocation.
Unclear lines of responsibility compound the trust issue when AI‑driven decisions lead to adverse outcomes. When asked who should bear accountability for an AI error that creates a financial loss or compliance breach, respondents were almost evenly split among IT leaders, finance leaders, and the end‑users who acted on the AI’s recommendation. This fragmentation suggests that many organizations have not yet established explicit governance frameworks that designate an AI owner, define model‑monitoring protocols, or establish escalation paths for anomalous outputs. Furthermore, 45 % of finance leaders admitted that their teams frequently act on AI‑generated suggestions without any human review, highlighting a potentially dangerous reliance on automation that outpaces supervisory controls. To mitigate these risks, companies should institute AI governance boards that include representatives from finance, IT, legal, and internal audit, mandate model explainability for high‑impact decisions, and enforce periodic audits that verify both accuracy and compliance with corporate policies.
AI’s influence is also reshaping talent management within finance, creating new expectations for skills and performance. More than half (55 %) of finance leaders said that proficiency with AI tools has become a meaningful differentiator when recruiting new staff, while 86 % indicated that an employee’s usage of AI now factors into performance evaluations and promotion decisions. This shift underscores the growing premium placed on digital fluency, as organizations seek individuals who can leverage machine‑learning insights, manage autonomous agents, and interpret algorithmic outputs. However, the same technology that promises to eliminate mundane tasks is inadvertently contributing to worker fatigue. Three‑quarters (75 %) of respondents reported that increased AI usage has heightened feelings of burnout, likely due to the constant need to monitor, validate, and intervene with AI systems, as well as the pressure to upskill rapidly. Leaders must therefore balance the drive for AI adoption with thoughtful change‑management practices, providing adequate training, realistic workload expectations, and wellness support to prevent talent erosion.
Drawing from these insights, finance leaders can adopt a pragmatic roadmap to curb shallowfakes, strengthen AI governance, and realize the full benefits of automation. First, implement continuous‑monitoring controls that flag even low‑value anomalies—such as repeated rounding of mileage claims or incremental expense inflations—using machine‑learning models trained on historical patterns. Second, foster a culture of fairness and transparency by regularly reviewing compensation structures, recognizing employee contributions, and communicating the ethical impact of seemingly minor infractions. Third, establish a formal AI governance framework that assigns clear accountability, mandates explainability for high‑risk decisions, and requires human‑in‑the‑loop verification for transactions exceeding defined thresholds. Fourth, optimize AP automation by identifying the root causes of manual interventions—such as non‑standard PO formats or missing vendor data—and investing in straight‑through processing capabilities, supplier portals, and exception‑reduction initiatives. Finally, strengthen supplier relationships through proactive communication, realistic payment schedules, and joint process improvements that enhance reliability for both parties. By addressing both the human and technological dimensions of risk, finance teams can protect working capital, preserve trust, and position themselves as strategic partners in sustainable growth.