In a Heidrick & Struggles survey of executives, most CFOs said their companies have not yet adopted AI in key areas such as auditing, accounting, financial reporting and FP&A. A Teradata survey released last week found that only 12% of very large companies have deployed AI across their enterprise.
But these “studies” suggest that finance functions are far behind on the GenAI adoption curve, and they make a big assumption: they suggest that AI products and services like Large Language Models (LLMs) are ready for adoption, and all CFOs need to do is line up the funds and deploy the tools.
For most mid-sized companies in the finance department, this isn't true. For them, the first task is to get a strong foundational understanding of the current wave of AI technology and how it will change their organizations and workflows. “Many CFOs are falling behind because they don't know where to start,” says Wayne Burson, CEO of BDO USA.
There are numerous descriptions of logical use cases for GenAI in the finance sector, such as processing unstructured data and incorporating it into predictions, extracting non-standard language from business contracts, acting as a helpdesk for suppliers with questions about invoices, and using tools like ChatGPT to collect and structure information for internal reporting.
“Everyone is figuring out what's real and what's possible,” says Vivek Saxena, senior vice president and leader of finance and accounting services at Genpact. “But there's a big problem. [not] “Having the right data. AI models depend on good data.” The success of GenAI adoption will also depend on the ability to manage potential cultural adaptations within the organisation, such as employees’ over-reliance on AI output or reluctance to use the technology.
Raising awareness
One way to see what GenAI can do for the financial industry is to let employees experience it. A key goal of accounting firm Marcum's GenAI initiative was to “become an organization with an AI culture in mind,” says Peter Scavuzzo, Marcum's chief information and digital officer. To do that, employees needed to be comfortable enough to contribute to the company's “AI journey.”
Marcum has given its in-house AI tools to knowledge workers so they can crowdsource accounting and advisory use cases. “If someone in the tax department is trying to solve a problem using AI, we have a conversation with them and ask, 'What's the interesting problem you were trying to solve? What were you trying to do? What challenges did you face?'” Scavuzzo says. A month ago, Marcum surpassed 100,000 prompts, with the goal of opening up its library of use cases to the public.
“This is a cultural journey shift and a change management effort,” Scavuzzo said.
Re-establishing expectations
If GenAI is truly going to make workers more productive, workers (and their CFOs) need to know “what it can do, how to build the right prompts to get you what you need, when and where to use it, and how it fits into your daily workflow,” WalkMe CEO Dan Adica said on the company's earnings call on May 22.
Just as important as letting employees try the tool is making sure they have realistic expectations of the GenAI tool. Many executives expect absolute answers and irrefutable facts from the AI output. If they don't get those, they tend to think the AI is completely useless. But learning about AI means understanding what it can and can't do. Scavuzzo likens using AI to an American football team moving down the field. With AI, you can't go 100 yards from end zone to end zone, but you can get into field goal range nearly 100% of the time, he says. “Who wouldn't want to snap their fingers and go 70 yards?”
Workers need to be aware of what GenAI is good at and what it's not good at. “The technology is certainly moving in the direction of providing more accurate and precise answers, but we can't just blindly accept it,” Scavuzzo says.
Think about ROI
Running GenAI tools in-house can be costly, and costs vary widely at this point. Genpact's Saxena said inputs into the total costs may become more clear in the next 12 to 18 months. At this point, they may be hard to quantify. Not only will companies need to have the necessary tools and systems infrastructure in place, but some employees will spend time and effort on governance, including creating usage guidelines and deciding which use cases to approve.
In finance, “the CFO organization doesn't want to be left behind when the CMO and CIO are doing things in different parts of the business,” Saxena says. But that probably means moving forward without absolute answers about whether new ROI metrics will be needed to justify future spend, especially when it comes to measuring AI's contribution to employee productivity and employee engagement.
be patient
“There are definitely use cases for AI that are incredibly beneficial for finance organizations,” says Jotham Ty, founder and CEO of Gappify, which provides accrual automation software for accounting teams. But the market is still sorting out which use cases to tackle and how to deploy them.
“How this gets rolled out, how it gets audited, how it gets prioritized, I think all of those things are still under review,” Tai said.
Tai says GapFi has a good idea about how to incorporate AI into its products, but is taking a cautious approach to launching and promoting it. “We don’t want to jump in until we have a clear message and a good way to explain to prospects and customers how to implement AI and ensure the auditability of transactions,” Tai says.
“We know some people are jumping the gun and some of our peers are touting features that will be coming next year. We're not doing that,” Tai says. “At the end of the day, as accountants, we value trust, and if we destroy that trust by advertising something we don't have, that could be a deterrent.”
Tai also feels the market could be useful for education: “One of the things that will start to emerge over the next few quarters is a better understanding of the different types of AI that are out there,” he said.
But being patient while waiting for some of that to happen doesn't mean being passive. Armed with AI tools, Scavuzzo says, smaller, more agile competitors with much less capital might be able to use the technology to level the playing field. Or, entrenched competitors in a space might be too slow to adopt AI tools, and an outside company might see an opportunity to disrupt the space.
As a result, CFOs need to think broadly about AI, Scavuzzo says. The question isn't just, “How can AI help solve my current problems?” CFOs also need to ask what AI means for their organization's customers, commercial markets and business partners.
A shorter version of this story appeared in the May 31 issue of Finance & Accounting Technology Briefing.
