- stock.adobe.

7 hurdles to using AI in accounting and finance

Setting realistic expectations and using the same quality control measures that employees must follow can ensure success when you apply AI to financial data and processes.

As more organizations deploy AI in accounting and finance, they're coming up against a common set of hurdles.

Consultants who helped clients implement AI in accounting and finance operations and use it themselves identified seven typical challenges.

Lack of off-the-shelf software

ERP vendors are shipping real products, and niche players abound (see sidebar).

Nevertheless, a dearth of packaged tools is the main reason few companies have deployed AI in accounting and finance, said Robert Kugel, senior vice president and research director at Ventana Research.

"We're really waiting for the vendors to build these capabilities into their systems," Kugel said. "I don't want to say this stuff is hype, because it's real [but] it's just going to take time."

In other words, you'll likely need some customization.

Off-the-shelf niche software can handle 60% to 70% of the use cases that companies are considering, but some internal development is still required, said Adrian Tay, managing director of finance and CFO services at Deloitte Consulting.

"You definitely require that expertise in house to be able to get data and make sure you understand the algorithms," he said.

Insufficient data

Virtually every company produces an avalanche of data, but that doesn't mean the finance team has the right data. "There are quite a number of CFOs who don't have confidence in the data set," Tay said.

Applying machine learning to bad data risks unreliable analysis, he said.

Unrealistic expectations

Will Bible, Partner, Audit & Assurance Practice, Deloitte USWill Bible

Using AI in accounting and finance is no panacea.

If people don't expect perfection from AI they are less like to be disappointed, said Will Bible, partner in the audit and assurance practice at Deloitte US.

"Sometimes, if you have a process but it was only doing it right half the time, the person on the other end might distrust the system so much that it's not even worth having [AI] involved. You're having to correct it all the time," Bible said.

Employees who mostly use AI as a first-pass tool can better tolerate inaccuracies while training the system, he said.

Inadequate change management

CFOs and business leaders need to address resistance issues.

John Van Decker, Vice President and Analyst at GartnerJohn Van Decker

Some companies are still reluctant to move their financial systems to the cloud, where most AI offerings reside, said John Van Decker, a vice president and analyst at Gartner. Once they do, they need effective change management to take advantage of AI.

"It's not a matter of flipping the switch," Van Decker said.

AI can perform sophisticated matching to bring purchase orders and Accounts payable (AP) processes together, he said. But problems arise if leaders fail to understand or communicate how the new processes will affect AP jobs and set thresholds for what AI can handle on its own. Implementing AI in finance requires considering the affected processes before choosing technology.

"It's not like we're just moving to a solution and we could just forget about it for another 10 years," Van Decker said. AI needs care and feeding.

Distrust of AI

Another issue holding back the use of AI in accounting and finance is the technology's "black box" issue.

Adrian Tay, Managing Director, Finance/CFO Services, Deloitte ConsultingAdrian Tay

CFOs say that AI works on a very complicated algorithm that's "in the box," and question whether they can learn to trust it, Tay said.

Two solutions: Compare AI to human output and hire people with the talent to guide, monitor and govern the technology, he said.

Some companies choose AI forecasting tools that are less accurate than ones that use highly complex algorithms because they understand their methods better, he said.

Deloitte's auditing practice employs a "human in the loop" approach, Bible said. The AI software, named Argus, handles some tasks automatically, but auditors make the final judgment.

"We've retained the same quality-control framework that we've had in place for every other audit process, which means multiple levels of review."

Vendors start with narrowly focused AI

ERP vendors are among the key players shipping AI products.

  • Oracle has Adaptive Intelligent Apps in Oracle ERP Cloud for supplier categorization and intelligent payment discounts.
  • SAP outfitted S/4HANA ERP with AI for financial planning and analysis (FP&A), treasury management and invoice-to-cash, invoice-to-pay and procure-to-invoice processes.
  • Sage Intacct released information that its Intelligent GL (general ledger) will be available this year.
  • Workday offers four applications built on the Adaptive Insights analytics platform acquired in 2018, including anomaly detection in its SaaS planning tool; optical character recognition for expense reports; supplier invoice automation; and journal insights, which detects GL anomalies.

Other financial management vendors are jumping in.

Prophix, which makes corporate performance management software, launched AI-based Virtual Financial Analyst last fall.

Blackline released information that it will use Google Cloud Platform to add machine learning to its rules-based robotic process automation tools for financial close management in ERP.

Niche AI vendors include, which makes machine-learning tools for AP, cash flow and other common workflows. Mindbridge AI Auditor is machine learning that combs accounts receivable, accounts payable and GL data for errors or fraud. sells trainable bots that scan documents and handle invoicing, procurement, expenses and payroll.

Concerns about bias and liability

AI's use in accounting and finance hasn't become popular because CFOs distrust the recommendations made by the "black box." The worry is that AI-assisted risk analysis will produce false positives or negatives, said Bob Woods, a partner at PwC.

Bob Woods, Partner at PwC USBob Woods

"Today, you can have a human sitting and reading a report, looking at a set of data and making a judgment that says, 'This is a risk to the company -- for example, a corruption risk,'" he said. "AI is taking some of the judgment out of that process."

That's why running AI pilots in parallel with traditional processes and comparing notes is critical, Wood said.

Fear of job loss

The rise of AI has engendered fears of job loss as machines take over work performed by people. The risk applies to using AI in accounting and finance.

Future workforces will be leaner, especially in operational finance -- functions such as order-to-cash and "transactional" accounting, essentially, bookkeeping, according to a report from Deloitte. On the other hand, there will be growth in analytical jobs and specialties like tax and treasury management.

Displaced workers will need retraining for these new jobs, Van Decker said.

"You can't just take an accounts payable clerk and turn them into a financial analyst," he said. "There has to be an investment in retraining going forward, [and] we're seeing a lot of that already."

Before smartphones and easy laptop connectivity, people at Deloitte had to print reports and put them in binders, Bible said.

"Now [employees] are doing things like analytics their first year on the job," he said.

Robert Kugel, Senior Vice President and Research Director, Ventana ResearchRobert Kugel

Jobs have become more complicated and specialized as employees navigate client IT systems to perform audits. Technology has led to more jobs, not fewer.

Despite these hurdles, CFOs should be "fast followers" of AI and put the required competencies in place so they are ready to act, Kugel said. They needn't staff up with "propeller heads," but finance and accounting pros with adequate understanding of AI.

"A lot of people who are trained as accountants, even a lot of people in [financial planning and analysis] don't understand how the technology works," Kugel said. "It's all magic, and because it's magic, there's no ability to really grow with the technology. That's not going to fly in the 2020s."

Next Steps

Finance industry giants disclose AI challenges

Dig Deeper on ERP implementation

Data Management
Business Analytics
Content Management