AI in accounting: Weighing the pros and cons

April 2, 2025 10:00 AM

Whether you think artificial intelligence will usher in a great new era in accounting, or signal the profession's imminent downfall, you can find someone who will agree with you.

"We as a society are attaching all of our hopes and fears on AI," Bill Whyman, founder and principal of consultancy Tech Dynamics, told attendees at the AICPA's AI in Accounting & Finance Symposium, held last week in New York City. "Marc Andreessen says all the fear is irrational; he's worried we're going to miss out on all the opportunities from AI. On the other hand, two out of three winners of the Turing Prize say AI represents an existential threat to humanity."

Speakers at the symposium laid out a host of ways — both positive and negative — that AI will impact accounting, from commoditizing large parts of the profession's traditional work to creating whole new areas of demand for attest services.

While there was no consensus on whether AI's long-term impact will be positive or negative, it was clear that will be significant.

For example, Gartner managing vice president Matthew Kiel noted, "By 2029, one-third of finance staff will occupy 'shared jobs' where AI and the person are responsible for a single job role together."

At the same time, the transformation to an AI-driven world will generate opportunities for everyone's favorite most-trusted advisor.

"Public accounting firms are farther ahead than those in business and industry, and that creates opportunities," explained AICPA president and CEO Mark Koziel. "The firms can help their clients in industry. We are well-positioned to take these technologies and bring them to our clients."

"AI is not going to disrupt the accountant — it'll change what the accountant does, but it will not replace the accountant," Koziel concluded. "As long as we keep up with the skills we need, we'll continue to be a profession that prospers far into the future."

With that in mind, here are 10 takeaways from the symposium that demonstrate the imminent major impact AI will have — and how much we still have to figure out about it.

1. Don't get FOMO yet

Antony Weerut - stock.adobe.com

It can often feel like everyone else has jumped on the AI bandwagon except you — but the fact of the matter is that it's still early days for most people, with the majority of businesses and most accounting firms only just beginning to dip their toes in.

"Be careful of surveys that say, '75% of companies are using AI!'" Tech Dynamics' Whyman said in his kickoff presentation. "Often it's just one little sandbox operation with no real-world applications."

Nevertheless, it isn't a topic you can ignore, with all the attendees agreeing that accountants need to at least begin educating themselves on how AI will impact their clients and their firms.

2. Confusion in the regulatory environment

assetseller - stock.adobe.com

One area where it's clear that AI is still in the early stages is regulation, where multiple actors are taking often-contradictory stances.

"We're already seeing very different approaches to AI regulation in the U.S., EU and China," explained Whyman, adding that the federal approach to AI in the U.S. has undergone a 180-degree turn in just the last few months. The Trump administration revoked a hundred-plus-page Biden administration executive order that focused on protections and safeguards against AI, and replaced it with its own two-page EO that focuses on preserving and extending American dominance in the area, while calling for an action plan to be developed by the middle of this year.

"As the focus moves away from AI safety, we're seeing states step into the gap," Whyman continued, noting that over 781 state bills have been introduced in 2025 so far, as opposed to 743 in all of 2024.

"I don't think we're going to get global consensus on AI regulation — it's too soon and too early," he added.

Business leaders, naturally, are concerned about the prospects of regulatory overload, which may explain an explosion of lobbying activity: "Ten years ago, there was one lobbyist registered for AI in Washington, D.C.," said Mark Peterson, executive vice president of advocacy at the AICPA & CIMA. "Today, there are 465."

Uncertain regulatory regimes locally or nationally don't have to mean stasis for accounting firms, however. Avani Desai, the CEO of Schellman, noted that her Top 100 Firm has been working with regulatory frameworks from Europe, because there aren't any equivalent U.S. standards or regulations. "Work with regulators early," she recommended. "Don't wait for the mandate."

3. AI isn't right for everything

kartos - Fotolia

Much like the cloud more than a decade ago, AI is the next big platform shift for technology, according to Tech Dynamics' Whyman.

"It's a general purpose tech," he explained. "It will be embedded in all software, it will get better, quicker, and it will be the ubiquitous interface on screen and by voice."

"But we're just starting," he warned. "And it's not ideal for every application — if you need the exact same answer for a question every time, then gen AI may not be the right thing for you. We have databases for that."

As AI becomes more ubiquitous — and more varied — both solutions developers and users will need to bear in mind the limitations of different AI models, and the possibility that AI might not be the right hammer for the nail they're looking at.

4. Data is key

magicmary - stock.adobe.com

One major hurdle for everyone involved in AI is the quality and accessibility of the data they have to work with.

"Accounting is basically a big-data problem; it's the process of reconstructing money that has already moved or soon will," said Jeff Seibert, the co-founder and CEO of AI-based accounting solution Digits.

"We spend a lot of time cleaning data," he explained. "Deduplication is a nightmare."

At least with bad data you can clean it up; there are also plenty of sources of accounting data — from payroll systems to corporate treasury accounts — that simply may not be available to AI accounting systems, which puts a limit on their effectiveness for the moment.

For now, firms need to keep an eye on the data they're using.

"What makes our AI systems resilient?" asked Schellman's Desai. "We hired a content manager whose job is to make sure that the data we're using is reliable."

5. Saving time for what?

Sathaporn - stock.adobe.com

It's commonplace to suggest that technology — both artificial intelligence and the more commonplace forms of automation — will free up accountants to focus on more productive work.

But that's not all it will free them up to do, and there's no guarantee that all that extra time will be put to good use, according to a recent survey about AI conducted by Gartner.

"Our respondents are saving 5.4 hours a week because of AI," said managing vice president Matthew Kiel. "The question is, what are they doing with that time?"

It turns out that fully 69% of those gains leaked away in non-work-related activities.

6. AI isn't perfect

In none of its many current iterations is AI anywhere near perfect — and, in fact, perfection may not be in the cards at all, particularly in the context of accounting.

"Getting to 70% accuracy is easy," explained Digits' Seibert. "It then gets exponentially harder. I don't think it's ever going to be broadly perfect."

"But that's good," he added. "That's why you'll always need accountants."

7. AI will make you more thorough

BillionPhotos.com - stock.adobe.

For all its current limitations, as well as the many cautions and caveats that come with it, symposium attendees were universally convinced of the positive game-changing potential of AI in accounting.

Aaron Harris, the chief technology officer of Sage, offered comprehensiveness as an example: "You'll be able to be more thorough as a result of AI," he said. "For centuries, because of limits on human resources, the industry often has had to rely on sampling. But AI can look all the data you have, so you can be incredibly thorough."

"You can expand the range of content you can look at — gen AI is incredibly good at analyzing unstructured data like corporate charts, images and video content, and more," he said. "The AI is equipping you to look at things beyond what you usually would."

"Instead of 'reasonable assurance,' we can offer 'absolute assurance' because we can look at every transaction," added Desai. "We hope that will let us offer better audits."

8. Clients don't know anything about AI

Picasa/elnariz - Fotolia

"I've yet to sit down with a client who could identify the top three things they hope to achieve with AI," said Dom Megna, a principal at PwC Tax.

"They can all agree that they shouldn't be paying the same amount if we start to use AI," he added, "but they don't understand the extra value we're bringing with the AI."

On the positive side, clients' ignorance offers an opportunity for their trusted advisors to come in and advise them on their AI journey.

9. Get ready for agents

Wanan - stock.adobe.com

The next frontier in AI will be "agentic AI," where autonomous systems have the ability to operate on their own, without human intervention.

That doesn't mean a human won't have to be in charge of them, however.

"This is the last generation that will manage purely human workers," explained Heather Unruh, senior vice president of solution engineering at Salesforce.com. "We need to start thinking about managing people and autonomous agents. And then we're going to have agents that are managing agents that are managing agents."

But it's important to remember that, while agentic AI isn't the end of the line: "Agentic is the next step, but it's not the last step," said Whyman. "There will almost certainly be something following agents."

10. Start small

Anton - stock.adobe.com

The FOMO many accountants are feeling amid the perceived rush to AI may convince them they need to jump in at the deep end, but experts warned against biting off more than they can chew.

"Everyone wants to solve the biggest toughest problem they have right away," said PwC's Megna. "Why don't you solve something easier first?"