Paid Program: How Auditors Maintain Trust in the Age of AI

How Auditors Maintain Trust in Financial Reporting in the Age of AI

Independent perspectives help bridge the gap between technological complexity and market trust.

Millions of investors rely on financial reports to make investment decisions. Increasingly, those reports may contain information generated by company AI systems. A recent survey by the Center for Audit Quality (CAQ) reveals 40% of institutional investors believe one of the biggest risks associated with AI implementation is lack of human oversight to ensure accurate AI outputs, while 25% cite overreliance on automation. As AI transforms corporate finance, independent auditors serve as the essential bridge to transparency and trust. They ensure the information investors depend on—whether generated by humans or AI—remains reliable. The following fictitious case study is inspired by real-world audit challenges. It’s meant to illustrate how the combination of traditional audit expertise and understanding of emerging technologies can uncover new risks—ones that could threaten investor interests. We talked to real public company auditors for insights into how they have evolved their methods to address how AI shapes financial reporting. Their long-standing skills in professional skepticism, testing and financial analysis have equipped them to understand and assess risks with these new AI systems.NevoGuide Solutions, a fast-growing public technology company, has positioned itself as an AI-first organization. The company began using GenAI to forecast financial performance—promising faster analysis, greater precision and deeper insight. But as AI becomes embedded in financial operations, a critical question emerges: Who is monitoring these machine-generated outputs for reliability in financial reporting? “Finance wasn’t first out of the gate with AI adoption,” says Jennifer Kosar, assurance AI leader for PwC US. “But once finance leaders understood what the technology could do, what the risks were and how to manage them, adoption really accelerated.” That acceleration is creating new challenges for the auditors tasked with verifying the reliability of financial information in scenarios like NevoGuide’s. As AI enters financial reporting, auditors apply their traditional expertise in new areas—evaluating not only the accuracy of information, but also the reliability of the AI-driven processes and controls that generate it.

Part 1: When AI Sounds Too Good to Be True

At NevoGuide, the use of GenAI to create detailed forecasts of expected customer payment behavior—which informs the company’s estimate for allowance for doubtful accounts, a key figure for investors—seemed to be a breakthrough. GenAI analyzed accounts receivable aging, customer payment histories, credit profiles and economic indicators to forecast collections, producing detailed narratives that explained why certain accounts posed risks. The analysis it produced was more granular than a human could perform and was delivered with explanations that appeared plausible.

“With human estimates, you can usually trace the assumptions and rationale behind the conclusion,” says Brian Miller, assurance managing principal of digital transformation and innovation at BDO. “But with generative AI, especially when used as a narrative layer over complex models, outputs can sound convincing when the underlying logic isn’t transparent.” Routine audit procedures to understand NevoGuide’s financial reporting processes revealed the company was relying on outputs from its GenAI without verifying its reliability. Further, NevoGuide’s GenAI forecasting system developed customer collection estimates based on algorithms the company couldn’t fully explain or replicate. NevoGuide had no way to verify whether the AI was properly weighing different inputs or potentially overemphasizing certain data patterns from its training. And because NevoGuide used a third-party model, the company—and audit team—had no visibility into the model’s underlying training data or methods. “Users often accept AI-generated results if they sound plausible,” warns Richard Jackson, EY Americas assurance chief technology officer; EY global and Americas assurance AI leader. “The greater risk is that even developers may lack insight into how or why the model produced a specific output—especially when working with third-party systems.” AI is often referred to as a “black box”—as there isn’t visibility into why or how the technology arrives at its outputs and conclusions. In these circumstances, human oversight is critical. It calls for not only verifying the reliability of AI-generated outputs, but also instituting a comprehensive approach to AI governance that supports the effective use of AI throughout the company.

Why It Matters

Cases like NevoGuide’s show how auditors’ focus on processes and controls that support financial reporting enables them to identify risks arising from GenAI use. Their expertise in evaluating management’s systems helps prevent overreliance on AI-generated outputs that could misstate results.

Part 2: Dealing With the Black Box

Faced with the black box of NevoGuide’s AI system, auditors couldn’t solely rely on traditional methods to respond. They had to adapt their audit procedures to the new technology.“When we can’t audit an ERP’s algorithms directly, we usually build an independent model using interpretable methods,” Miller explains. “We replicate the forecasting process with available inputs to produce our own estimates and compare those results to the client’s AI outputs.” This independent approach allows auditors to test AI conclusions without necessarily needing to understand the complex algorithms behind them. The audit team at NevoGuide developed their own AI model and compared the outputs to NevoGuide’s model. They discovered meaningful differences between the auditor’s independent expectation and the company’s AI-generated forecasts: NevoGuide’s GenAI had gradually begun favoring outdated customer collection patterns while undervaluing current economic factors. The system had drifted from its original performance parameters while continuing to generate plausible explanations to accompany forecasts.

Why It Matters

This discovery showcases the value of auditors’ professional skepticism. Auditors looked beyond surface results to understand underlying processes. Their risk-based approach—focusing scrutiny on areas where AI most influences financial reporting—enabled them to identify systematic problems the company’s own reviews missed.

“With generative AI, especially when used as a narrative layer over complex models, outputs can sound convincing when the underlying logic isn’t transparent.”

— Brian Miller, Assurance Managing Principal, Digital Transformation and Innovation, BDO

Part 3: Getting Oversight Right

Digging deeper into NevoGuide’s processes, the auditors discovered the company also lacked a comprehensive inventory of AI systems that could affect financial reporting and did not have processes to monitor whether these financially relevant AI systems continued to operate effectively over time.

In response, NevoGuide strengthened its AI governance by designating a chief AI officer to coordinate efforts across departments. It centralized responsibility for managing AI inventories, establishing ongoing monitoring protocols and implementing human oversight policies for AI outputs. This centralized approach addresses what Kosar identifies as a key challenge: “As AI adoption widens, governance becomes more complex—and without clear accountability, risks grow.”NevoGuide’s management implemented robust human-in-the-loop protocols, training employees to critically evaluate the reliability of GenAI outputs rather than accepting results that seemed reasonable. The company also established regular model performance reviews and retraining cycles to ensure AI systems adapt to changing business conditions.

“As AI adoption widens, governance becomes more complex—and without clear accountability, risks grow.”

— Jennifer Kosar, Assurance AI Leader, PwC US

Why It Matters

Auditors don’t just check the numbers—they review management’s processes and controls around AI use to strengthen trust and transparency and ensure financial integrity.

Part 4: Building Trust in Automated Systems

NevoGuide’s challenges reflect common gaps auditors encounter as companies implement AI: lack of comprehensive AI inventories, insufficient data governance frameworks and inadequate human oversight of AI-generated outputs. As companies get up to speed on GenAI, audit firms are rapidly building expertise to respond to the technology’s transformative power. “We’re integrating foundational AI and data science concepts into our firm-wide training programs so accountants can understand key concepts like machine learning, data preparation, modeling validation and bias detection,” Miller shares. “We also run rotation programs where practitioners spend six months to two years with our analytics teams to learn modeling and data quality before returning to client-facing roles with a deeper understanding of emerging technologies.” The goal isn’t to turn auditors into data scientists, he assures. “It is to ensure they can critically evaluate AI-driven processes and ask the right questions about data sources, model assumptions and output reliability.” Jackson emphasizes this evolution reflects broader quality management requirements. “As clients deploy sophisticated GenAI systems, audit firms must demonstrate they have the knowledge and skills to effectively respond to client use of emerging technologies,” he affirms. As AI is further integrated into financial systems, the role of the auditor is essential to promote trust and confidence in the information investors rely on for decision-making. “Auditors will be called on not just to test the numbers prepared by a company,” Kosar notes, “but also to help explain how those numbers came to be.” “Our profession provides independent, objective assurance over complex systems,” Jackson says. “AI adoption represents a new chapter for business leaders, and with it, a new opportunity for the profession to evolve, strengthen trust and build confidence.”

Why It Matters

As cases like NevoGuide’s multiply, auditors remain the critical link between innovation and market trust, ensuring transparency and sustaining confidence in AI-driven financial reporting.

The NevoGuide Solutions case may be fictional, but the challenges are real and ongoing. Every day, auditors perform similar procedures to help investors access reliable, decision-useful information. Their work, though often invisible, forms a crucial foundation of trust, enabling our financial system to function effectively and drive growth in our economy.