Earmark CPE

Earn CPE Anytime, Anywhere

AI

When Bots Listen to Robots and Real Money Disappears

· ·

Picture this: a computer on stage playing songs to an audience of computers. No humans involved, just machines performing for machines in an endless digital loop. Yet somehow, millions of dollars change hands.

This isn’t science fiction. It’s happening right now on streaming platforms, and it’s just one of the mind-bending fraud schemes explored in this episode of Oh My Fraud. Host Caleb Newquist opens with a relatively new conspiracy theory called the Dead Internet, which suggests that most online activity, including posts, likes, followers, and streams, isn’t human anymore. It’s “bots talking to bots, talking to bots,” creating an information superhighway filled with self-driving cars that have destinations but no passengers.

But what happens when someone exploits this artificial ecosystem for real money? That’s exactly what we’re about to find out.

The $121 Million Email That Fooled Silicon Valley

Between 2013 and 2015, a Lithuanian man named Evaldas Rimašauskas pulled off something that shouldn’t have been possible. He convinced two of the world’s smartest companies, Google and Facebook, to wire him $121 million. His method wasn’t sophisticated hacking or complex algorithms. He simply pretended to be someone else.

Rimašauskas impersonated Quanta Computer, a real Taiwan-based hardware manufacturer that actually did business with both tech giants. He set up a company in Latvia under Quanta’s name and opened bank accounts in Latvia and Cyprus. Then his team got to work, calling Google and Facebook customer service lines to gather intelligence, including names of key employees, contact information, and other details that would make their lie believable.

Through phishing emails and what Caleb describes as “a maze of phony invoices, contracts, letters, and corporate stamps,” Rimašauskas created enough confusion to convince someone at Google to update the bank account they had on file for Quanta Computer. In 2013, Google sent $23 million to his account. Two years later, using the same playbook, Facebook wired him $98 million.

The money flowed through accounts across Latvia, Cyprus, Slovakia, Lithuania, Hungary, and Hong Kong. And here’s the kicker: these amounts were so insignificant to Google and Facebook that they “went virtually unnoticed.” As Caleb puts it, “$23 million and $98 million aren’t even rounding errors on the amount of revenue for Google and Facebook. It’s less than pocket change.”

Eventually, someone at Google caught on. Rimašauskas was arrested in March 2017, extradited to the U.S. that August, and pleaded guilty to wire fraud in March 2019. He got five years in prison, and both companies got their money back.

From IT Mogul to Music “Producer” to Alleged Fraudster

Our second story shifts from simple impersonation to something far stranger. Meet Michael Smith, a 52-year-old with a resume that reads like three different people’s lives smashed together.

According to the research, Smith made his first fortune in the 1990s with an IT business where he allegedly wrote “one of the main fixes for the Y2K millennium software bug.” He then ran chains of medical clinics, which landed him in trouble in 2020 when he and two associates paid $900,000 to settle Medicare and Medicaid fraud allegations.

But here’s where it gets weird. At age 39, Smith decided to become a music industry player. Despite having no apparent musical background, he somehow ended up judging a BET hip-hop competition called “One Shot” alongside DJ Khaled, T.I., and Twista. As Wired magazine described it, he was “a relatively unknown record producer with a checkbook” among actual stars.

When Caleb asked producer Zach Frank if he’d ever heard of anyone building a successful music career starting in middle age, Zach’s response was telling: “It’s extremely, extremely rare. Not without money, at least.”

The Streaming Revolution and Its Discontents

To understand Smith’s alleged fraud, you need to understand how dramatically the music industry has changed. Zach, who comes from a family of professional musicians, explained how streaming completely upended the business model.

In the old days, people bought physical albums for $12-15 at stores like Tower Records. Artists made real money from album sales. Then came Napster and peer-to-peer sharing, which Caleb admits using extensively in college. “People were listening to all this music completely in its entirety for free,” he recalls.

Today’s streaming platforms like Spotify and Apple Music operate on a subscription model. Users pay monthly fees for unlimited access, and artists get fractions of pennies per stream. Spotify made $17 billion in 2024 and claims 70% goes to the music industry, but individual artists see almost nothing.

The numbers are staggering. According to Spotify’s former chief economist, more music is released every single day in 2025 than in the entire year of 1989. And here’s what makes it worse: bigger artists negotiate better deals, while smaller artists, as Zach puts it, “get screwed.”

Building an Army of Fake Listeners

This is the landscape Smith allegedly decided to exploit. Starting in 2017, he orchestrated what the Department of Justice calls a scheme to steal millions in royalties by fraudulently inflating music streams.

The mechanics were brilliant in their simplicity. First, Smith created thousands of bot accounts using fake email addresses and names. He even told a coconspirator to “make up names and addresses” but to “make sure everyone is over 18.” He paid $1.3 million in subscription fees because, as Zach explains, paid subscribers generate higher royalty rates than free users.

By October 2017, Smith had 1,040 bot accounts spread across 52 cloud service accounts. Each bot could stream about 636 songs per day, generating approximately 661,440 total daily streams. At half a cent per stream, that meant $3,307 daily, $99,000 monthly, or $1.2 million annually.

But Smith had a problem: he needed content. Lots of it.

When AI Makes Music for Bots to Hear

Initially, Smith used music catalogs from coconspirators and even tried selling his streaming service to other musicians desperate for plays. But as he wrote in May 2019, “I can’t run the bots without content and I need enough content so I don’t overrun each song. If we get too many streams on one song, it comes down.”

His solution? Artificial intelligence. Smith partnered with Alex Mitchell, CEO of an AI music company called Boomy, who began providing thousands of AI-generated songs each week.

The song and artist names were gloriously terrible. Song titles included “Zygotic Washstands,” “Zygoptera,” and “Calvinistic Dust.” Band names ranged from “Calm Knuckles” to “Camel Edible.” As Caleb jokes, “I don’t know what camel edibles are. Perhaps they are THC gummies for camels.”

To demonstrate just how far AI music has come, Zach used Udio.com during the podcast to generate two complete songs about Oh My Fraud in just 10-15 seconds. The results were unnervingly good, professional-sounding tracks that could easily pass for human-created music. “There’s a lot of AI music on Spotify at the moment without people knowing it’s AI,” Zach notes.

Smith used VPNs to hide that all streams came from one location and spread activity across thousands of songs to avoid detection. When flagged for “streaming abuse” in 2018, he protested: “We have no intentions of committing streaming fraud.”

By February 2024, Smith’s scheme had generated 4 billion streams and $12 million in royalties.

Folk Hero or Fraudster?

The reaction to Smith’s indictment has been surprisingly divided. Some see him as a criminal who stole from real artists through the “stream share” system, where royalties are distributed based on each rightsholder’s proportion of total streams. Others view him as a folk hero exposing an exploitative system.

The case raises uncomfortable questions. When the band Vulfpeck released an album of complete silence and asked fans to stream it while sleeping—earning $20,000 before Spotify banned them—was that fraud or performance art? As Zach asks, “If someone’s playing blank music, who are they to say that’s not real?”

Smith has hired the prestigious law firm that defended Diddy and plans to fight the charges vigorously. This will be the first major streaming fraud case fully litigated, potentially setting precedents for how we define fraud in digital spaces.

What We Learned

As Caleb reflects at the episode’s end, these cases reveal something profound about our digital economy. Google and Facebook, companies worth trillions with founders worth hundreds of billions, got tricked by simple schemes. A middle-aged entrepreneur with a checkbook created a phantom musical empire that earned millions.

For accounting professionals, these are warnings about the future of fraud detection. When documentation can be perfectly faked, when bots are indistinguishable from humans, when AI creates content that only machines consume, traditional audit procedures become obsolete.

These cases force us to confront questions about power, technology, and authenticity in the digital age. When companies make billions while creators earn pennies, algorithms determine value instead of human appreciation, and the line between real and artificial completely disappears, that’s when people start rooting for the fraudsters. Not because they’re right, but because the system itself feels so wrong.

Listen to the full episode to hear Caleb and Zach grapple with these questions, including those AI-generated songs that sound disturbingly human. Because in an age where machines create for machines while extracting real value from real people, understanding these frauds helps preserve what makes us human in an increasingly artificial world.

Your CPA Exam Scores Might Be Lost and Your AI Bookkeeper Is 57% Accurate

· ·

“No kings means no paychecks, no paychecks, no government.” When Treasury Secretary nominee Scott Bessent dropped this line in a Fox News interview, Blake Oliver and David Leary weren’t sure if they should laugh or be terrified. As David put it: “That’s the most un-American thing anybody could say.”

In episode 458 of The Accounting Podcast, Blake and David dig into a series of accountability failures that would be funny if they weren’t so serious. From the Trump administration creating a brand new IRS “CEO” position to dodge Senate confirmation, to NASBA somehow losing track of CPA exam scores, the organizations supposed to maintain standards can’t even maintain their own data.

The IRS Gets a CEO (Because Who Needs the Constitution?)

The Trump administration’s latest move isn’t subtle. It created a new “CEO” position for the IRS that doesn’t require Senate confirmation. As Blake explains, “If the president just creates a new role that has the same responsibilities but doesn’t get checked by the Senate, then that’s just a run around the rules.”

The plan goes deeper than personnel changes. Gary Shapley, an advisor to Treasury Secretary nominee Scott Bessent, wants to weaken IRS lawyers’ involvement in criminal investigations and eliminate extra procedural steps for sensitive cases involving elected officials and tax-exempt groups. These aren’t reforms—they’re removing the safety rails.

“Where’s the AICPA on this?” David asks. The AICPA wrote a letter about the government shutdown’s impact on taxpayers but stayed silent on bypassing Congress to appoint IRS leadership. Blake doesn’t mince words: “They don’t. They are not willing to take a stand on something that matters because they’re afraid of political blowback.”

According to Wall Street Journal reporting that Blake and David discuss, Shapely has already compiled a hit list. The targets? George Soros and affiliated organizations, major Democratic donors, and left-leaning nonprofit groups.

The hosts make an important point that transcends politics. “The Obama administration targeted right wing groups,” Blake notes, agreeing with a viewer comment. “This is why you don’t want to give the government too much power. The other side gets the gun eventually, then points it at the other side.”

When Accounting Organizations Can’t Do Accounting

If you think government accountability is bad, wait until you hear about the profession’s own organizations.

Professor Joseph Ugrin, who creates the CPA Success Index published by Accounting Today, discovered NASBA’s 2024 data is essentially garbage. Between 25% and 40% of candidate scores are simply missing. Plus, Iowa community colleges appear in the data despite state law requiring bachelor’s degrees to sit for the exam.

“NASBA has access to all the transcripts submitted by the candidates,” Blake points out. “So there’s no reason why they couldn’t correctly classify what schools they went to.”

David speculates, “This smells like somebody at NASBA tried to use AI to summarize some stuff and screwed it up.” Whether it’s AI or old-fashioned incompetence, Ugrin can’t publish the Success Index this year because the data is unusable.

Meanwhile, the Chicago Teachers Union hasn’t released required financial audits for over five years, despite paying $80,000 for audit services in 2025 alone. When members finally got federal filings, they showed only 18% of spending goes to representing teachers. The other 82%? Overhead, politics, and “leadership priorities.”

As David asks incredulously: “How did it go past one year?”

The issue isn’t confined to Chicago. Forty-three Arkansas cities can’t get state funds because they can’t find CPAs to do required audits. “The auditors are retiring. They’re not being replaced,” Blake explains. Small-town America is literally running out of accountants.

AI to the Rescue! (Just Kidding, It’s 57% Accurate)

While real problems go unsolved, the profession is being sold AI magic beans.

One marketing CEO’s experience with QuickBooks’ new AI features reads like a horror story. “Although trained on transactions, QuickBooks frequently miscategorized payments based solely on dollar value,” he wrote. If a vendor sent one $1,000 invoice, the AI recorded all future invoices as $1,000. Contractor payments were recorded under “QuickBooks payments” instead of the contractor’s name. The company spent thousands on accountants trying to fix problems that couldn’t be fixed.

“QuickBooks sits at the heart of our business,” the CEO explained. “When AI upgrades destabilize that core, the consequences ripple across the organization.”

The hosts shared another headline that calls AI’s accuracy into question. Microsoft’s AI agent in Excel achieves 57.2% accuracy on spreadsheet benchmarks. As Blake says: “57.2% accuracy is not going to cut it. Not even 98% accuracy is going to cut it.”

Yet companies like Docyt claim AI will let one accountant manage 300 clients. The hosts’ response? “I’ve talked to firm owners that are super efficient,” David says. “Their best bookkeepers maybe handle 45 clients a month.”

Blake’s experience backs this up: “A typical bookkeeper could do 20 to 30 on average. And my all star could do 40 to 60.” The idea of 300 clients per person? “You would have too many questions coming in emails,” Blake explains. “I don’t think there’s an AI tool that can do that.”

Blake’s ideal practice would have ten outsourced controller clients, meeting weekly with each. “Once I got the ten clients, I could probably do it in four hours a day.” That’s realistic. Managing 300 clients with AI? That’s fantasy.

The hosts haven’t seen AI actually eliminating jobs. “I have yet to talk to an accountant that says, oh, we implemented this thing and now we got rid of two of my staff,” David states. Even at their own company, which uses AI extensively: “We’re not getting rid of anybody. We just hired more engineers.”

The $300 Trillion Oops

Just when you thought it couldn’t get wilder, David shares the stablecoin story that should terrify everyone.

Paxos, which provides stablecoin infrastructure for PayPal, accidentally minted $300 trillion in stablecoins. Not million. Not billion. Trillion. For context, the US deficit is $2 trillion.

“You understand how a stablecoin works in theory.” David says. “A dollar goes in, you get a stablecoin worth a dollar back. What if I told you none of that is true?”

The company claimed it was a “technical error that briefly appeared for 20 minutes,” then they “burned” the excess tokens. But as David points out, if companies can just create and destroy them at will, this proves stablecoins aren’t actually backed by dollars.

This matters because Ripple just bought a treasury management firm for $1 billion, putting cryptocurrency at the center of corporate cash management. “Accountants are going to be touching this stuff,” David warns. “It’s going to be here next year.”

Time to Pay Attention

This episode of The Accounting Podcast is a reality check for a profession facing multiple crises simultaneously. The IRS is being restructured to avoid constitutional oversight. Professional organizations can’t maintain basic data integrity. AI is being forced on businesses with disastrous results. And small towns can’t find CPAs to do basic audits.

“We don’t need a king,” David emphasizes about Bessent’s comments. But between government overreach, organizational incompetence, and technological snake oil, the profession is being pulled in all the wrong directions.

The hosts’ frustration is justified. When Blake asks why the AICPA won’t stand up for constitutional principles, when David wonders how organizations go years without audits, when they both laugh at the idea of one person managing 300 clients, they’re asking the questions the profession should be asking itself.

Listen to the full episode to hear Blake and David’s complete breakdown of these interconnected failures. In a profession built on trust and verification, their willingness to be brutally honest is exactly what’s needed.

After 50 Years in Internal Audit, Richard Chambers Sees the Profession’s Greatest Risk Yet

· ·

“Who’s going to provide the skepticism, the intellectual curiosity, and the institutional knowledge to our audit teams in ten years? Because the rest of us are going to be gone.”

Richard Chambers drops this stark warning after 50 years in internal audit. His concern isn’t about losing jobs to technology. It’s about the growing gap between how we’ve always trained auditors and what the profession now demands.

On this episode of the Earmark Podcast, host Blake Oliver sat down with Richard, Senior Advisor for Risk and Audit at AuditBoard. He brings a unique view of internal audit’s transformation. When he started in 1974, fresh out of college and working in a bank’s internal audit department, the job was all about checking financial records and looking backward. Today? Financial risks make up only 25% of audit plans. The rest involves cyber threats, AI governance, supply chain chaos, and what Richard calls “perma-crisis”—our new normal where tariff rates can change three times in a single day.

Most companies use AI, but only a quarter have set up proper governance over it, according to AuditBoard research. That gap presents massive risk and opportunity for internal auditors who can bridge it.

From Bean Counting to Risk Navigation

Internal audit has changed dramatically since Richard joined that bank in 1974. Back then, it was all ledgers and reconciliations—purely financial work focused on last year’s numbers. Today, financial risks are just a quarter of what internal auditors examine.

“The profession has matured,” Richard explains. “While we still do some work in the financial space, that’s really a small percentage of internal audit’s focus.”

The real game-changer has been what Richard calls “perma-crisis.” It started with the COVID-19 pandemic and hasn’t stopped. “We’ve been lurching from one risk-induced disruption to another,” he says, listing the cascade: pandemic, forty-year-high inflation, supply chain breakdowns, wars in Europe and the Middle East. “We’re in our sixth year of it, and I would submit this is the new normal.”

This constant chaos makes traditional planning almost useless. Richard found that nearly 60% of internal audit departments had already changed their 2025 plans by May. When tariff rates can swing wildly in a single day—Richard recalls hearing three different numbers from Washington in one day—annual planning is dangerous.

“You can no longer have any confidence that one scenario is the only one you have to worry about,” Richard emphasizes. Organizations need what he calls “scenario risk management,” or planning for multiple possible futures at once.

This need for flexibility shifts how internal audit works with other departments. The old model was called “three lines of defense”: management controlled risks (first line), oversight functions monitored them (second line), and internal audit was the last barrier before disaster (third line).

But pure defense isn’t enough anymore. In 2019, the Institute of Internal Auditors dropped “defense” from the name. The new message? “Independence does not mean isolation.”

Richard uses a ship analogy that really hits home. Organizations are like vessels at sea that need lookouts watching in all directions and talking to each other. “If your internal auditors are looking in one direction and your risk managers are looking in another,” he warns, “but they aren’t sharing what they’re seeing, then you don’t know whether there are gaps.”

AI: The Top Risk and Best Opportunity

Three years ago, AI wasn’t even on internal audit’s risk list. Today, it’s number one, pushing even the talent crisis to second place.

“Pre-2022, before ChatGPT came out, we weren’t asking about it,” Richard admits. Once he started surveying the profession, AI rocketed up the list: middle of the pack the first year, third place the next, then straight to number one.

This isn’t just another tech disruption. After watching five decades of change, Richard doesn’t mince words: “In the five decades I’ve been in internal audit, there’s never been a greater risk to this profession in terms of becoming irrelevant.”

The scariest part? When Richard asks why audit teams aren’t using AI more, the top answer is, “We don’t really understand it enough.” That hesitation could be fatal.

Yet Richard himself uses AI daily as his “research assistant.” He asks it to identify industry risks, outline articles, analyze data. “It takes me longer to write the prompts than it takes to give me the answer,” he notes.

The use cases are obvious and powerful. Risk assessments that used to happen annually can now be continuous. AI can scan for threats humans would never spot. Data analysis that took weeks happens in minutes. Even audit reports can be AI-generated.

But the trap is that AI excels at exactly the work that trains new auditors. Entry-level graduates traditionally learned by doing routine tasks. Now AI does those tasks better and faster.

“College graduates have traditionally been able to ease into professions by doing some of the more rudimentary tasks,” Richard explains. “But AI is prime for rudimentary tasks.”

This creates a vicious cycle. Companies hire fewer entry-level auditors. Without that pipeline, who develops the judgment for complex work? Richard’ solution: “We shouldn’t refrain from hiring them. We should be willing to bring them in and help them leap the learning curve.”

“AI won’t replace internal auditors,” Richard predicts, “but it will replace internal auditors who don’t use it.”

The Human Superpowers AI Can’t Touch

“To assess culture, you also have to be able to rely on your sense of smell.”

A chairman of the board of a large Indian company shared this wisdom with Richard years ago, and it perfectly captures what separates humans from AI. Technology can analyze documents and data. But it takes human instinct to sense what happens when nobody’s watching.

Richard identifies three “human superpowers” that AI cannot replicate: professional skepticism, intellectual curiosity, and relationship skills. These aren’t soft skills; they’re the core value of internal audit.

Take culture assessment. Richard has done two major research projects showing how toxic culture can destroy organizations. But judging culture requires reading between lines, sensing unspoken tensions, and understanding human motivations. As Blake pointed out during the conversation, “The body language, the way people talk to each other, all of that is context that AI just cannot have access to.”

The audit committee relationship shows this even more clearly. Richard chairs an audit committee and knows these relationships need more than data transfer. They require courage to “grab them by the face” and focus them on hidden risks.

“If we’re content to just answer the questions they ask,” Richard warns, “then we’re not really serving our organizations well. We have to help them understand the questions they need to be asking.”

This shift, from giving answers to finding the right questions, represents a huge evolution. While AI can list potential questions, there’s something fundamentally human about knowing which questions matter.

Most critically, Richard identifies one role that must stay human: assessing AI’s own governance. “I shudder to think that there may be a day where we ask AI to assess its own governance,” he says. “We would never do that with anyone else.”

The challenge is developing these human skills when the traditional path is disappearing. Without routine work to learn on, how do new auditors develop judgment?

We need to help new auditors develop skepticism, intellectual curiosity, and institutional knowledge from day one. Teach them to ask “why” before teaching them “how.”

As Richard reflects after 50 years, “What a difference from the bean counter view of internal audit. You get to be so curious as an internal auditor these days.”

The Next 50 Years Start Now

Richard’s journey from a bank to internal audit’s leading voice shows a profession that has transformed before and must do so again.

The collision of perma-crisis and AI doesn’t doom internal audit. It clarifies its purpose. When tariffs change three times daily, cyber threats evolve by the hour, and AI makes decisions we don’t fully understand, organizations desperately need professionals who ask the hard questions.

Not “What does the data say?” but “What isn’t the data telling us?” Not “How do we implement AI?” but “How do we govern what we can’t fully understand?”

The saying “independence does not mean isolation” applies to both organizational relationships and the human-AI partnership. Tomorrow’s successful auditors won’t resist AI or surrender to it. They’ll orchestrate a sophisticated dance between computational power and human intuition.

The fact that entry-level work is vanishing while judgment becomes more critical demands new thinking about professional development. Organizations can’t wait for fully-formed auditors. They must cultivate intellectual curiosity from day one.

For accounting and tax professionals watching internal audit’s future, Richard warns those who avoid or fear AI will become irrelevant. But he also extends an invitation: those who combine technology with human capabilities will find themselves at the center of organizational decision-making.

Listen to the complete conversation to understand why this moment represents internal audit’s greatest challenge and its most exciting opportunity. After five decades in the profession, Richard reminds us the question isn’t whether internal audit will survive the age of AI. It’s whether individual auditors will choose to evolve with it.

Deloitte’s $440,000 AI Fabrication Scandal Exposes the Accounting Profession’s Deepest Fears

· ·

A startup founder discovered $2.1 million in embezzlement by his co-founder in just 18 minutes using Claude AI. The company’s internal auditors, external auditors, and even the CFO had completely missed it. Meanwhile, Deloitte was forced to refund the Australian government hundreds of thousands of dollars after delivering a report filled with AI-generated fabrications.

In this episode of The Accounting Podcast, hosts Blake Oliver and David Leary dig into these stories. They explore how AI is both exposing massive frauds and creating embarrassing failures, examine the chaos from the government shutdown, and question whether traditional accounting services still matter when 86% of major companies use broken charts that nobody even notices.

When AI Catches What Humans Miss (And Creates What Shouldn’t Exist)

The accounting profession is experiencing an AI identity crisis. On one hand, artificial intelligence can spot complex fraud that teams of professionals completely miss. On the other hand, professionals are using it to generate work that looks legitimate but is actually riddled with fabrications.

Let’s start with Deloitte’s spectacular failure. The Big Four firm charged the Australian government $440,000 AUD (about $290,000 USD) for a 237-page report on welfare compliance systems. The problem? It contained over 20 AI-generated errors, including completely made-up quotes from federal court judgments and references to non-existent academic papers.

Chris Rudge, a Sydney University researcher, spotted the errors immediately. One fabrication attributed a non-existent book to constitutional law professor Lisa Burton Crawford on a topic completely outside her field. “I instantaneously knew it was either hallucinated by AI or the world’s best kept secret because I’d never heard of the book, and it sounded preposterous,” Rudge said.

Even after getting caught, Deloitte insisted its findings and recommendations were still valid. This prompted Australian Labor Senator Deborah O’Neill to observe that Deloitte has “a human intelligence problem.”

But here’s where it gets interesting. While Deloitte was using AI to create fake references, a startup founder used it to uncover real fraud. He exported his company’s QuickBooks data into Claude AI and asked one simple question: “What’s wrong with this picture?”

In just 18 minutes, the AI found what everyone else had missed: 17 fake companies routing $2.1 million to his co-founder’s personal accounts through shell companies. The AI spotted patterns humans overlooked, including fake vendors paid on 23-day cycles while real vendors were paid on 28-day cycles, and payment amounts that followed Fibonacci sequences, which humans subconsciously create when making up numbers.

The founder has since turned this into a business, selling AI-powered fraud detection prompts for $10,000 each to 47 clients. He’s probably making more money from his fraud-detection business than from his original startup.

As Leary points out, this creates both an opportunity and a threat for accounting firms. “The real risk of AI taking accounting jobs isn’t that AI will take the job away. Clients are just going to say, ‘I can do that myself. I don’t need to pay somebody $400,000 to do a half-assed ChatGPT thing.’”

Government Shutdown: When Critical Systems Break Down

The conversation then turned to the government shutdown’s impact on air travel and tax services. The situation has become genuinely dangerous, with cascading failures that reveal how fragile our systems really are.

Air traffic controller-related delays jumped from a typical 5% to 53% as workers called in sick rather than work without pay. Oliver experienced this firsthand when his flight was delayed for hours with no official explanation, though flight attendants privately blamed air traffic control shortages.

The scariest incident happened at Burbank Airport in Los Angeles, where the tower went completely unmanned. “When that happens, there is a backup procedure, which is that the pilots have to do their own air traffic control,” Oliver explains. “They get on a shared frequency and have to communicate with each other. There’s no intermediary. So that not only slows things down. It also creates risk. There’s a huge risk of these planes crashing into each other because they miscommunicate.”

The economic impact is staggering. The US Travel Association estimates $1 billion in weekly losses to the travel economy. Over 750,000 federal workers have been furloughed, while more than a million work without pay. For TSA screeners earning an average of $51,000, the situation is untenable. “If they don’t get paid, they are not paying their bills,” Oliver notes. “They’re going to go drive for Uber to pay the bills.”

The IRS shutdown creates serious problems for accountants. Nearly half of IRS staff have been furloughed. While electronic returns continue processing and automated refunds still flow, human support has collapsed. Phone support is essentially gone, paper returns sit unprocessed, and audits have stopped. Yet interest and penalties continue to accrue, and all deadlines remain in effect.

Adding to the chaos, Trump fired over 4,100 federal workers instead of furloughing them. The Treasury alone lost 1,446 employees, including about 1,300 IRS workers. “It’s the first time in modern history that mass firings have happened during a funding lapse,” Oliver observes.

The administration also created a new “CEO of the IRS” position to bypass Senate confirmation, appointing Frank Bisignano, former CEO of Fiserv, who still owns about $300 million in company stock. This creates obvious conflicts of interest, especially since Fiserv is involved in launching digital stablecoin initiatives. “This is why you have to have hearings. You can’t just appoint somebody to a position,” Leary emphasizes.

When Independence Becomes a Joke

Next, Oliver and Leary discussed how financial entanglements are destroying audit independence while regulators focus on trivial violations.

Take BDO’s current crisis as an example. The firm took a $1.3 billion loan at approximately 9% interest from Apollo Global Management to finance its employee stock ownership plan. The debt forced the company to lay off employees, freeze travel, and conduct emergency cost reviews across all divisions.

But while BDO was giving First Brands a clean audit opinion, Apollo was actively shorting the company. First Brands collapsed months after BDO’s clean audit. “If I’m BDO and I audit a company that is being shorted by a company I took a $1 billion loan from, where’s the independence?” Leary asks. “What is the fraud triangle? Opportunity, rationalization, and financial pressure. All the parts of the fraud triangle are here.”

Meanwhile, EY is celebrating a “dramatic audit quality turnaround,” with its deficiency rate dropping from 46% in 2022 to below 10% in 2025. They achieved this miracle by firing 132 public company audit clients. In other words, the problematic audits didn’t disappear. They just moved to Deloitte and KPMG. “Have we actually achieved anything here? Or have we just shifted the bad audits somewhere else?” Oliver wonders.

The hosts also discussed a new scheme where crypto promoters target CPA firm clients. The Truevestment Bitcoin Legacy Fund wants CPAs to help raise $150 million from their clients, which institutional investors will then match before merging into a Nasdaq entity—essentially a SPAC wrapped in Bitcoin speculation.

The marketing compares buying Bitcoin today to “buying the Dow at 900.” But as Leary points out, when the Dow was at 900 in the mid-1960s, it consisted of companies like AT&T and General Electric—”companies that made things” and created real value, not speculation.

Why Nobody Cares About Financial Reports Anymore

Perhaps the most damning revelation from the podcast’s recent news roundup is that 86% of major companies are using broken charts in their financial reports. A CPA Journal study found bar charts with misleading axes, pie slices that don’t match percentages, and deliberate distortions to exaggerate performance. Of 1,584 charts reviewed, 12% had fatal flaws that completely misrepresented the data.

“The fact that so many of them have errors and nobody’s pointing them out indicates to me that nobody’s reading them,” Oliver observes. Indeed, 10-K filings get downloaded an average of just a few dozen times.

The hosts even shared a bizarre example where social media bots criticizing Cracker Barrel’s new logo caused the stock price to tank. According to Wall Street Journal data, 44.5% of posts about the logo change were from bots. “Maybe nobody cares about your charts because nobody even cares about the financial statements,” Leary suggests.

What This Means for Your Firm

The key insight from Hector Garcia stuck with David: “AI is never going to do perfect accounting, but it’s going to do it good enough.” For most clients, “good enough” financials that they can generate themselves might be perfectly adequate.

Accounting professionals can embrace AI for meaningful fraud detection and insights, or watch clients realize they can generate “good enough” work themselves. As this episode of The Accounting Podcast makes clear, the traditional value proposition of professional accounting services is crumbling. The firms that survive will be those that identify and deliver human value that transcends what AI can do: strategic insight, ethical judgment, and genuine expertise that no algorithm can replicate.

Listen to this episode to understand not just the challenges facing accounting, but what you need to do differently starting today.

The Accounting Platform That Achieves 96.5% Automation Reveals How They Did It

· ·

“No one’s going to be outcompeted by the AI itself. You are going to be outcompeted by firms that really adopt this aggressively,” warns Jeff Seibert, whose company just hit 96.5% accuracy in automated bookkeeping—something that seemed impossible just a few years ago.

In this milestone 100th episode of the Earmark Podcast, Blake Oliver sits down with Jeff Seibert, co-founder and CEO of Digits, to explore how AI is fundamentally changing the architecture of accounting software. Seibert brings fresh eyes to accounting—he previously led consumer product at Twitter and built Crashlytics (now running on six billion smartphones). His frustration was simple: Why could product teams access real-time analytics while business owners waited weeks for black-and-white spreadsheets?

Founded in 2018, Digits set out to reimagine accounting in the age of machine learning. While traditional software treats transactions as meaningless text in rigid databases, Digits achieves near-perfect automation by treating financial data as interconnected objects that learn from patterns across millions of transactions.

The 30-Year-Old Problem Holding Back Accounting

As Seibert sees it, the fundamental issue facing bookkeeping automation is that every major accounting platform—QuickBooks, Xero, and even NetSuite—runs on relational databases designed 20-30 years ago. These systems treat transactions as simple text entries with no understanding of what they mean.

“QuickBooks is just going to see an Uber transaction as “U-b-e-r”. It just sees text,” Seibert explains. “It doesn’t understand the data, it doesn’t know what Uber actually is.”

This limitation explains why Intuit, with all its resources, has yet to deliver meaningful automation. The answer is architectural. Each QuickBooks company exists in its own isolated database, preventing the software from learning patterns across businesses. The constraints run so deep that QuickBooks still can’t handle having a vendor and customer with the same name—it appears they chose “name” as the primary database key decades ago.

Digits takes a completely different approach using what’s called a vector graph data model. Everything becomes an object—Uber is an object, your expense categories are objects, your bank accounts are objects. Transactions become connections between these objects, creating a web of financial relationships the AI can understand.

This mirrors how large language models (LLMs) work, converting transactions into vector embeddings, essentially plotting them in multi-dimensional space where similar items cluster together. When trained on 170 million transactions representing nearly $1 trillion in business activity, patterns emerge that would be obvious to humans but invisible to traditional software.

“When you have that scale of data and you see how everyone has booked Uber before, you start to see patterns,” Seibert notes. “The model starts learning. If it sees Lyft in your accounting for this client, it then knows how to book Uber.”

How AI Agents Actually Work (Hint: Like Clever Interns)

The accounting world is buzzing about “AI agents,” but what are they really? Seibert explains, “An agent is simply an LLM that you run in a loop. You give it a task, it attempts the task, you ask if it completed it. If not, it continues until it’s done.”

Think of them as clever interns who never get tired. Digits has been running these agents in production since January 2024, primarily for researching unfamiliar transactions.

The system uses three layers of intelligence. First, it checks if this specific client has seen this transaction before. If yes, it books the transaction exactly the same way. Second, if the transaction is new to this client but familiar to the platform, it uses its global model trained across all users. Third, for completely novel transactions, the agent literally Googles them.

“What would you do as an accountant? You would probably Google it,” Seibert explains. “What do our agents do? They literally Google it, research the transaction, build a dossier about it.”

As a result, only 4-5% of transactions now require human review, compared to the 20% that typically slip through even well-maintained rule-based systems. Notably, the system maintains strict confidence thresholds. Any transaction it is unsure about gets flagged for human review. It never guesses when uncertain.

The upcoming reconciliation feature shows how sophisticated these agents have become. The system pulls statements directly from banks or extracts them from PDFs, then matches transactions with pixel-level precision. “You can literally click on the transaction and see it on the statement and vice versa,” Seibert says. This builds trust with accountants who need to see exactly where the numbers come from.

What This Means for Your Firm’s Future

As of August, Digits hit 96.5% accuracy, up from 93.5% in spring. Each percentage point represents thousands of transactions that no longer need human touch. But it begs the question: how do you price services when the work happens automatically?

“If you’re charging purely per hour right now, then automation may make that challenging,” Seibert acknowledges. But forward-thinking firms are already adapting. They’re moving to fixed-fee models for routine work like monthly closes, which become increasingly profitable as automation reduces time investment. Many use a hybrid approach, charging fixed fees for the close, and hourly rates for advisory work.

At a flat $100 per month (with volume discounts for accounting partners), Digits offers predictable pricing that contrasts sharply with QuickBooks’ constant increases. The platform even offers specialized SKUs for ledger-only or reporting-only clients, accommodating diverse practice needs.

The staffing implications are real but not apocalyptic. Junior bookkeeping roles focused on data entry will diminish. But Seibert points out this could make the profession more attractive: “You don’t want to just sit there doing data entry all day long. You want to learn how to advise businesses.”

Seibert recommends firms start small when implementing automated bookkeeping. “Pick one client in your firm and see what you can achieve,” Seibert challenges. Choose a simple, digital-native business like consultants, SaaS companies, or agencies with predictable electronic expenses. Build confidence, then expand to complex cases.

Building Trust Through Transparency

With financial data flowing through AI systems, security is crucial. Digits addresses this with architecture developed at Seibert’s previous companies, where they handled crash data from billions of smartphones.

Everything stays within Digits’ systems; they don’t send raw data to OpenAI or other third parties. All data is encrypted at rest using per-object envelope encryption, where each object has its own encryption key. Even if breached, stealing one key wouldn’t compromise the system.

The platform is SOC 2 Type 2 certified, with complete audit trails showing who changed what and when. You can even grant granular access, like giving your marketing manager visibility into only marketing expenses. “They can see marketing, all the transactions booked to marketing, and nothing else,” Seibert explains.

Importantly, when AI does the work, you can trace exactly what happened. Click on any transaction to see the activity log. This solves the common problem of clients making changes in QuickBooks without anyone knowing.

The Competitive Reality Check

Seibert’s warning deserves repeating: “No one’s going to be outcompeted by the AI itself. You are going to be outcompeted by firms that really adopt this aggressively.”

This isn’t hypothetical. Firms using advanced automation already serve more clients with similar-size teams, offer competitive pricing while maintaining margins, and provide real-time insights that clients increasingly expect.

You don’t have to become a tech expert. Set aside time each month after the close to try new tools. Watch YouTube videos about AI agents (though Oliver warns to avoid the hype channels). Most importantly, maintain healthy skepticism. As Seibert notes about AI doing math, “If it’s not 100% correct, what’s the point?”

Remember, AI agents are like clever interns. They’re eager, overconfident, and need supervision. They excel at tedious, repetitive tasks but need human judgment for nuanced decisions. The goal isn’t to replace accountants but to eliminate the work accountants wish they didn’t have to do.

Taking the First Step

Thoughtfully evaluate how these innovations can augment your practice. Start with one simple client. See what 96.5% automation actually feels like. Build confidence, then expand gradually.

Listen to the full episode to hear Seibert’s complete vision and practical guidance on everything from selecting pilot clients to restructuring pricing models. The tools to eliminate tedium while amplifying expertise aren’t coming; they’re here, proven, and improving rapidly. How quickly and thoughtfully can you integrate it?