By: John S. Morlu, II CPA
After years studying blockchain and AI, I discovered an unexpected business model: fixing what happens when powerful technology meets unwarranted confidence.
For years, accountants heard the warning:
“AI is coming for your job.”
Apparently, AI got lost on the way.
Because from where I sit, artificial intelligence may be creating an entirely new category of accounting work.
And my marketing manager and I have occasionally had to laugh about it.
Not because AI is bad.
Quite the opposite.
AI is extraordinary.
I have spent roughly eight years immersing myself in emerging technologies, including blockchain and, later, large language models. I became interested in what happens when technology moves from being an interesting experiment to something businesses actually depend on.
Now we are beginning to see the answer.
And sometimes the answer arrives carrying a 90-page report.
Welcome to the Age of the $7,000 Prompt
We have seen a new type of client problem emerge.
The details vary, and the examples here are deliberately generalized to protect confidentiality, but the pattern is increasingly familiar.
Someone has a financial problem.
Instead of starting with an accountant, auditor, tax professional, lender adviser, or other qualified expert, the person starts with AI.
That is perfectly reasonable.
Then something interesting happens.
The AI produces an answer.
The answer has headings.
It has tables.
It has bullet points.
It may even have an executive summary.
It looks magnificent.
There is only one small problem.
Nobody has established that it is correct.
That little detail can become expensive.
Case No. 1: The Financial Statements Looked Fantastic
Consider a generalized example.
A business owner needs financing.
The accounting records have problems. The balance sheet does not make sense. Accounts may be misclassified. Transactions need reconciliation. The financial statements need professional attention.
The owner has another idea:
AI.
Upload accounting reports.
Upload bank information.
Upload supporting records.
Ask an AI system to analyze everything.
Minutes later, beautiful financial statements appear.
The owner is delighted.
They look professional.
They look organized.
They look as if an entire finance department spent three weeks preparing them.
So the owner submits the information for financing.
Then questions come back.
Because lenders do not lend money based on font selection.
They care whether the numbers reconcile, whether assets and liabilities actually exist, whether income is supportable, whether cash flows make sense, and whether the financial information tells a credible economic story.
The SBA itself uses financial information to evaluate matters such as repayment ability and creditworthiness.
A beautiful balance sheet that does not balance economically is still a bad balance sheet.
Calibri cannot save you.
Case No. 2: ChatGPT, Meet the IRS
Another generalized situation involves a business facing an audit or tax problem.
The owner turns to ChatGPT.
This can actually be useful.
AI can explain terminology.
It can organize issues.
It can identify questions to ask.
It can summarize documents.
It can help a taxpayer become far better prepared for a conversation with a professional.
The problem begins when assistance becomes authority.
After enough prompting, almost any complicated problem can produce an impressive-looking strategy.
There may be numbered steps.
There may be references to regulations.
There may be a section titled:
“Recommended Strategic Approach”
That heading feels wonderful.
Unfortunately, the IRS does not become frightened when your chatbot produces bold text.
Eventually, somebody still has to determine:
- What actually happened?
- What do the records prove?
- What tax rule applies?
- What position is defensible?
- What documents support it?
- What should be communicated?
- And who is accountable for the professional judgment?
AI can help you prepare for the fight.
It cannot sit across the table and become your CPA.
Case No. 3: The 90-Page Masterpiece Nobody Understands
Then comes perhaps my favorite modern business phenomenon:
The AI Megareport.
A business owner is thinking about selling a company.
The owner uploads financial information into an AI system.
The machine goes to work.
Out comes approximately 90 pages of analysis.
Charts.
Ratios.
Forecasts.
Strategic observations.
Valuation language.
Risk analysis.
Maybe enough acronyms to qualify as a consulting firm.
Then the owner sends the report to a financial professional with one simple request:
“Can you explain this to me?”
And there, ladies and gentlemen, is the new economy.
We have reached the point where AI can generate analysis faster than humans can understand the analysis it generated.
That creates an entirely new professional problem:
Who interprets the machine?
AI Is Not Stupid. That Is Precisely the Point.
It would be easy—and wrong—to dismiss these systems as toys.
They are not.
Research has found that advanced AI models can perform extremely well on accounting examinations.
One published study found that ChatGPT-4 and Claude Opus passed CPA-style exam testing under the researchers’ methodology. Another study found major improvements as models became more advanced and were given better tools and prompting.
Research published in 2026 went even further: ChatGPT 5, Gemini 2.5, and Claude Sonnet 4.5 achieved averages above 96% across the assignments, exams, and projects in an undergraduate financial accounting course, although weaknesses remained on a complex, multi-step project.
So no, the serious argument is not:
“AI doesn’t know accounting.”
That argument is already obsolete.
The better argument is:
Knowing accounting answers is different from being accountable for accounting conclusions.
That distinction may define the next decade of professional services.
The Most Dangerous AI Error May Look Completely Professional
Traditional mistakes often announce themselves.
A spreadsheet formula returns an error.
A trial balance does not balance.
A document is obviously incomplete.
AI introduces a stranger problem.
The answer can be wrong beautifully.
OpenAI itself acknowledges that language models can hallucinate—producing plausible but false statements—and that hallucinations remain a challenge even as models improve.
Google gives similar warnings for Gemini. Google specifically tells users not to rely on Gemini responses as financial, legal, medical, or other professional advice and warns that responses can be inaccurate.
That is important.
Because human beings have a weakness for presentation.
Give us a poorly formatted correct answer and we become suspicious.
Give us a beautifully formatted wrong answer with six headings, three tables, and an executive summary and suddenly:
“This looks right.”
No.
It looks formatted.
Those are different professional standards.
This Is QuickBooks All Over Again
Accountants have seen this movie before.
Accounting software became easier.
People concluded that accounting itself had become easy.
It had not.
Software made recording transactions easier.
It did not eliminate the need to understand what the transactions meant.
That distinction has created enormous amounts of cleanup work for accounting firms.
A company can operate accounting software for years while quietly accumulating:
- Unreconciled accounts
- Duplicate transactions
- Incorrect classifications
- Unsupported balances
- Bad opening balances
- Incorrect equity
- Questionable receivables
- Mysterious liabilities
- And the famous account known throughout civilization as “Ask My Accountant.”
Then one day the company needs a loan.
Or an audit.
Or investors.
Or a sale.
Suddenly everybody discovers that clicking Save was not an accounting policy.
AI may produce the same phenomenon at much greater speed.
Welcome to AI Interpretation
That leads to an emerging opportunity for professional firms.
Call it AI Interpretation and Assurance.
Or, because accountants apparently cannot resist acronyms:
AIG — AI Interpretation Group
The proposition is simple.
Bring us the output.
We examine the underlying records.
We test the assumptions.
We reconcile the numbers.
We identify unsupported conclusions.
We distinguish useful AI analysis from confident nonsense.
Then we tell you what can actually be relied upon.
The joke around the office practically writes itself:
You pay AI to produce the answer.
You pay us to determine whether the answer is an answer.
And if it is wrong?
You pay us again to fix it.
Artificial intelligence has finally discovered recurring revenue.
But There Is a Serious Business Model Here
Behind the joke is a major change in professional services.
For decades, accountants largely transformed raw information into analysis.
AI can now perform part of that transformation astonishingly quickly.
The next high-value service may therefore move one level higher:
analysis → verification → interpretation → judgment → action
That changes what clients pay professionals for.
Not typing.
Not formatting.
Not basic calculations.
Not producing another 40-page report nobody requested.
Clients increasingly need someone who can answer five harder questions:
- Is this information reliable?
- What did the AI misunderstand?
- What important facts are missing?
- What professional standards or rules actually apply?
- What should management do next?
That is a much more valuable conversation.
The Professional Moat Is Moving
The accounting profession should pay attention.
The competitive advantage of the future will not be:
“I can calculate faster than AI.”
Good luck with that strategy.
Neither will it be:
“I can read 4,000 pages faster than AI.”
Enjoy retirement.
The durable advantages are more likely to include professional judgment, skepticism, domain expertise, verification, context, ethical responsibility, communication, and accountability.
Recent research comparing AI systems with advanced accounting students found an especially interesting weakness: models performed well on journal entries and basic calculations but struggled more when they had to critically evaluate solutions, detect errors, and distinguish correct information from incorrect information.
That is exactly where professional skepticism becomes valuable.
AI generates possibilities. Professionals determine what survives scrutiny.
The Winning CPA Will Use AI, Not Fight It
None of this means accounting firms should resist AI.
That would be like accountants in 1985 declaring war on spreadsheets.
The better model is:
AI + professional judgment.
Use AI to accelerate research.
Use it to summarize large documents.
Use it to identify anomalies.
Use it to generate questions.
Use it to explore scenarios.
Use it to draft explanations.
Use it to analyze large volumes of information.
Then apply controls.
Verify sources.
Reconcile the records.
Challenge assumptions.
Protect confidential information.
Apply professional standards.
Document the reasoning.
And have a qualified human take responsibility for the conclusion.
That is not anti-AI.
That is grown-up AI.
Eight Years Later, the Bet Is Paying Off
I became deeply interested in blockchain and artificial intelligence years before most clients were asking accountants about either one.
At the time, much of it seemed theoretical.
Today the practical lesson is becoming clear.
Technology rarely eliminates expertise overnight.
Instead, it moves the location where expertise creates value.
Calculators did not eliminate accountants.
Excel did not eliminate accountants.
QuickBooks did not eliminate accountants.
Cloud accounting did not eliminate accountants.
AI probably will not eliminate strong accountants either.
But it may eliminate a great deal of low-value accounting work.
And I am perfectly comfortable with that.
Let the machine do the boring part.
Give me the strange part.
Give me the balance sheet that somehow has negative cash.
Give me the 90-page valuation report whose owner cannot explain page two.
Give me the tax strategy produced after 47 increasingly desperate prompts at 1:14 a.m.
Give me the AI-generated financial statements that look ready for Wall Street while the bank reconciliation is hiding under the desk.
That is where professional judgment earns its keep.
Thank You, AI
So yes, I am enjoying the AI revolution.
Not because AI fails.
Because AI is becoming incredibly powerful.
And powerful tools create an enormous premium for people who know when the tool is right, when it is wrong, and when nobody has enough evidence to know yet.
That is the business opportunity.
The future CPA may spend less time producing information and more time validating it.
Less time calculating.
More time interpreting.
Less time preparing reports.
More time deciding whether those reports deserve to exist.
So when someone asks me whether AI will replace accountants, my answer is increasingly simple:
Bad accountants should probably be nervous.
Good accountants should probably buy better computers.
Because AI is not merely creating competition for the profession.
Sometimes it is creating customers.
And somewhere tonight, an entrepreneur is uploading an unreconciled QuickBooks file, three bank statements, two credit cards, a box of receipts, and a dream into an AI chatbot.
Tomorrow morning, the report will look magnificent.
We will be here when the lender reads it.
Author: John S. Morlu II, CPA is the CEO and Chief Strategist of JS Morlu LLC, a licensed public accounting and management consultancy firm in Woodbridge, Virginia. He has more than 20 years of professional experience in auditing and advisory work, including service as Auditor General of Liberia and FAR and DCAA compliance work at Unisys Federal Systems. He has led international audit engagements and works across the firm’s government contracting, assurance and advisory practices. He is also the founder of ReckSoft, FinovatePro, Fixaars, Signal Playbook AI and Ratevora.