It's 9pm on a Tuesday. You're still in QuickBooks.

Not because the work is hard. Because your client finally answered your questions at 6pm — and now you're racing to enter three weeks of categorizations, missing statements, and clarifications before close.

This is the actual rhythm of modern accounting. Long stretches of waiting. Then a flood. Then a late night.

The close isn't slow because closing is hard. It's slow because everything that should have been answered weeks ago lands in your inbox at the worst possible moment.

On Tuesday, join me for a webinar with Andrew Robinson of Puzzle to see how AI is starting to break this cycle.

We'll cover:

• AI that emails clients automatically the moment a transaction needs clarification — no more end-of-month scramble

• Live AI chat: cash projections, headcount planning, variance analysis, tax estimates — answered in seconds, not days

• Bank rec from a PDF statement (great for credit unions and treasury accounts the feeds keep mangling)

• Payroll JEs, settlement reports, and screenshots turned into clean entries

• The shift from batch close to continuous data entry — and what that means for how firms staff and price

📅 Tuesday, June 2 · 10:30am PT / 1:30pm ET

🎓 Free NASBA-approved CPE for attendees via Earmark

Register: https://streamyard.com/watch/GTHfFJ44vX4H

The SEC and FASB are finally asking.

Officials from both bodies recently confirmed they’re preparing for optional semi-annual reporting.

Companies would file twice a year instead of four times, opting in by checking a box on their 10-K. A new form, the 10-S, would replace the quarterly 10-Q for companies that choose it.

It's voluntary. Companies that prefer quarterly reporting keep it. That means the market itself will signal what investors actually need.

David Leary and I covered the news on Episode 487 of The Accounting Podcast.

Here's how I see the trade-offs.

The Case For Semi-Annual Reporting

It ends the brutal quarterly reporting cycle. Quarterly reporting means four internal closes, four audit cycles, four rounds of disclosure review, and four earnings calls every year. Cut that in half, and you free up real capacity for finance teams and audit firms.

It relieves pressure on an already thin talent pool. Accounting has a well-documented talent shortage. If we want to attract and retain great people, we have to stop burning them out on a frantic quarterly reporting cycle. Semi-annual reporting supports retention and makes accounting look like a desirable career to someone weighing their options.

It could reduce short-termism. David made the point on the podcast that quarterly reporting creates quarterly pressure. Most business models can’t pivot in 12 weeks. The pressure to show quarterly progress incentivizes earnings management over actual business building. Semi-annual reporting loosens the ratchet.

The Counterargument

Investors get information less frequently. That's the core objection, and it's legitimate. For a company going through a big strategic shift or an operational crisis, six months is a long time.

The adverse selection problem. If companies with good news to share stick with quarterly reporting and companies with bad news opt into semi-annual, the 10-S could become a signal, and not a reassuring one.

Voluntary doesn't mean simple. Two different reporting cadences create comparability problems for analysts covering sectors where some companies report quarterly, and some don’t.

The Bigger Question

Beyond the logistics, the debate over quarterly versus semi-annual reporting assumes that what we're currently reporting is useful. I'm not sure that's entirely true.

The SEC and FASB plan to address how often companies file.

We also need to ask what we’re capturing.

In the 70s, tangible assets like factories and inventory told nearly the entire story. Today, it’s the opposite.

Intangible assets like brands, customer relationships, proprietary technology, workforce capabilities, and data now account for roughly 92% of market value for S&P 500 companies.

But under U.S. GAAP, we expense these drivers of value or ignore them entirely.

So when regulators talk about not disclosing what doesn’t matter to investors, I have to ask: Do earnings reports tell investors what drives value creation in a knowledge economy?

Often, the honest answer is no.

This Is Only a Start

I support the shift to semi-annual reporting. It’s a necessary relief valve for a profession under immense pressure.

But this is a minor reform.

We need to think critically about what belongs on a balance sheet in 2026.

What would it take to make financial statements genuinely useful again?

Listen to the full discussion on The Accounting Podcast.

It signals health. It drives valuations.

Most accounting firms have no idea what theirs is.

We’re too busy staring at billable hours, utilization rates, and realization metrics to notice we’re running subscription businesses.

Clients come back every year for tax returns, audits, and bookkeeping. That's recurring revenue. That’s a subscription, even if we don’t sell it that way.

But we ignore the one metric that actually tells you how your subscription business is performing.

Standard-Setters Are Missing the Point

I've been thinking about this more since this Bloomberg story landed in my inbox. AI startups use ARR as their primary growth metric, but it’s slippery. CEOs can manipulate it to impress investors.

The problem is there’s no accounting definition for ARR.

Neither FASB nor the SEC has defined it. It doesn't exist in GAAP. Companies make it up as they go.

That's a transparency problem in the tech world. But it’s even worse in accounting: we ignore the metric entirely.

The Subscription Economy Caught Up to Accounting

The shift to fixed-fee pricing has been building for years.

CAS practices were the early adopters because hourly billing doesn't work when you've automated 80% of the data entry that you used to bill for.

So firms moved to fixed monthly fees. But they never adopted the mindset of a subscription business.

Think about what ARR tells you:

• The total value of your returning clients

• The value you're retaining year-over-year

• Which clients are powering growth, and which are one-time distractions

None of that is visible on a utilization dashboard.

Where ARR Gets Slippery

The Bloomberg story gained traction because ARR is easy to manipulate.

Dishonest founders stretch the truth by annualizing monthly revenue from customers who won’t stick around.

If a client signs up for a trial and you multiply their monthly fee by 12, you count revenue that may never materialize.

Blending all of that into a single ARR number without accounting for churn (the percentage of customers you lose over the period) gives a misleading picture.

In Episode 484 of The Accounting Podcast, my co-host David Leary pointed out that B2B software contracts are getting shorter. Sub-one-year deals grew 4% in 2023 and 13% in 2026. "Recurring" is getting murkier.

The Challenge for FASB

FASB has spent decades refining standards for obscure derivative transactions. We have hundreds of pages on revenue recognition, lease accounting, and financial instruments.

But there’s zero authoritative guidance on ARR, the metric driving hundreds of billions in investment decisions.

If I were in charge of accounting standards, SaaS metrics would be the priority. Define ARR, Churn, and Net Revenue Retention (NRR). Give auditors something to test and investors a number they can trust.

We’re in the business of making financial information reliable. Right now, we’re failing the subscription economy.

Start Tracking ARR

We don’t have to wait for FASB to define ARR to start learning from it.

Build an ARR report for your own practice. List every client and their fixed monthly or annual fee. Sum it up.

Then track it monthly.

You'll see patterns your billable hours report never showed you. You’ll see which service lines are growing. You’ll see where you’re losing ground.

When FASB eventually gets around to defining churn and ARR, you’ll already understand what it means in practice because you use it. And you can help FASB define it.

You already have the data. You just haven't organized it this way yet.

When I shared NASBA's demand letter publicly last week, I said the profession deserves an honest conversation about how its institutions work.

That conversation just got bigger.

CFO.com picked up the story: NASBA warns CPE provider to stop 'unfavorable' comments.

The piece walks through what NASBA's letter said, what I actually argued at the AICPA Learning & Development Symposium, and the reply I sent back.

NASBA declined to comment on the substance to CFO.com, saying only that they handle these matters privately, not through public commentary. I’m still waiting for their reply.

If you want my full take (what I said, why I stand behind it, and why this matters to all educators in our profession), here's the original post with links to the demand letter and my reply.

The panel: "The AI-Native Firm: Unifying Tax, Advisory, and Assurance."

A few of the other questions I’m bringing to the table:

• How does AI change what we can do compared to 18 months ago?

• When AI handles compliance, what does the advisor's role become?

• What does a CPA need to unlearn to make that leap?

Joining me on stage:

• Jim Bourke — Withum

• Becky Munson, CPA, CITP — EisnerAmper

• Sean McLean, PMP — Elevate CPA Group

• Ashok Parmar — Deloitte

• Geni Whitehouse — Information Technology Alliance (ITA)

The summit is invitation-only—you can request yours at blackore.ai/summit.

If you're already on the list, see you there!

It was the choice you made when you valued lifestyle over scale. You could have a great work-life balance, but you’d never get rich.

That ceiling just turned into a floor.

Thanks to agentic AI, even the smallest firms can achieve the highest margins.

And now we have an example:

Sam Leon runs The Millennial CPA out of Richmond, Virginia. He has one office and zero employees. 70% of his budget goes to tech.

His firm was just featured in Accounting Today's 2026 Best Firms for Technology list, and it's a preview of the future.

We discussed this story (and more) in episode 486 of The Accounting Podcast.

The Grunt Work Is Gone

According to Sam, AI handles the preparation while he focuses on the review.

Tax workpapers that used to take a human 3 to 5 hours to build from financial statements now take AI 5 minutes. He can finish up year-over-year comparisons (which once ate up an entire afternoon) before his coffee gets cold.

He still applies professional judgment. But he isn’t spending hours on admin, document collection, data entry, and workpaper preparation.

His hiring strategy is simple. Don’t hire until the AI hits a wall.

Margin vs. Price

Larger firms should pay attention. Sam isn't competing with you on price. He's competing with you on margin.

He takes home a higher percentage of every dollar he bills because his cost structure looks nothing like yours.

Traditional firms have partners averaging 50+ hours a week to keep the lights on.

Solo practitioners using this model reclaim their time.

The largest firms face a talent retention problem driven by a model that demands people grind through years of low-paid work on the promise of a partnership that fewer than 1% will ever reach.

Sam has none of that.

The New Math of Going Solo

For accountants dreaming of going out on their own, this changes the math in a real way.

The traditional barrier to a solo practice was the operational load.

Doing everything yourself is exhausting. That's why most solo practitioners either burn out or hit a growth ceiling.

AI changes the definition of "doing it yourself."

Document organization, data extraction, and workpaper creation can be automated workflows.

You don't need a software engineer. You just need to invest in the right tech and verify the output. That's a skill any CPA can develop.

The New Competitive Threat

Your biggest competitor probably isn't a national player moving into your market.

It's the solo practitioner with low overhead, automated workflows, and the ability to maintain quality while operating at a fraction of your cost.

There will be more of them. The friction of running a one-person shop drops every year.

So, will large firms adapt before the margins compress?

Listen to the full discussion on Episode 486 of The Accounting Podcast.

The message, in plain English: stop criticizing us, or we'll pull your CPE sponsorship.

If a standard-setter can silence its sponsors by contract, every CPE provider in the country has a problem.

We discussed this in episode 485 of The Accounting Podcast.

Here's What I Said

In March, I spoke at the AICPA Learning and Development Symposium. I argued that the traditional model of CPE course development is backward.

I also pointed out that polling questions in live webinars mostly prove you have a working index finger, not that you learned anything.

Here's the heresy:

Today, instructors are advised to write a description and learning objectives, and to create a detailed outline before teaching the class.

I proposed flipping it. Let the expert teach first. Record it. Then have AI read the transcript and create the course from the actual knowledge shared.

In a 30-minute demo, I showed L&D professionals how to build a high-quality self-study course in the time it used to take to write the outline.

Expertise first, paperwork later.

What NASBA Calls “Unprofessional”

NASBA's director of compliance pointed to a clause in the sponsor agreement requiring sponsors to act in a manner that is "professional, appropriate," and "reflects favorably on NASBA."

On that basis, they're asking me to "immediately cease making any unfavorable, unprofessional, or inappropriate comments" about them — or lose Earmark's sponsorship.

NASBA hasn't defined any of those terms. In practice, that means NASBA gets to decide which speech from CPE sponsors is acceptable.

Why This Matters To All CPE Providers

The CPE system has real, well-known problems.

You can click "yes" on a polling question while answering emails and get credit. You can scan your badge at a conference, sleep through the session, and still earn hours.

Everyone in the profession knows this. The question is whether we're allowed to say it out loud.

A sponsor agreement should ensure quality, not enforce a speech code. By weaponizing these contracts to stifle debate, NASBA is choosing rigid compliance at exactly the moment the profession needs rapid progress.

I want NASBA to succeed at its mission. That's precisely why I'm willing to criticize how it's being carried out. Professional skepticism is a core value of the CPA profession. Sharing ideas to improve the process isn't unprofessional. It's the job.

My Response

I've sent NASBA a written reply. You can read it here.

I'm not looking for a fight. I want to help modernize CPE. But I'm not going to stop saying what I think.

If you're a CPE provider, a state board member, or a CPA who has noticed the same things I have, I'd like to hear from you.

The profession deserves an honest conversation about how its institutions work, and it can't have one if the institutions get to define which conversations are allowed.

That's the argument Richard Lynch, managing principal at Sikich, made when we sat down for episode 113 of the Earmark Podcast. And it's hard to poke holes in it.

The pyramid-shaped firm has worked for decades. Lots of staff at the bottom, fewer partners at the top. That model existed because junior roles did real work — and in doing that work, they learned. But AI is taking over the work. So if there's nothing left at the base, the pyramid becomes a diamond: fewer entry-level people, more in the middle, different expertise at every level.

That changes everything. How you hire. How you train. How you bill. How you measure whether someone is actually getting better at their job.

Richard and I dug into all of it on the pod. And now you can earn free CPE credit just for listening.

Register here: https://www.earmark.app/c/3402

Here's what else we covered:

• The "super accountant" and what it actually means: AI isn't replacing accountants. It's compressing the learning curve. Professionals who adapt will do in an hour what used to take a week. Those who don't may find the market has already moved on.

• Why the billable hour is on borrowed time: When AI does the work in minutes, billing by the hour stops making sense. We get into which metrics should replace it and why timesheets persist even when everyone knows they're broken.

• Training without doing: Junior accountants used to learn by doing the work. If AI handles the work, how do you develop the next generation? This is the question keeping me up at night. There's no easy answer.

• What the firm of 2030 actually looks like: Richard maps out who gets squeezed out in the transition, who thrives, and how leaders can get ahead of it rather than react to it.

• Progression or extinction: Firms that wait to see how AI plays out may not get a second chance. Lynch says the time to act is now and shares what that looks like in practice at a firm like Sikich.

How to earn free CPE

Listen to episode 113 of the Earmark Podcast, take a short quiz, and get your CPE certificate.

Register here: https://www.earmark.app/c/3402

A developer on Reddit built a full double-entry accounting system using Claude and an MPC server. No QuickBooks subscription or Xero license. Just AI, a .NET console app, and a local SQLite database.

He calls it Tidwell, and it does more than you’d expect.

Drop in a photo of a receipt, and Claude categorizes it, picks the right expense account, and posts the entry. The database enforces accounting rules. Debits must equal credits.

The system supports a chart of accounts, bank reconciliations, and financial reports, including a P&L, balance sheet, trial balance, and general ledger.

It can even handle QuickBooks imports.

He built it because he didn’t want to pay for QBO and figured Claude could handle the interface layer.

Turns out it works.

The End of the Standardized Tech Stack

Right now, you might be thinking, “Some developer’s weekend experiment doesn’t threaten Intuit’s $17 billion market cap.”

And you’re probably right, at least for now.

But that’s not what caught my attention when David Leary and I discussed this on episode 483 of The Accounting Podcast.

What caught my attention was the question it raises for the profession. If a self-taught developer can clone the core functionality of QuickBooks in a weekend, what does that say about the defensibility of accounting software?

Around the same time, another post surfaced, this one about a developer who fed an old version of QuickBooks Desktop into Claude Code and got back an open-source clone. A working one.

Meanwhile, Xero published a blog calling itself an “AI-native operating system” and listing roughly 20 buzzwords per paragraph.

My co-host David’s take is that the post was written for the stock market, not for customers.

Xero, Intuit, Sage, and Oracle are all near 52-week lows right now. Investors are nervous that AI will seriously disrupt the market for accounting software.

They’re right to be nervous.

What the Accounting Software Moat Looks Like

The real moat for QuickBooks and Xero is the ecosystem: the integrations, accountant relationships, and institutional inertia.

Xero spent over a decade barely making a dent in QuickBooks’ market share, even though it had a genuinely better product for a long time. Switching costs in accounting software are brutal.

But AI changes the equation for switching costs. If an AI agent can sit on top of whatever system a client is already using, then the underlying platform matters less. The interface layer is the AI.

That’s exactly the bet a company called Artifact is making.

It launched a platform called Omni that functions something like a Zapier for accounting firms. But instead of connecting apps with rules, it deploys AI agents to operate your existing tech stack.

You describe your workflows in plain English. The agents build and execute them. You don’t have to rip out your existing software. The AI just learns to use it in the way a human would.

It’s an interesting idea. Whether it works at scale is another question.

The Problem This Creates for Accounting Firms

There’s a catch, though, and David put it well: accounting firms spent the last 15 years building standardized tech stacks: “We’re a Xero shop. We use Dext for receipts. We use Melio for bill pay.”

Standardization is how firms scale. AI threatens that control.

Clients are building their own accounting systems. Some of them actually work.

So when a client shows up with a homemade Claude-powered bookkeeping setup, you, as their accountant, have to figure out how to work with it.

That’s already happening.

We’ve seen this movie before. Every major technology wave in accounting, from cloud software to bank feeds to OCR receipt scanning, promised to reduce the work.

Instead, it created more of it. More mess. More reconciliation. More clean-up work.

AI will likely do the same thing. The clients who build their own AI-powered accounting systems will make a mess. We’ll be the ones untangling it.

What Needs to Happen

The Tidwell project is impressive as a proof of concept, but it’s not going to replace QuickBooks for most small businesses. The reasons go beyond technical capacity.

There’s no audit trail that an outside reviewer can independently verify. There’s no integration with payroll, sales tax, or dozens of other systems a business actually needs to run.

There’s no support when something breaks.

But it shows the interface layer of accounting software is largely automatable. AI can handle the conversation. The rules and the database can enforce the accounting.

The firms and platforms that recognize this early will build systems where AI handles routine work while humans focus on review, judgment, and client relationships.

That model makes sense. AI can handle the parts of accounting that shouldn’t require an accountant’s judgment in the first place.

A Question Worth Taking Seriously

Xero’s blog post claims it’s transforming from a “system of record” into a “system of action.”

Maybe.

But if AI agents can operate a system of record on behalf of the user, the underlying system of record becomes a commodity.

That’s the question accounting software companies should be losing sleep over.

For accounting firms, when your clients start showing up with DIY AI-powered bookkeeping systems, do you have a plan for working with them?

Or will you be the one sending a QBO invite to a client who’s already moved on?

Chelsea Summers of Inside Public Accounting put it plainly: if AI makes your work faster and you're billing by the hour, what happens to your fees? You'd have to cut them. And no firm is going to do that voluntarily.

That's the tension the IPA survey data keeps surfacing. All the talk about AI transforming the profession — and the pricing model that punishes efficiency hasn't budged.

Chelsea and I talked about this and a lot more on Episode 112 of the Earmark Podcast. And now you can earn free CPE credit for listening.

Register here: https://earmark.app/c/3370

Here's what else we explored:

• Offshoring is paying off in the numbers: Over half of the largest accounting firms now use offshore staff, and the data backs it up. Firms with offshore teams had 8.1% organic growth versus 7.5% for those without, and 9% better margins. The model has shifted too: offshore staff are no longer processing in centers. The work has moved up the value chain, from data entry to manager-level work and even client-facing advisory.

• Top firms don't grind harder: The IPA names its "best of the best" firms based on 30 operational metrics. Those firms have revenue per FTE of $272,000 versus $220,000 for everyone else. But their utilization rates and hours worked are nearly identical to average firms. The difference is leverage — 17.7 staff per partner versus 11.8 — and pricing confidence. Partner bill rates at top firms run $588 versus $448 at the average.

• Firms are underpricing, and they know it: When firms raise prices aggressively, clients stay. Chelsea's heard it over and over: raise rates 40%, and they all stay. The market can absorb it. Firms that aren't raising prices every year are leaving money on the table and may be signaling a lack of confidence in their own work.

• AI ROI is real, but narrow for now: The clearest returns are in workflow automation, document processing, and tax research. The blockers aren't technical. They're partner skepticism and change management. Firms that tabled AI conversations to focus on busy season are now months behind, in a space where months matter.

How to earn free CPE

Listen to Episode 112, take a short quiz, and get your certificate. It's free.

Register here:https://earmark.app/c/3370

In the latest episode of the Earmark Podcast, I sat down with Richard Lynch, managing principal at Sikich, one of the largest CPA firms in the US. Richard's been thinking seriously about what AI means not just for how firms operate, but for the entire structure of the profession.

We covered a lot of ground in this conversation:

• The rise of the "super accountant" — a new kind of professional who uses AI to skip the grind and get to high-value advisory work faster

• Why the pyramid firm model is shifting to a diamond, and what that means for hiring, training, and career development

• The stubborn grip of the billable hour, and why even firm leaders who hate timesheets can't seem to let them go

• What firms must change now — in billing, performance metrics, and culture — to land on the side of progression rather than extinction

Richard says the question isn't whether change is coming. It's whether your firm will adapt fast enough to survive it.

If you lead a firm, manage a team, or are early in your accounting career, this episode is worth your time.

Listen in: https://podcast.earmarkcpe.com/113

Not hacking. Not a one-off forgery. A business model built around fabricating the compliance reports that companies use to prove their data security to vendors and customers.

SOC 2 reports are trust anchors. Companies share them with prospects and partners to demonstrate their controls are real and their security practices hold up to scrutiny. If AI can generate convincing SOC 2 documentation cheaply and at scale, what does a real one actually mean anymore?

David and I dug into this and a lot more on episode 480 of The Accounting Podcast. And now you can earn free CPE credit just for listening.

Register here: https://earmark.app/c/3348

Here's what else we got into:

• How taxpayers are actually using AI for tax prep: The data might surprise you — and not in a good way.

• AI in finance workflows: Tools like Claude are quietly showing up in accounting and finance departments. We talk about what that means for practitioners.

• FAA staffing risks: This one seems off-topic until it isn't. Worth watching if you care about systemic risk in regulated industries.

• Threats to state accountancy boards: What's being proposed, what's at stake, and why CPAs should be paying attention right now.

How to earn free CPE

The episode is live on Earmark as a free CPE course. Listen, pass a short quiz, and get your credit.

Register here: earmark.app/c/3348

Accounting's billing model was designed for factories.

Not metaphorically. Literally.

In the early 1900s, accounting firms took cost accounting theory — built to manage factories and railroads — and applied it to themselves. Their logic: our people are machines that produce hours, and we sell those hours to clients.

That made sense for a 20th-century factory. It makes zero sense for a knowledge business.

I learned this firsthand when I was a bookkeeper at the start of my accounting career. I automated 80% of my data entry work using cloud tools. If I'd stayed on an hourly billing model, I would have lost 80% of my revenue overnight. So I switched to fixed fees immediately. My clients got more service. I did less work. Better margins for everyone.

That's the trap of hourly billing: the more efficient you get, the more you get punished. The most productive accountants aren't the ones logging the most hours. They're the ones who've decoupled revenue from time.

The big firms haven't cracked this. Their practice management systems are built around timesheets. Even when they attempt fixed fees, it's a hack layered on top of an hourly infrastructure. And their staff have no incentive to innovate, because efficiency just means more clients, not more money.

AI is about to make this worse.

AI can make a good accountant 10x more productive. One person can now do work that used to take a team. How do you reconcile that with billing by the hour? You can't. The entire relationship between hours and revenue will be completely destroyed.

The answer is to think like a SaaS company. Stop tracking utilization rates and realization percentages. Start tracking:

• Annual recurring revenue (ARR)

• Customer lifetime value (LTV)

• Churn rate

• Cost to acquire a customer (CAC)

These metrics tell you whether your firm is actually healthy. The old metrics just tell you how many hours people sat at their desks.

I went deep on all of this with Alex Hoffman on Syft Analytics’ Beyond Insights podcast. We covered the factory mindset, AI agents, how to productize your services, why niching is one of the best moves a firm can make right now, and why I call myself a “lazy CPA.”

Watch the full episode here: https://youtu.be/afNcI70FOPw

Zero.

That’s from a National Bureau of Economic Research study of nearly 6,000 executives across four countries. And yet those same executives expect AI to boost productivity by 1.4% over the next three years.

Economists are calling it Solow’s paradox: the same gap between investment and results that defined the early computer age. You could see computers everywhere but in the productivity statistics.

The standard advice is to be patient. Transformative technologies take time to show up in the data. I get it. But I also agree with Alexander D. Hilton, the author of this CFO.com opinion piece: most firms are thinking about AI the wrong way.

The Cost Accounting Trap

The typical approach to AI focuses on the time a task takes today. Multiply it by the loaded labor cost. Calculate how much faster AI can do it. That factors into the return on investment.

That’s cost accounting logic. And, as Hilton points out, it’s the same logic Eliyahu Goldratt dismantled decades ago with his Theory of Constraints. Manufacturers ran every machine at full utilization because the local math said to.

The result was excess inventory, longer lead times, and hidden costs that never showed up in the numbers. The machines were busy. The system was slow.

The same mistake is playing out with AI right now.

When Klarna replaced 700 customer service agents with a chatbot in 2024, the cost accounting looked great. Until customer satisfaction collapsed and they started rehiring humans.

Making something faster isn’t the same as making the system produce more.

The Right Question: Where’s the Bottleneck?

Goldratt put it plainly: technology only brings benefits if it removes a limitation. The question isn’t whether AI is capable. It’s whether it removes the specific constraint that’s limiting your firm’s output.

That’s throughput accounting. Instead of asking “how much did we save?”, you ask, “how much more can we generate?”

Throughput is the rate at which the organization produces revenue. If an AI initiative speeds up a step that isn’t the bottleneck, throughput doesn’t move.

The CFO.com piece cites a real example: an AI tool accelerated document generation by a factor of 240. Cost accounting would call that transformational. But the bottleneck in the system was subject-matter expert validation. That’s a manual, human-dependent review step that couldn’t be parallelized.

The AI made the non-bottleneck faster. Nothing else changed.

A Firm-Level Reality Check

Right now, AI vendors are pitching tools that can prepare tax returns.

Perplexity just launched a product that drafts full federal tax returns from uploaded documents. TaxGPT says it automated 1040 prep. Basis raised serious money to build AI agents for accountants.

On Episode 482 of The Accounting Podcast, David Leary and I talked through what all this actually means for practitioners.

If your ROI logic is, “AI prepares a return 40% faster, so I save X hours, so the ROI is Y,” you’re probably measuring the wrong thing.

Is tax prep the bottleneck in your firm? In my experience, it isn’t.

The constraints are getting the right documents from clients and the time required for partner review. If AI speeds up data entry and form preparation, but you still have to chase down client documents, and the partner review queue still moves at the same pace, what’s changed? Your junior employees are less busy. But your throughput is identical.

Before you commit to any AI tool, push for honest answers to three questions:

1. What specific bottleneck does this address? If no one can name the constraint limiting firm revenue, you’re probably optimizing a non-bottleneck.

2. Does this increase throughput, or just reduce local cost? A process that runs faster but feeds into the same downstream bottleneck hasn’t changed your output.

3. What new costs does this create? Retraining, integration, validation workflows, or human oversight? If the net effect on throughput after subtracting that overhead is negative, the tool destroys value regardless of how good the demo looked.

Are You Doing the Right Math?

AI can be genuinely transformational for an accounting firm when it’s applied to the actual constraint. I’m seeing it happen.

But the firms that come out ahead will ask the harder question first: where is our constraint, and does this tool actually remove it?

The math for AI works out. You just have to make sure you’re using the right accounting method.

If you want to hear more of our discussion on this topic, check out Episode 482 of The Accounting Podcast.

A compliance startup seemingly sold hundreds of companies fake SOC 2 reports.

If the allegations hold up, it’s a brazen example of assurance fraud.

And it raises serious questions about how we’ve built the trust infrastructure around these certifications.

The story comes from a detailed Substack post by an anonymous author called “DeepDelver.” They allege Delve, a VC-backed compliance automation startup, generated the appearance of SOC 2 compliance without the underlying substance.

I haven’t independently verified these allegations. But they’re specific enough to take seriously.

What the Allegations Say

According to DeepDelver, Delve’s platform is built around pre-populated policies, templated risk assessments, canned security simulations, and pre-written board minutes.

And it’s all presented as if clients actually did the work. But the controls describe activities that never occurred.

Board meetings didn’t happen. Security simulations weren’t performed. Trust pages show controls as implemented before any work was done.

The author claims to have analyzed 322 public Delve trust pages. Of those, 321 showed identical SOC 2 control sets.

That’s tough to reconcile with Delve’s marketing claim that it customizes programs for each client.

Even worse, the report alleges draft report sections containing auditor conclusions were pre-generated before clients completed their compliance work.

If true, that’s fabricating auditor judgment.

How Does Something Like This Get Signed Off?

SOC 2 reports are AICPA attestation engagements. A licensed CPA firm accredited by the AICPA has to conduct them. The SOC 2 badge on a company’s website represents an independent professional opinion.

It’s not just a marketing asset.

So how do you allegedly game that system? You find firms willing to rubber-stamp work they didn’t do.

We’ve seen this before.

The SEC eventually charged BF Borgers CPA PC with fabricating audit documentation in more than 1,500 SEC filings. The PCAOB flagged deficiencies in 29 out of 30 audits it inspected.

But for years, the consequences were minimal. Inadequate penalties create inadequate deterrence.

When the downside of getting caught is smaller than the upside of cutting corners, corners will be cut.

Who’s Watching?

The harder question here is regulatory. Who’s actually responsible for investigating this?

The AICPA controls accreditation. State boards govern individual CPAs. The SEC gets involved when it affects public companies.

But for private companies relying on these certifications to win enterprise contracts, it’s less clear who’s checking.

Is the AICPA auditing the badges on company websites? Are the accredited firms being monitored for the volume and uniformity of the reports they sign off on?

These are questions worth asking.

The Real Cost

Companies displaying fake SOC 2 reports deceive their customers and create real security risks. A SOC 2 certification is supposed to tell enterprise buyers a vendor implemented meaningful controls around data security, availability, and confidentiality.

If those controls don’t exist, the certification provides false cover for real vulnerabilities.

And frankly, it also undermines legitimate compliance work. The firms that invest in actual controls and pay for genuine audits are competing against vendors that allegedly bought a badge.

As a CPA, it’s painful to watch. We spend decades building the “assurance” brand, only to have it diluted by a service sold like a software product that generates board minutes out of thin air.

If these allegations prove out, this won’t be the last story like it. The demand for compliance certifications isn’t going away. The market for fast and cheap alternatives will keep growing.

Oversight hasn’t kept pace.

Since we recorded the episode, Delve has fired back. They posted a response calling the allegations misleading and disputing the claims.

I’m not here to play judge and jury. But whether these specific allegations are 100% accurate or not doesn’t change the underlying problem.

The SOC 2 market is built on a foundation of “trust me.” And right now, that foundation looks pretty shaky.

If the AICPA doesn’t police the use of its own logo, the market will continue to treat SOC 2 as a commodity, or worse, a scam.

It’s time for the profession to decide: are we auditors or are we just helping startups check a box?

Business and finance (i.e., accounting) sit at the outer edge of the chart.

David Leary and I talked about it on episode 479 of The Accounting Podcast, and it sparked an interesting conversation because the same week we dug into that chart, two of the biggest firms in the world made opposite bets on what it means.

The Big Four are Split on the Future of Your Job

PwC’s U.S. CEO told The Guardian that partners and staff who haven’t fully adopted AI risk will be pushed out of the firm. They cut 5,600 employees last year and shifted hiring towards data specialists and engineers.

They’re building AI-powered tools and charging clients a subscription fee to access them, no PwC professional required. That’s what they’re calling the future.

EY, meanwhile, just doubled its CPA exam pass bonus to $10,000.

One firm is betting on machines. The other is betting on people.

As we discussed on the pod, the problem with PwC’s model is that when you sell an AI tool with your logo on it, you’re competing against every other AI tool, including the one the client already has on their laptop.

The value of the Big Four brand was never “we have software.” It was “we stand behind the answer.”

Is human judgment really that easily replaceable?

Automating it away doesn’t scale the business model. It undermines the reason clients pay a premium.

EY doubling down on the CPA is a bet that credentialed humans (those who can be held accountable and understand the work the AI is doing) will be worth more, not less.

The Subtext in the Anthropic Chart

The Anthropic chart actually tells a more nuanced story than the headlines suggest. Yes, business and finance has a near-ceiling theoretical exposure. But the observed exposure (what’s being automated today) is still low.

David has been asking on LinkedIn for weeks whether anyone can name one accountant they personally know who lost their job because a company implemented AI. Still at zero.

History shows that labor-saving technology doesn’t tend to reduce work. It raises standards and creates more work. Your clients won’t need fewer accountants. They’ll expect more from the ones they have.

Know Your Level, Then Skip One

I left public accounting as a manager. Give that version of me today’s tools, and I’m doing partner-level work inside of six months.

That’s not a brag. It’s what AI does to the experience curve. It compresses a decade of knowledge accumulation into a much shorter climb.

AI is going to blow up the staff-to-partner ratio. A big firm running a large client engagement today might assign 10 people to it. In a few years, that’s a team of three.

In small firms, one person will routinely do what used to take a team.

That changes how you should be thinking about your career right now.

If you’re a staff accountant, stop thinking like staff. Use AI to produce manager-level output: analysis, judgment calls, and client communication. Do it before the headcount reductions catch up with you.

If you’re a manager, the same logic applies. Push yourself to director or partner-level work.

If you’re a partner, you have two options. Use AI to develop your staff faster, or use it to run leaner.

Both work. Doing neither doesn’t.

84% of the PCAOB's actions and 98% of its penalties came before the previous chair (appointed by Biden) resigned in July. After that, things went quiet. The SEC only brought two actions all year.

And it's probably not bouncing back anytime soon. The PCAOB's 2026 budget cuts enforcement funding by 15%.

When enforcement drops this sharply, the effect isn't just statistical — it's behavioral. Auditors know they're being watched less closely, and that changes incentives at the margin.

Some firms may push back less on aggressive client accounting. Others may cut corners on procedures they're confident won't get scrutinized.

This isn't speculation — researchers who study regulatory behavior have documented this pattern repeatedly: enforcement levels signal the cost of non-compliance, and when that cost falls, compliance tends to decline alongside it.

The investor angle matters here too.

Audit enforcement exists because investors in public companies depend on audited financial statements to make decisions. When auditors face meaningful consequences for shoddy work, they have strong incentives to do it right. When they don't, the downstream risk lands on investors — usually long after anyone can do anything about it.

The PCAOB was created after Enron and WorldCom precisely because voluntary compliance wasn't cutting it. Gutting its enforcement budget while enforcement actions are already at historic lows sends a troubling signal about where priorities lie.

A 15% cut to enforcement funding doesn't happen in isolation. It's a policy choice about what the PCAOB is for.

If enforcement is being deprioritized, it raises the question of whether the agency will focus more on inspections, standard-setting, or something else entirely — or simply shrink.

For auditors, the practical implication is that the external check on quality is weakening. That puts more pressure on internal culture, firm leadership, and individual judgment to maintain standards.

Those who take the long view will keep doing quality work regardless. But history suggests not everyone will.

According to McKinsey, sixty percent of finance and accounting tasks could be automated with existing technology. So why doesn’t it feel that way?

Because “could be automated” and “can actually rely 100% on the output” are two very different things. And right now, there’s a big gap between them.

We’re stuck between two bad options. Full manual mode, which feels increasingly slow, or autonomous AI that just does stuff and won’t tell you why.

Neither one is actually usable in accounting, where we’re expected to work faster than ever, AND every number has to be defensible.

Sasha Orloff, co-founder and CEO of Puzzle, joined me on a recent livestream. He laid out a roadmap for approaching AI in accounting that works.

He calls the destination “governed automation.” But the analogy that got me was a parenting story.

“Go Clean Your Room”

Every parent has said it. And every parent knows the problem: you tell your child to go clean their room. Your child disappears for 20 minutes, and now they’re back, telling you it’s done.

But what’s the probability that it’s actually done?

It’s probabilistic. Maybe the floor is clear, but clothes are stuffed under the bed.

“Clean” can mean something entirely different to a child than it does to you. They answered the instruction. They just answered it according to their own interpretation.

That’s how most AI in accounting products works right now.

You give it a task, it does its best guess, and says it’s done. You have no idea what happened in between.

Governed Automation

The alternative version, the deterministic one, looks like this: before your child goes anywhere, they ask a few clarifying questions.

Do you mean just the floor? Or the whole room?

Do I have to fold the clothes or just put them away?

Once you agree on the definition of “clean,” they execute each step. Then they come back and show you: floor cleared, clothes folded, bed made. Done.

You can walk in and verify each condition. Yes or no, not “probably.”

That’s governed automation. And it’s the right design for accounting.

The Difference Between “Done” and “Actually Done”

There’s no “probably correct” in a general ledger.

Either the depreciation schedule is right, or it isn’t. Either the software categorized the transaction correctly, or it didn’t.

You can’t have a system running autonomous best-guess operations against books the client is going to stake business decisions on (and that you’re staking your reputation on).

As Sasha pointed out, you have to build a deterministic system on top of probabilistic technology. That’s the piece most accounting software companies miss.

Puzzle uses a large language model (LLM) to interpret what you want, then converts that intent into a series of concrete, auditable steps.

The AI drafts a plan, shows you the plan, asks if that’s what you meant, and waits for you to say yes before it touches anything.

Nothing happens without explicit permission.

You Still Have to Define “Clean”

Back to the room analogy for a second, because Sasha’s extension of it is great.

He pointed out that when the system doesn’t work, it usually fails for one of two reasons. Either:

1. The AI hasn’t been given the underlying functions to execute the steps (i.e., it doesn’t know what a “drawer” is), or

2. The user gave vague instructions.

“Go clean your room,” without specifying what clean means, is a recipe for disappointment. Same with “reconcile these accounts” without specifying your firm’s standards, your capitalization threshold, or your categorization rules.

The AI isn’t magic. It’s a very fast, very capable executor of instructions, but only instructions it understands.

By design, the accountant remains responsible for clearly defining those instructions.

You set the rules, review the output, and sign off.

The AI does the processing. Humans do the thinking.

You’re outsourcing labor, not judgment.

Trust Is Still the Product

The accounting profession has always been about trust.

Clients trust you with their numbers. AI makes that trust relationship more complicated, not less.

The answer isn’t to hand everything over to a machine. It’s to build systems that log every decision, trace every change, and ensure nothing happens without your explicit sign-off.

The best implementation of AI takes your prompt, asks questions to convert it into a deterministic checklist, and follows that checklist step by step with your approval at every stage. That’s how you get AI you can actually use as part of your close process. It’s a documented, auditable workflow you designed and signed off on.

That’s what I want to see in every AI tool accountants use. And it’s what buyers should demand when they evaluate new software.

“Go clean your room” is a terrible instruction.

“Clear the floor, fold the clothes, shirts in the shirt drawer, pants in the pant drawer, make the bed, and show me when its done” is a workflow.

Build the workflow.

If you want to see what this looks like in practice, watch the full livestream here.

And if you’re ready to try it yourself, Puzzle is opening the waitlist for AI Close now.

But when you look closer, there's still a long way to go.

Despite making up 44% of the CPA workforce, women account for only 26% of lead engagement partners in the S&P 500, according to a recent analysis in The CPA Journal, “Glass Ceiling Shattered?”

In other words, women are getting the title of partner but not the biggest clients.

The disparity is even more visible in particular sectors. For example, Energy, Real Estate, & Utilities audits have <20% female lead partners.

It’s not enough to promote women to partner if the most prestigious, high-fee engagements remain behind a legacy 'boys' club' wall. If we want true equity, we have to look at who is holding the keys to the S&P 500.

The song: “Baby, Where The Hell Is My Audit?”

Here’s the backstory:

In November, Massachusetts voters gave the state auditor authority to audit the legislature. It passed by a landslide. The legislature refused to cooperate. Diana DiZoglio sued them in February. The case is now heading to the state supreme court.

But instead of waiting quietly, DiZoglio belted this out in front of a crowd:

“Oh, baby, where the hell is my audit? Do they think we forgot it? ... I only fear he’s taking time to destroy all the records.”

We played the whole thing on The Accounting Podcast.

We also covered:

• Whether audit opinions should be probabilistic instead of pass/fail

• How AI agents turn vague requests into deterministic checklists (same thing good managers do)

• Why work won’t vanish due to AI — a continuation of the “Cowan Paradox”

• PwC’s warning to partners: embrace AI or get replaced

• What the BDO lawsuit means for auditors

• Whether tech CEO predictions about AI job losses are overblown

Watch the full episode.