Why CFOs should care about generative AI

The content of this article was composed after prompting Chat GPT-4 to distill a raw transcript extract of a 60-min conversation Deloitte Consulting's #FinanceAI leadership team. Only minor formatting, organization, and re-wording occurred during reviews.

Why does generative AI matter for CFOs?

Everywhere we turn, someone’s talking about generative artificial intelligence (GenAI)—from ChatGPT to Bard and beyond. We are on the cusp of an “iPhone moment”—a major revolution in our personal and business lives. For corporations, GenAI has the potential to transform end-to-end value chains—from customer engagement and new revenue streams to exponential automation of back-office functions such as f inance. Chief financial officers (CFOs) have the opportunity to guide the strategy for their AI-enabled organization. To start, CFOs should:

1. Understand how the finance function could be impacted

GenAI has potential to evolve every part of the finance function. Some examples may include:

Transactional finance: Activities such as billing, payments, and collections will largely be automated through AI. Finance will be able to analyze millions of contract records (for example, supplier agreements, leases) and easily decipher inconsistencies. Variance analysis (for example, actuals vs. budget) and scenario analyses will be easily produced in the format the user prefers. Reporting formats, frequency, and level of granularity will be highly flexible based on the user’s prompts.

Controllership: CFOs aren’t likely to trust AI to produce Securities and Exchange Commission (SEC) filing financials any time soon. However, highly reliable drafts for internal (for example, profit and loss by business unit and geography) and external financial reporting (for example, 10K and earnings releases) will save significant time during month and quarter end. AI will also dramatically automate activities such as journal entries and reconciliations, plus provide intelligent detection of anomalies.

Insights generation: Finance has historically struggled to meaningfully integrate financial, operational, and commercial data. Financial reporting data is often stale, limiting usefulness. Our hypothesis is that Generative AI will make it easier to ask questions and immediately get responses created from multiple disparate data sets, in real time—that’s insight.

Thoughts on talent: Finance talent will inevitably evolve. Like all prior technological disruptions, talent will learn and elevate to focus on new tasks.  Traditional finance and accounting transactional work will be automated, and new work will emerge. The “art of asking questions”—or prompt engineering, is rapidly becoming a must-have skill. Finance professionals will use their deep domain knowledge to help tune large language models (LLMs). They will help validate the AI’s outputs. The overall size of finance teams will shrink, but CFOs may need to invest in higher cost specialized, deep domain knowledge roles to scale AI strategy.

2. Leverage areas to prioritize and pilot use cases

For many companies, AI strategy conversations are happening across the enterprise (not just in finance). CFOs should be significant influencers in their enterprise AI strategy. Strategic collaboration between key leaders (for example, CEO, CFO), will play a vital role in the success of AI deployment and adoption. The journey should begin with a strategy and use cases to build proofs of concept with well-governed and accessible data. It doesn’t have to be perfect. The key is to learn fast and scale quickly.

3. Understand the risks

Adopting AI and data-driven tools in business come with certain risks that need to be addressed:

Governance and output validation: “Garbage in, garbage out” is likely true for GenAI. If models are trained on incomplete or poorly governed data, the output will not be reliable. CFOs should institute a process to validate the LLM outputs to build trust in the models over time.

“Hallucination”: Currently, LLMs have the tendency to confidently provide incorrect answers. AI can also inadvertently reinforce stereotypes or exclusions based on the training data set.

Unadjusted risk tolerance: AI may not comprehend the risk tolerance of an organization. Ensuring a transparent controls process (security and segregation of duties) will be paramount as AI is leveraged at scale.

4. Take the first step

As CFOs prepare to embrace AI, it’s crucial to begin conversations with your technology team to scope proofs of concepts that can be worked on over a six-month period. To maximize your chances of success, focus first on data-intensive areas that could benefit from automation or data analysis.

CFOs can develop an informed strategy around AI deployment and lead the way for the rest of the enterprise. Competing priorities, messy data, and legacy systems may pose challenges—but prototype anyway. Organizations that start early will set the foundation to iterate on their failures, double-down on successful strategies, and ultimately, lead the race in AI-driven transformation.

There will be skeptics, but the promise of AI is real, and CFOs should lead the way.

Authors: