Will AI eventually make work faster and cheaper? Jury’s still out, but the international division of Big Four accounting firm KPMG is already operating under those assumptions. Last week, the Financial Times reported that the firm “threatened” to drop its auditor, Grant Thornton, if it didn’t pass along savings from the use of AI tools.

In addition to its AI case, KPMG made the case that its tax situation was fairly simple and that Grant Thornton had more familiarity with its tax situation. Surprisingly, it worked. Per the report, the cost of Grant Thornton’s audit services for the firm fell 14% from 2024 ($416,000) to 2025 ($357,000), saving the firm $59,000.

However, while the bold ask might have worked for KPMG, it might have also opened a Pandora’s Box for the company. This could be the lesser-discussed risk of AI adoption.

Fair ask or slippery slope?

KPMG might feel validated in its ask, but it has probably created a slippery slope for itself and other companies that use artificial intelligence (AI) tools in their business. Not only was the company successful in demanding a lower fee, but it prominently and directly connected the fee reduction with the use of AI tools (similar to ones that it openly experiments with).

The message is clear: If your auditor’s services are replaceable, then why wouldn’t yours be too? After all, it’s called the Big Four, meaning there are at least three other similar firms to choose from if they simply say “no, we won’t pass along savings.”

That could mean that KPMG’s modest $59,000 in savings might end up costing the firm significantly more money in the long run. By seeing KPMG’s success in scoring savings on its own audit, clients might be able to leverage this knowledge to make similar demands and reduce their own audit costs.

AI as a commodity

The accounting profession is unlikely to be going anywhere any time soon, but as AI tools challenge norms, Big Four firms might find that the moats they built in public accounting are more easily bridged.

The AI tools they tout might be part of that problem, as AI models become a commodity product for financial and accounting functions. For example, Goldman Sachs has already laid some groundwork on how an AI tie-up might work in financial services, recently partnering with Anthropic on a six-month-long project to automate mundane accounting tasks.

This trend is likely to accelerate further, as tech firms aim to sell models and purpose-built products to businesses. Salesforce has been all-in on selling its new Agentforce AI platform to enterprises, while OpenAI is hiring salespeople to pitch agents to businesses.

There are big questions about how these new tools or technologies might affect the professional services business in the long run, including the accounting profession, which is facing an exodus of talent and a dearth of young people to fill in the empty seats.

An AI ouroboros

AI optimists swear by the tech’s potential to improve productivity and deliver cost savings, things which could both be deflationary. That could be potentially catastrophic for the professional services business.

Worse? They now have a case study for how it could look in this drama between two large accounting firms. It should serve as a warning about the risk of implementing the new technology and being loud about the benefits.

Most of the warning is for industry people, who are becoming too reliant on expectations about what the technology has to offer. After all, businesses now know to ask for cost savings from vendors, or to replace them altogether as more resourced firms can experiment with, implement, or build solutions of their own.

But perhaps worst of all, they have come to expect that AI will be a cost and time-saver, while also boosting efficiency. And at this stage, that outcome seems increasingly likely to be challenged by the prohibitively high cost of investment by tech giants. After all, AI services are relatively cheap today; they’re like loss leaders.

That could put firms in an awkward situation where they are urged to pass on savings based on today’s beliefs about the technology, only for the real costs of Big AI’s capital expenditures to sneak up on them.

In a way, it sort of anecdotally feels like an Uber situation, just with a few more players. Where rideshare apps like Uber started out as cheap alternatives to their regulated taxicab competition, we know what eventually happened once everybody was hooked on the convenience factor …