How CFOs Can Jump into Gen AI for Finance and Accounting

For busy leaders who don’t have robots doing their work for them, I’ll get straight to the punchline. Artificial intelligence won’t take away our jobs any time soon, but I do think we’re going to see how transformative generative AI for finance and accounting can be in the very near future.

The other day I tried the generative AI chatbot that Workiva is developing within its unified platform for financial reporting, ESG, and audit and risk teams. I asked it to write a generic foreign currency risk factor to insert in an annual financial report. You know what? It started typing what I thought was an excellent first draft. 

I can envision generative AI one day helping teams produce reports faster, whether it’s by identifying the right ESG data to pull into financial reports or checking data and processes before external auditors do.

Generative AI tools are still evolving, but technology in general has been part of finance and accounting for years. As more countries develop ESG regulations, ESG practitioners will have no choice but to embrace technology to keep up with compliance.

Let’s look at how modern teams are already using certain tech tools, the potential of generative AI, and what CFOs can consider as you explore opportunities to advance your organisation—or your career—with gen AI. 

"Companies need to start playing with generative AI,” said David Wood, a professor in the School of Accountancy at Brigham Young University who is researching the use of generative AI in internal audits. “They need to experiment with it. They need to start seeing use cases and discussing it and putting budget aside to exploring this or they will be replaced.” 

Technology for CFOs’ teams over the years

Over the last several years, CFOs and controllers have led digital transformation projects to overhaul ERP systems, modernise general ledgers, and more. 

Your organisation may have implemented robotic process automation (RPA) to tackle structured, repetitive tasks such as data entry or reconciliation. Perhaps your team took on a project to automate the most tedious parts of financial reporting, from data collection and preparation to workflows to updating data and final outputs. (Workiva has played a major role in that, helping some teams shave days, if not hundreds of hours, off their financial reporting processes.)  

Users of the Workiva solution for SEC reporting know we’ve enabled machine learning so that our platform can suggest US GAAP taxonomy XBRL^® tags to add to the data you report to the U.S. Securities and Exchange Commission. By analysing how peers in your industry have tagged similar values, the Workiva platform also can flag outliers in XBRL tags.  

Abigail Zhang, assistant professor of accounting at the University of Texas at San Antonio, has researched the use of AI. She and her co-authors found that its use among some firms was associated with more accurate management earnings forecasts. Meanwhile, auditors have used it to analyse an entire population of transactions, not just a sample, to identify anomalies, as noted by the Public Company Accounting Oversight Board (PCAOB).

Generative AI will quickly expand what is possible in the world of work. “Generative AI is at the heart of one of the biggest revolutions in computing that we have experienced,” said Gautam Subbarao, Vice President of Product Management for Machine Learning, Data, and Analytics at Workiva.  

Generative AI for finance and accounting teams

Generative AI applications are using large language model (LLMs) to create new narratives, images, or videos, which we can then review for use in financial reports, ESG reports, or disclosures.  

Gen AI has the potential to speed up tasks we weren’t necessarily expected to ace as accounting and finance majors (writing or illustrating the first draft of an annual report, for example) so we can use our brainpower on finance, accounting, or audit issues that require our expertise. One European company has experimented with asking a generative AI chatbot to help with risk assessments, performing some audit tasks, and drafting audit reports, as noted in David’s research with his co-authors.

Yet there are risks. Sometimes generative AI produces wild and creepy hallucinations or is simply off base with its outputs. In financial reporting, our credibility rests on getting the numbers right. I don’t ever want to be wrong. 

My suggestion would be to treat gen AI chatbots as assistants, with plenty of review and oversight over the output.

“Start early, but be cautious,” said Abigail, who said first movers in adopting generative AI can have a competitive advantage.

How CFOs can balance risk and opportunities in generative AI

So where should accounting and finance teams start with generative AI? 

In my view, CFOs should prioritise internal uses such as preparing first drafts of an earnings announcement or coming up with an initial summary. You could rehearse for board presentations by asking it to brainstorm questions a CEO might face or drafting a description of a risk and control. (I’d still want the highly skilled people on my team to review outputs from generative AI before publishing them externally.)

Remember that the models that fuel generative AI are based on historical data, so they don’t yet provide real-time data the way the most popular search engines do. Members of your team will need to review, fine-tune, and augment outputs from a generative AI assistant. 

The quality of the outputs also will depend on the quality of what you ask an AI-fueled bot to do: Specific prompts with some context and direction of what kind of output you want will lead to more relevant responses.

Artificial intelligence will be more valuable to CFOs and their teams if it can understand the complex nuances of accounting, finance, sustainability, audit, and risk work. It needs to be trained with domain-specific data to make that happen. 

Based on conversations with Workiva customers, my colleagues and I know some companies are experimenting with ChatGPT to shorten summaries or as a starting point for drafting a presentation. Some generative AI applications learn from the questions you ask it, unless you explicitly ask them not to. That could limit the type of information that you feel comfortable sharing with the bot.

For our part, Workiva has enabled 300 of our customers to test generative AI within our platform. We aren’t sharing their queries with the public or storing their data. 

“We are stewards of the data that our customers entrust in the Workiva platform. Data security and data privacy have been center stage as we enable generative AI as part of the Workiva platform, and we are making no compromises on that,” Gautam said. “Workiva’s approach to generative AI is anchored both in delivering productivity gains along with responsible implementation.” 

Generative AI isn’t always going to create perfect content for financial reporting teams, ESG teams, or auditors, but humans aren’t always perfect either, David noted. 

In any case, gen AI can take the load off creating a first draft from scratch. Abigail, the assistant professor, said AI’s ability to accelerate work for us is still in progress. “My philosophy is that as long as these tools are better than our older approach, then that is OK,” Abigail said.

The future for AI in finance, accounting, and audits

I started this post by making the bold prediction that generative AI won’t take our higher-level finance and accounting jobs any time soon. 

In June, when the PCAOB asked the public for comments on a proposal that would guide the use of tech-assisted audit procedures, board member Kara Stein noted the potential of AI. Then she added, “Technology or computer-assisted analysis is a tool that can enhance, but cannot replace, professional skepticism and professional judgment.” 

Abigail noted that very basic, entry-level, manual, rules-based tasks can easily be replaced by robots. But she agrees with those who say the use of AI will spur the creation of new jobs for people who know how to work with AI effectively.

“Auditors are still very important because they have to learn how to interpret the output from AI and have critical thinking to make some important judgments,” she said.

Generative AI for accounting, finance, risk, and business reporting may still be in its early days, but I’m glad to have a front-row seat to seeing how it develops.

Best practices for generative AI in accounting and finance

• Emphasise data security and privacy
• Prioritise internal uses, such as creating first drafts, where a wrong answer won’t devastate your reputation as a CFO, controller, or member of the team
• Have a subject matter expert (a human one!) review AI-generated responses before incorporating them into your document. “We view generative AI as a means to supercharge and accelerate work, but it is not a replacement for subject matter expertise,” said Gautam Subbarao, Vice President of Product Management for machine learning, data, and analytics at Workiva
• Start with use cases that call for creating new content; traditional machine learning applications can tackle more quantitative, predictive tasks
• Prompt generative AI assistants with specific requests that include context since the quality of responses are sensitive to the type and depth of what you prompt it to do
• Aim for gen AI chatbots with industry- or domain-specific expertise that will provide the most value for accounting, finance, risk, and ESG (environmental, social, governance) teams

Subscribe for updates on how Workiva is building the future of generative AI for accounting, finance, audit, risk, and ESG teams. 

Don’t wait! Register for a free Amplify account and stream select sessions until Oct. 31, 2023. Explore how financial reporting, ESG, and GRC intersect. 

XBRL^® is a trademark of XBRL International, Inc. All rights reserved. The XBRL^® standards are open and freely licensed by way of the XBRL International License Agreement.