New technology that can process large volumes of information offers accountants and auditors a chance to shift people and resources to other needs, but accounting specialists said the technology needs human oversight to be effective.

Robotics, machine learning, and other forms of artificial intelligence hold promise for accounting. Technology for generating and processing natural language can turn financial tables into written documents, give structure to large volumes of unstructured financial data, and pore through long documents and contracts to extract needed information, according to Mike Schor, a partner at Deloitte.

But they still can’t see and hear the way humans do, creating tension between the promise of advanced technology and the need to keep an eye on that technology.

“There are more questions than answers now in terms of what this means for the future,” Laura Phillips, Google’s director of finance compliance, said Nov. 12 at an annual Financial Executives International accounting and reporting conference.

Some companies are ahead in adopting advanced technology for accounting, so it behooves auditors and accountants to explore how to take advantage of them, Schor said.

“In five years, we’ll be exponentially ahead of where we are now, so jump in now,” he said.

Success at Cigna

New technology allows auditors to look at transactions across a business to boost the efficiency and quality of an audit, gain new insights, and pick up areas of financial risk they might not have seen otherwise, said Josh Jones, a partner at Ernst & Young LLP.

Cigna is one company that is already on board with leveraging new technology for accounting purposes.

A pilot program to use robotic technology to perform simple tasks proved so successful that the insurance company rolled out a full review of its operations to see how the technology could be expanded, said Mary Hoeltzel, vice president and chief accounting officer at Cigna.

“We have to compete aggressively for resources within the business, but the cost savings were so significant that it was easy for us to get the funding,” she said.

Speed Behooves Accuracy

But the processing speed of new technology makes it all the more important to have humans monitoring it, Schor said.

If a routine bank statement format moves the key data from the top left to the bottom right of a scanned page, for instance, machine-reading technology won’t pick that up. Humans have to do that—and quickly.

Machine processing is now so fast that, if there are errors or embedded biases in the works, “things that can go wrong will proliferate very quickly,” Schor told Bloomberg Tax. “You need monitoring and controls to identify and fix problems right away.”

What if Robot Fails?

If one bot does the work of 10 people, a business will still have to have someone in charge of the bot, said Craig Schmidt, a governance and oversight leader at Wells Fargo & Co.

Updating continuity-of-service procedures to account for robot failure, he said, “will be critically important in the future.”

Companies, auditors, and regulators are all competing for the same limited talent pool to run the new advanced technologies of robotic process automation (RPA). That poses a recruiting and training challenge.

“You don’t have to be an IT aristocrat to put in RPA, but we do need people who are interested in this and can be trained,” Hoeltzel said.