Accounting
Both auditors, management must prepare for AI impact
As AI works its way into more and more business processes, it has become increasingly important for auditors to understand where, why, when and how organizations use it and what impact it is having not only on the entity itself but its various stakeholders as well.
Speaking at a virtual conference on AI and finance hosted by Financial Executives International, Ryan Hittner, an audit and assurance principal with Big Four firm Deloitte, noted that since the technology is still relatively new it has not yet had time to significantly impact the audit process. However, given AI’s rapid rate of development and adoption throughout the economy, he expects this will change soon, and it won’t be long before auditors are routinely examining AI systems as a natural part of the engagement. As auditors are preparing for this future, he recommended that companies do as well.
“We expect lots of AI tools to inject themselves into multiple areas. We think most companies should be getting ready for this. If you’re using AI and doing it in a way where no one is aware it is being used, or without controls on top of it, I think there is some risk for audits, both internal and external,” he said.
Andrey Popov/stock.adobe.com
There are several risks that are especially relevant to the audit process. The primary risk, he said, is accuracy. While models are improving in this area, they still have the tendency to make things up, which might be fine for creative writing but terrible for financial data reporting. Second, AI tends to lack transparency, which is especially problematic for auditors, as their decision making process is often opaque, so unlike a human, an AI may not necessarily be able to explain why it classified an invoice a particular way, or how it decided on this specific chart of accounts for that invoice. Finally, there is the fact that AI can be unpredictable. Auditors, he said, are used to processes with consistent steps and consistent results that can be reviewed and tested; AI, however, can produce wildly inconsistent outputs even from the same prompt, making it difficult to test.
This does not mean auditors are helpless, but that they need to adjust their approach. Hittner said that an auditor will likely need to consider the impact of AI on the entity and its internal controls over financial reporting; assess the impact of AI on their risk assessment procedures; consider an entity’s use of AI when identifying relevant controls and AI technologies or applications; and assess the impact of AI on their audit response.
In order to best assist auditors evaluating AI, management should be able to answer relevant questions when it comes to their AI systems. Hittner said auditors might want to know how the entity assesses the appropriate of AI for the intended purpose, what governance controls are in place around the use of AI, how the entity measures and monitors AI performance metrics, whether or how often they backtest the AI system, and what is the level of human oversight over the model and what approach does the entity take for overriding outputs when necessary.
“Management should really be able to answer these kinds of questions,” he said, adding that one of the biggest questions an auditor might ask is “how did the organization get comfortable with the result of what is coming out of this box. Is it a low risk area with lots of review levels? … How do you measure the risk and how do you measure whether something is acceptable for use or not, and what is your threshold? If it’s 100% accurate, that’s pretty good, but no backtesting, no understanding of performance would give auditors pause.”
He also said that it’s important that organizations be transparent about their AI use not just with auditors but stakeholders as well. He said cases are already starting to appear where people unaware that generative AI was producing the information they were reviewing.
Morgan Dove, a Deloitte senior manager within the AI & Algorithmic Assurance practice, stressed the importance of human review and oversight of AI systems, as well as documenting how that oversight works for auditors. When should there be human review? Anywhere in the AI lifecycle, according to Dove.
“Even the most powerful AIs can make mistakes, which is why human review is essential for accuracy and reliability. Depending on use case and model, human review may be incorporated in any stage of the AI lifecycle, starting with data processing and feature selection to development and training, validation and testing, to ongoing use,” she said.
But how does one perform this oversight? Dove said data control is a big part of it, as the quality and accuracy of a model hinges on its data stores. Organizations need to verify the quality, completeness, relevance and accuracy of any data they put into an AI, not just the training data but also what is fed into the AI in its day to day functions.
She also said that organizations need to archive the inputs and outputs of their AI models, without this documentation it becomes very difficult for auditors to review the system because it allows them to trace the inputs to the outputs to test consistency and reliability. When archiving data she said organizations should include details like the name and title of the dataset, and its source. They should also document the prompts fed into the system, with timestamps, so they can possibly be linked with related outputs.
Dove added that effective change management is also essential, as even little changes in model behaviors can create large variations in performance and outputs. It is therefore important to document any changes to the model, along with the rationale for the change, the expected impact and the results of testing, all of which supports a robust audit trail. She said this should be done regardless of whether the organization is using its own proprietary models or a third party vendor model.
“There are maybe two nuances. One is, as you know, vendor solutions are proprietary so that contributes to the black box lack of transparency, and consequently does not provide users with the appropriate visibility … into the testing and how the given model makes decisions. So organizations may need to arrange for additional oversight in outputs made by the AI system in question. The second point is around the integration and adoption of a chosen solution, they need to figure out how they process data from existing systems, they also need to devote necessary resources to train personnel in using the solution and making sure there’s controls at the input and output levels as well as pertinent data integration points,” she said.
When monitoring an AI, what exactly should people be looking for? Dove said people have already developed many different metrics for AI performance. Some include what’s called a SemScore, which measures how similar the meaning of the generated text is to the reference text, BLEU (bilingual evaluation understudy), which measures how many words or phrases in the generated text match the reference text, or ROC-AUC (Receiver Operating Characteristic Area Under the Curve) which measures the overall ability of an AI model to distinguish between positive and negative classes.
Mark Hughes, an audit and assurance consultant with Deloitte, added that humans can also monitor the Character Error Rate, which measures the exact accuracy of an output down to the character (important for processes like calculating the exact dollar amount of an invoice), Word Error Rate, which is similar but does the evaluation at the word level, and the “Levenshtein distance,” defined as the number of single character edits needed to fix an extracted text to see how far away the output is from the ground truth text.
Hittner said that even if an organization is only just experimenting with AI now, it is critical to understand where AI is used, what tools the finance and accounting function have at their disposal to use, and how it will impact the financial statement process.
“Are they just drafting emails, or are they drafting actual parts of the financial statements or management estimates or [are] replacing a control? All these are questions we have to think about,” he said.
Artificial intelligence has indeed led tech-forward firms (including those in this year’s
On the positive end, firms such as the Texas-based Franklin Alliance reported that adopting AI technology has dramatically increased their capacities as bots take on repetitive manual tasks with an ease and a speed far past more conventional automation setups, allowing accountants to focus more on higher value tasks.
“What’s been most impressive about the AI tools we’ve explored is their ability to dramatically reduce the time spent on repetitive, manual tasks—things like document summarization, data extraction, and even early-stage tax prep. In the right context, these tools create real efficiency gains and allow our team to shift focus to higher-value advisory work,” said Benjamin Holloway, co-founder of Texas-based Franklin Alliance.
madedee – stock.adobe.com
For some, like Illinois-based Mowery & Schoenfeld, these efficiencies have been most impressive on the internal administration side, with AI effectively taking care of the non-accounting work that nonetheless keeps many firms afloat, especially where it concerns meetings.
“Truly most impressive and a huge time savings for us has been AI’s ability to record and summarize Team meetings. Circulating notes and reducing administrative burden on such activities has freed up much capacity, both for our admin side and for partners or management who are not able to be at every meeting,” said Chris Madden, director of information technology.
Others, like top 10 firm Grant Thornton, emphasized AI’s benefits in client-facing activities and noted that it has been especially meaningful in its risk advisory services at least partially due to the firm’s recently-launched CompliAI tool, designed specifically for this area.
“The tool uses generative artificial intelligence and was developed using Microsoft technology, including Microsoft Azure OpenAI Service. CompliAI’s ability to quickly analyze vast datasets and identify potential risks has proven invaluable in combining Grant Thornton’s extensive global controls library with generative AI models and features, including AI analysis, ranking and natural language processing capabilities. As a result, our employees can run control design and assessment tasks in minutes, versus days or weeks. This means clients enjoy faster operational insights, which could amount to a new level of efficiency and a path toward transformative growth,” said Mike Kempke, GT’s chief information officer.
Another positive frequently mentioned, such as by top 25 firm Cherry Bekaert, has been the accessibility and ease of use for many AI solutions even for those without strong technical capacities. Assurance partner Jonathan Kraftchick said this means they did not need to wait long before they began seeing results.
“The most impressive aspect of AI has been its ability to add value with minimal ramp-up time. Many of the tools we’ve implemented have a low barrier to entry, allowing users to start experimenting and seeing results almost immediately. Whether it’s drafting content, conducting accounting research, summarizing meetings, normalizing data, or detecting anomalies, AI has consistently helped accelerate tasks and enable our teams to focus on higher-risk or higher-value areas,” he said.
Several firms, such as California-based Navolio & Tallman, also mentioned improvements to broad strategy and ideation, saying it’s been good for enhancing creativity and accelerating the early stages of their work.
“We’ve still seen value in AI as a jumping off point for ideas and strategy. It’s been helpful for brainstorming, drafting early versions of client communications, and supporting high-level planning conversations,” said IT partner Stephanie Ringrose.
Inconsistencies, inaccuracies, insufficiency, and insecurity
At the same time, firms over and over again said that while the strength of AI comes in handling simple jobs, it often lacks the precision and consistent accuracy needed for higher value accounting work. While it can certainly generate outputs at an industrial scale, trusting that those outputs are correct is another story for firms like Community CPA and Associates.
“AI is incredibly useful for certain types of tasks, such as summarization, data extraction, answering simple questions, drafting communications or documentation, brainstorming ideas, or serving as a sounding board. However, we have observed that most AI tools we’ve tried have difficulty with complex tasks that require lots of context, precision, or domain-specific knowledge. Oftentimes in these cases, AI tools will generate responses that are overly confident or wrong and are missing key information due to not being integrated with other systems or software we have,” said CEO Ying Sa.
Some, like top 25 firm Armanino, noted that these challenges mean that humans need to devote considerable time to ensuring the quality of AI outputs and intervening when the programs go off track.
“The primary disappointment stems from the occasional inaccuracies or biases inherent in AI-generated outputs, commonly referred to as ‘hallucinations,’ necessitating continuous human oversight to ensure reliability. Addressing these inconsistencies remains an ongoing challenge,” said Jim Nagata, senior director of cybersecurity and IT operations.
Top 25 firm Eisner Amper’s chief technology officer Sanjay Desai noted that these issues with accuracy and consistency can be found across AI solutions, though noted that the technology is still quite new and so many things are still in the process of being refined.
“The lows come from the gap between what’s possible and what works reliably in practice. We still need strong guardrails to define valid inputs and outputs, especially in sensitive use cases. Technologies like retrieval augmented generation (RAG) haven’t yet delivered the accuracy or consistency we need when working with proprietary or domain-specific data. Even in mature areas like audio-to-text transcription, we see issues—particularly with accurately identifying speakers in multi-person meetings, which affects the quality of recaps and follow-up actions. In short, while LLMs have come a long way, making them enterprise-ready still requires ongoing human oversight, thoughtful implementation, and continuous refinement,” said Desai.
Another issue reported by several firms was what firms like Navolio & Tallman saw as ongoing security risks from AI solutions that limits their ability to apply the technology to more sensitive use cases.
“The overall attention to security and privacy is still more limited than our industry requires, vendors have not yet aligned their pricing models with the impact their tools make to the business, and vendors still oversell their AI capabilities,” she said.
Top 25 firm Citrin Cooperman also noted–among other things–that the security of these solutions could stand to improve.
“The overall attention to security and privacy is still more limited than our industry requires, vendors have not yet aligned their pricing models with the impact their tools make to the business, and vendors still oversell their AI capabilities,” said chief information officer Kimberly Paul.
Another issue with AI that firms have reported is that solutions today don’t seem to integrate especially well with other programs, which limits the ability of these solutions to work across multiple systems in a single coherent workflow–under such conditions, AI solutions can wind up being siloed from the very areas it is needed the most.
“We believe one of the biggest gaps in current AI solutions is the inability to integrate into other AI solutions to work collectively across one process or workflow. There are many cases where one AI solution is very good at a specific task, while another is very good at another process or task, but the gap is the ability to integrate those solutions together to solve for an entirety of a process or a workflow,” said Brent McDaniel, chief digital officer for top 25 firm Aprio.
There is also the matter of data integration, which is needed for AI systems to gain a more holistic understanding of a firm’s needs. Without such integrations, AI becomes more limited in its ability to develop insights and provide actionable guidance, according to Tom Hasard, IT shareholder for New Jersey-based Wilken Gutenplan.
“We wish AI tools could fully synthesize all of our internal data and unique expertise—beyond the scope of general internet search—and provide detailed, context-specific answers for our team. In the near term, we envision an internal system that taps into our accumulated knowledge to assist staff in resolving complex client problems more quickly. Over time, this capability could be extended to give clients direct, on-demand access to our specialized insights, effectively scaling our expertise and delivering value in a more immediate and personalized way,” he said.
Beyond just data, lack of integration also limits the ability for AI to address complex problems due to lack of cross-disciplinary expertise, according to Kempke from Grant Thornton.
“Current AI solutions lack the deep cross-disciplinary expertise to be able to solve complex issues. AI today is optimized for specific fields and tasks but when it comes to solving problems that span multiple disciplines such as Tax, Legal and Finance, the current solutions are not yet capable of providing meaningful advice and guidance. Grant Thornton is already working with various AI partners on this issue and targets to be a very early adopter of the next iteration of AI that addresses this,” he said.
The AI wishlist
Many firms hoped that the next generation of AI solutions would address these sorts of problems in a way that will allow them to become true assistants capable of taking on complex tasks that require extensive judgment.
“We have found that AI currently lacks in the ability to replicate human creativity and complex decision-making. While AI excels at data analysis and task automation, it struggles with tasks requiring creativity and nuanced judgment. If AI could offer more sophisticated support in areas such as accounting and audit services, its value and impact in our daily lives would be significantly enhanced,” said Jim Meade, CEO of top 50 firm LBMC.
Desai, from Eisner Amper, also pointed out that AI isn’t very good at handling bad data, which is a problem considering that AIs run on data. This means that using AI effectively today still requires a great deal of data processing and sanitation to make information useful. If humans did not need to do so much manual cleanup to get data AI-ready, it would help make the technology even more efficient.
“One of the biggest gaps in AI today is its limited ability to handle bad data. Since data is the foundation of any AI strategy, it’s a challenge that most organizations still face— dealing with messy, inconsistent, or unstructured data. We wish AI could do more to identify, fix, and improve data quality automatically, instead of relying so much on manual cleanup,” said Desai.
Finally, Avani Desai, CEO of top 50 firm Schellman, said that AI needs to not only be safer, it needs to be visibly so, as trust and confidence in the technology is often key to adoption.
“I wish that AI could de-risk itself so that clients would be more open to using it and build client trust. If AI could more clearly demonstrate safety and responsible use, adoption would be much easier. Once people understand it’s here to help—and learn to use it responsibly—the fear will fade,” she said.
Representative Nicole Malliotakis said increasing the state and local tax deduction cap to $30,000 from $10,000 would reduce the tax burden of the vast majority of people in her district, indicating support for a proposal that is dividing Republicans.
“Every member needs to advocate for the particular needs of their district. Tripling the deduction to $30,000 will provide much-needed relief for the middle-class and cover 98% of the families in my district,” Malliotakis, a Republican representing Staten Island, New York and a member of the House tax committee, said in a statement to Bloomberg News on Friday.
Malliotakis’ nod of approval for a $30,000 SALT deduction cap comes as Republicans are fighting among themselves about how high to increase a tax break that has the potential to scuttle President Donald Trump’s entire tax package.
House Speaker Mike Johnson on Thursday said the $30,000 write-off limit is one of several options being discussed. That figure was rejected by several other New York Republicans, including Elise Stefanik, Nick LaLota, Mike Lawler and Andrew Garbarino. California’s Young Kim also rebuffed the idea.
Malliotakis’ district has less expensive property values and lower incomes than some of the other lawmakers pushing for a SALT expansion, making it politically viable for her to accept a lower cap than some of her colleagues.
White House Press Secretary Karoline Leavitt suggested on Friday that Trump would not weigh in on an appropriate level for a SALT cap, leaving it to lawmakers to resolve.
“There’s a lot of disagreement on Capitol Hill right now about the SALT tax proposal, and we will let them work it out,” she told reporters.
House Republicans’ narrow majority means that Johnson needs to win the support of nearly all his members to pass Trump’s tax-and-spending package.
Several of the SALT advocates have said that they are willing to block the bill unless there is a sufficient increase to the deduction. However, most members have not publicly stated how high the deduction must be to win their support.
The debate over SALT has proved to be a particularly thorny fight because it is a political priority for a small but vocal group of Republicans representing swing districts critical to the party maintaining a majority in the 2026 midterm elections.
Expanding the write-off is an expensive proposition, and Republicans have little fiscal wiggle room as they are sparring over ways — including cuts to Medicaid and levy hikes on millionaires — to offset the cost of the tax-cut package.
The House Ways and Means Committee is slated to consider the tax portion of the bill on Tuesday, including SALT changes.
Essential Strategies for Maintaining Data Security in Modern Bookkeeping
Investor Ric Edelman reacts to crypto ETF boom
The key issues and who stands to benefit
New 2023 K-1 instructions stir the CAMT pot for partnerships and corporations
The Essential Practice of Bank and Credit Card Statement Reconciliation
Are American progressives making themselves sad?
-
Finance1 week agoWarren Buffett to ask board to make Greg Abel CEO of Berkshire Hathaway at year-end
-
Personal Finance1 week agoMillions of older workers lost jobs during Covid. Prospects have improved
-
Finance1 week agoBerkshire meeting ‘bazaar’ features Buffett Squishmallows, 60th anniversary book and giant claw machine
-
Economics1 week agoWhy does America have birthright citizenship?
-
Economics1 week agoHow Donald Trump could rescue John Roberts
-
Economics1 week agoJobs report Friday to provide important clues on where the economy is heading
-
Accounting1 week agoThe Importance of Backing Up Bookkeeping Data
-
Economics1 week agoHow one Ivy League university avoided the president’s wrath