The ROI of generative AI: It’s growing rapidly, CFOs say

Positivity around generative artificial intelligence (AI) has exponentially risen among CFOs at large companies over the last year, but reliance on human involvement remains crucial to success, according to a recent survey.

It is not simple to determine the worth of generative AI, but nearly 90% of CFOs surveyed in December reported a very positive ROI, nine months after just 27% said the same. The latest CAIO Report conducted by PYMNTS Intelligence canvassed 60 CFOs at U.S. companies with at least $1 billion in revenue last year.

Generative AI, however, didn’t get the job done all by itself. Asked about 15 tasks that can be tackled by AI, more than half of the CFOs considered the tasks to be “mostly automated” in just two cases — generating code and creating data visualizations and reports.

A matter of trust?

CFOs don’t see human involvement in AI-assisted tasks going away, (see, “How Accountants Can Balance Technology and Critical Thinking”), but the level of trust they place in AI is rising. but the level of trust they place in AI is rising.

About three-quarters of the CFOs said they consider generative AI to be very or extremely important in financial reporting, financial planning and analysis, and strategic planning and decision support. In each of those areas, nearly all of the CFOs said they have high or complete trust in the integrity of AI’s output.

However, in one of the emerging areas for AI usage — generating new content — a survey-high 73.8% of CFOs reported that human involvement remains necessary.

Could that reliance wane?

It’s hard to say.

While trust in AI is on the rise, about half of the CEOs stopped short of declaring complete trust, instead asserting high trust. The top concern preventing complete trust was “output may not be very insightful,” cited by 22.4% of the respondents. Other top concerns included “results may be open to unauthorized access” as well as concerns about the possibility that important information may be missing and that errors and biases may be a part of AI output.

Those numbers speak to why the second-most-popular concern was “output may still require a lot of human supervision or output.”

Top uses for generative AI

• Eight of the 15 tasks highlighted in the survey were assisted by generative AI at 70-plus percent of the companies. Most commonly, CFOs reported the use of AI to generate automated responses to customer inquiries (83.3%); to run an automated cybersecurity management system and run an automated workflow management system (both at 75%); and to create data visualizations/reports and generate summaries and literature review (both at 73.3%).

• Just two of the tasks were being assisted by generative AI at less than half of the companies — generating code (30%) and innovating products and services (48.3%).

• All tasks were used by a higher percentage according to survey respondents than nine months earlier, the biggest jumps coming from generating summaries and literature review (16.7% to 73.3%); running an automated cybersecurity management system (28.3% to 75.0%); and identifying fraudulent behavior, errors, or inconsistencies (26.7% to 71.7%).

— To comment on this article or to suggest an idea for another article, contact Bryan Strickland at Bryan.Strickland@aicpa-cima.com.