The cost-effectiveness of AI in accounting is still unknown and will be affected by many factors

The Unpredictable Economics of AI Accounting

The biggest question facing small business owners isn’t whether or not AI will change Accounting, it’s how the economics of Ai accounting will actually play out when the dust settles.

Right now, nobody knows. The software vendors selling AI tools don’t know. The accounting firms implementing AI Accounting don’t know. Business owners paying for it don’t know. The economics of AI in accounting are changing, and they’ll remain unstable for years for a variety of reasons.

Think about this… As I write this, the LLM companies (OpenAi, Grok, Anthropic,…) are throttling usage of their free models and pushing users toward subscription plans. You have experienced this if you’re using these free tools significantly. The idea of “free AI” is slipping away.

Microsofts AI CEO, Mustafa Suleyman, just wrote: “For tne next couple of years at least, the entire AI industry is going to be defined by this fact: demand is going to wildly outstrip supply, and so what matters is which companies / products have margin to pay for tokens.” He continued, the products that can pay will improve the fastest because faster responses drive retention, retention creates data, and that data spins a flywheel of model improvement and adoption.

This begs the question: If models learn from data, and data is only going to be supplied by big businesses that can pay to use the model, then will the model be useful to small businesses that can’t afford to pay today, or that haven’t decided to participate in the AI transformation? Will the models be biased, or will the engineers be able to manage this?

This uncertainty creates risk for businesses trying to make decisions today. Do you invest in AI tools now, betting that the costs will stabilize? Do you wait, risking falling behind your competitors who move faster? Or do you stick with human-in-the-loop bookkeepers, knowing that approach might get more expensive as the accounting labor shortage deepens?

There aren’t any simple answers because the economic model for AI bookkeeping doesn’t exist yet. We’re watching it get built in real time, with all the trial, error, and unexpected consequences that come with major shifts in technology.

The KPMG Warning Shot Heard ‘Round the World

In early 2025, KPMG did something that sent shockwaves through the accounting profession. The Big Four firm demanded that its auditor, Grant Thornton UK, cut fees citing gains from AI efficiencies. The argument was simple: if AI makes audits faster and cheaper, those savings should be passed along to clients. That might not have been their best idea…

Grant Thornton resisted at first. KPMG reportedly threatened to find an accountant who would play ball. Eventually, Grant Thornton gave in. They cut their audit fee from $416,000 in 2024 to $357,000 in 2025, a 14% reduction.

KPMG just handed every business owner the perfect script for renegotiating their own accounting fees. “If AI is making you more efficient, why aren’t you passing those savings to us?” If you believe the headlines about AI, it’s a reasonable question. It’s a question accounting firms will hear repeatedly over the next few years.

But here’s the problem. KPMG may have won a short-term victory while creating long-term trouble for itself and the entire profession. Just reread Mustafa Suleyman’s quote. If clients can demand fee reductions based on AI efficiencies, firms face a revenue squeeze just when they need money to fund AI development and Accountants who are getting hard to find. Building and operating AI systems costs millions. If every client negotiates lower fees, where does the money come from to keep improving the technology?

There’s also the question of whether or not those AI efficiencies actually come to life. Grant Thornton’s UK audit leader wrote in December 2025 that work was becoming “faster” and “smarter” thanks to automation. But he also noted that audit quality depends heavily on expert human judgment, and that fees “reflect both the cost of our people and the cost of the technology that supports them.“

KPMG admitted “while it is true AI can create efficiencies, developing and operating AI systems can generate additional costs.” The firm acknowledged that audit pricing depends on complexity, scale, and the expertise and technology required.

So which one is it? Does AI reduce costs or increase them? The honest answer is both, depending on the timeframe and how you measure it. Ambiguity is exactly why the economics of AI are so unpredictable right now.

The 30 Percent Rule of AI Return on Investment

People tend to develop thumbules around concepts and technology. 

In 1965 Gordon Moore coined “Moore’s Law” and in 1975 he revised it. Initially, he suggested that technology could double every year. In 1975 he modified the law and stated that computing power would double every 24 months.

While several “rules” have popped up around AI over the last year, one that seems to be gaining ground is the 30% Rule of AI. The 30 Percent Rule states that only about 30% of repetitive administrative labor tasks can be handled by AI. The remainder of the work which is more strategic and creative will be left to humans. The second part of the 30% law is that 30 percent of the cost savings of AI ends up unrealized because it gets reinvested to verify or rework the AI output.

What this means is that ultimately, if true, AI will result in a 20% reduction is repetitive, white collar administrative jobs. If true, the KPMG cost savings with their auditor makes sense this early in the technology development cycle. But don’t get too excited just yet. We haven’t yet had to absorb the real cost of the AI infrastructure build-out and that will likely have a negative impact on the economics of AI.

What Factors Will Affect the Economics of AI?

The economics of AI will be impacted by a sizable list of factors. Those factors include:

  • 1) The labor shortage in the accounting industry
  • 2) Will AI become a regulated utility?
  • 3) How Insurance Companies will insure AI-powered decisions
  • 4) How courts will treat work product and attorney-client privilege in the use of AI
  • 5) Will the use of AI require new professions?
  • 6) How much of the savings will have be reinvested for quality control and rework?
  • 7) When will AI be available at scale?

The Labor Shortage Factor

AI is being implemented in the accounting profession at the same time a natural and severe labor shortage (for unrelated reasons) is making human talent more expensive and uncommon. About 65% of current CPAs are nearing retirement age. The accounting workforce has shrunk by over 17% since 2020 (not because of AI), with more than 300,000 professionals leaving the field. Universities are producing 20% fewer accounting graduates than they did in 2010.

This creates a strange dynamic. Businesses need AI partly because they can’t find a bookkeeper or accountant easily. But implementing AI requires skilled professionals who understand both accounting, bookkeeping principles, and technology. Those people are in even shorter supply than traditional bookkeepers.

The result? Short-term costs are likely going up, not down. Companies trying to become “AI-ready” need help cleaning their data, configuring systems, and training staff. That work requires experienced professionals who command premium rates. EY announced a 10%+ salary increase for accountants in 2025 as part of a $1 billion investment to address the talent shortage. Starting salaries for financial analysts are climbing fastest, and that doesn’t include signing bonuses and retention perks.

Over 90% of finance and accounting leaders report difficulty finding qualified professionals. The unemployment rate for accountants and auditors was just 2.0% in 2025, indicating extremely tight labor markets. Demand for bookkeepers, staff accountants, and controllers remains strong because even with AI, businesses still need humans to validate results, apply judgment, and make sure the results are accurate.

So in the near term, the combination of labor shortages and AI implementation costs could actually drive spending up for many small businesses, not down. The promised cost savings remain somewhere in future years. The economics of AI are nowhere near settled.

AI as a Utility

Looking ahead 5 to 10 years, AI in accounting will likely evolve into something resembling a utility, very much like cell phones or cable television. In 1988, a cell phone was a luxury. By 2008, it was considered core infrastructure, and various regulations emerged along the way to govern pricing, access, and service standards.

AI will probably follow a similar path. Some basic level of AI capability might be included in existing software subscriptions (which will likely cost more) or be available for free (maybe). But the real computing power required to run advanced AI accounting features will need to be paid for somehow, and that will be through tiered subscription models based on usage.

Small business owners will see their cost structure for bookkeeping will gradually shift. Labor costs for human bookkeepers won’t disappear entirely, they’ll still have a human in the loop, but they’ll be partially converted into monthly bookkeeping AI utility bills. You’ll pay for computing capacity the way you currently pay for electricity, internet service, or cell phone plans. 

This utility model brings its own uncertainties for the economics of AI. Will pricing be stable or will it fluctuate based on demand? Will the government regulate price and availability? During peak usage times, will costs surge like they do for electricity? When computing capacity is scarce, will prices spike? Will there be basic, standard, and premium tiers of service, and how much will each cost?

We don’t know because the infrastructure to support AI doesn’t fully exist yet. As discussed in the previous section, data center capacity is severely limited right now. Until supply catches up with demand, pricing will remain volatile and unpredictable.

The Insurance Question

There’s another economic factor that hasn’t been fully priced in to AI yet: professional liability and errors and ommissions insurance for AI-related errors.

Professional liability insurance for accountants and CPAs currently costs an average of $45 per month, or about $537 annually for typical coverage. But most policies were written before AI became commonly used among professionals. The terms and conditions may not explicitly cover AI-related claims.

As this is being written, commercial insurance providers are already issuing exclusions for Ai-powered human resources decisions and for Ai-powered marketing activities like chatbots. These aren’t going to be covered.

Some professionals may discover, after a claim is presented, that coverage for AI mistakes isn’t clearly included (or may be intentionally excluded) in their errors and ommissions policy. If a CPA relies on AI accounting output that turns out to be wrong, does the insurance cover the resulting liability? It depends on whether th use of AI tools satisfies the policy’s definition of “professional service” and whether losses flowing from AI use are covered.

Insurance companies are still figuring this out. As AI becomes more prevalent and claims start appearing, premiums may increase to reflect the new risks. Firms might need separate cyber liability policies to cover AI mistakes. The costs of these protections aren’t yet reflected in most firms’ pricing.

For small business owners, this matters because those increased insurance costs will eventually show up in the fees you pay. Even if AI makes some work faster, the added liability costs could offset part of the savings.

So far, the economics of AI has to account for a severe labor shortage, efficiency gains, the cost of multiple data center build-outs, liability, and we’re not done yet…

The Legal Questions

Are Ai searches discoverable?

Yes. Recent court rulings have decided that if a party can demonstrate “proportionality and relevance” the prompts entered into Ai tools are discoverable. But, the party seeking to gain access to the prompts does have the burden of proving proportionality and relevance.

Does using AI violate attorney-client privilege?

The question is poorly posed. The question that should be asked is “Does using AI waive attorney-client privilege? And the answer to that question is “maybe, maybe not”.

On February 10, 2026, The U.S. Southern District of New York issued a ruling stating documents generated using a publicly available AI tools are not shielded by attorney-client privilege or the work product doctrine.

On the same day U.S. District Court of the Eastern District for Michigan ruled that work product protection is applicable to materials generated by a third-party AI tool prepared by a pro se plaintiff in the course of litigation.

So what does this mean? Does it, or does it not violate or waive attorney-client privilege? The short answer is that for right now, we as a society haven’t decided, and the answer varies depending on where you are in the country, how certain terms are used, and who does the work, and probably who argues the question.

The Michigan ruling seems to suggest that a small business owner who uses the tool (pro se) might not be waiving attorney-client privilege, but if their CPA or Attorney uses the tool, that might be equivalent to waiving the privilege. 

How Does Ai Use My Information?

When you use an Ai tool to do work, the terms of service allow the Ai company to review prompts for training or disclose them to authorities, users have no “reasonable expectation of privacy” for that data.

When you enter financial information about your business into an Ai tool, you’re giving that company permission to use your business information in ways you many not yet have considered. 

So What Does All of This Mean?

As a society we haven’t yet decided what protections are afforded to Ai or the use of Ai. These sorts of questions still need to be litigated in the courts. Until then, this means risk. In business, risk is always a financial question. What we do know is that litigation will play a huge factor in the economics of AI.

Employment Shifts and New Roles

The long-term economics of AI also depend on how employment in the accounting profession evolves. Most projections suggest that AI will reduce employment in routine bookkeeping and data entry roles by 40% to 60% over the next decade. The Bureau of Labor Statistics projects 5% growth for accountants and auditors but a 5% decline for bookkeeping clerks through 2034.

AI might also create new roles. We’re already seeing demand for AI Implementation Specialists in accounting and bookkeeping, professionals who help companies migrate to AI-assisted workflows. Data governance officers, financial systems architects, and AI compliance analysts are emerging positions that didn’t exist a few years ago.

These new roles tend to command higher salaries than traditional bookkeeping positions. So while total headcount might shrink, payroll costs per employee could increase. The net economic impact isn’t clear yet.

What AI Cost-Effectiveness Means for Small Businesses

The unpredictable economics of AI bookkeeping create planning challenges for small business owners. You’re being told that AI will save you money, but the timeline for those savings isn’t clear. In the meantime, you might face higher costs for data cleanup, system implementation, and professional oversight.

The KPMG episode suggests that negotiating lower fees based on AI efficiencies is possible. But it also creates tension in the profession about whether those efficiencies are real and lasting. Firms are spending heavily to develop AI capabilities while facing pressure to cut prices. That’s not a stable economic model.

As AI evolves into a utility, expect your bookkeeping costs to shift from mostly labor to a mix of labor and technology subscriptions. There’s a good chance you’re going to need a “human-in-the-loop”. Whether that total cost ends up being higher or lower than today’s all-human approach remains to be seen.

The safest assumption is that AI won’t dramatically reduce your bookkeeping costs in the next few years. It might make certain tasks faster. It might reduce errors in some areas while creating new ones in others. But the economic benefits, if they materialize at all, will take time to realize.

In the meantime, budget conservatively. Don’t assume AI will cut your accounting costs in half. Don’t expect to eliminate your bookkeeper and replace them with software subscriptions. And don’t be surprised when the costs of AI accounting turn out to be higher and more complex than the sales pitches suggest.

The economics of AI will eventually stabilize. But we’re not there yet. And until we are, anyone promising you clear answers about what AI will cost is guessing.

Sources and References: