By Andrew Childress, FP&A Director with 10+ years in strategic planning experience
There are few who can claim not to have taken advantage of OpenAI’s generative AI (artificial intelligence) tool ChatGPT. The meteoric rise of this easily accessible AI tool grew to 173 million users in under six months.
A Salesforce survey of 500 IT leaders found that nearly 60% of executives believe generative AI is a potential game-changing technology. Yet Italian regulators and innovation leaders have raised privacy concerns, calling these services to pause or suspend user data collection. CFOs and finance professionals rely heavily on data. Do they share these concerns, or does the excitement about AI’s potential outweigh those concerns?
ChatGPT is the most high-tech AI that can generate text that we’ve seen so far. The speed at which this new type of AI can learn is something else and might give us more reasons to worry. With more and more people using it and no signs of it slowing down, how can we be sure companies like OpenAI, which make this sort of AI, won’t misuse the business data that users input? What happens if they sell or accidently spill this data to another business? What can FP&A learn from generative AI tools?
ChatGPT may still be a fresh face in the sphere, but it’s already showing just how valuable generative AI can be in eliminating humdrum tasks. Tackling routine work is more of a snooze fest for teams and doesn’t really add much in terms of value. One MIT study even demonstrated that workers appear to be 37% more efficient using ChatGPT.
Generative AI can take on time-consuming tasks like analyzing large data sets, generating reports, or writing emails. Given the data-heavy nature of financial planning and analysis (FP&A), artificial intelligence poses a huge opportunity for optimization.
FP&A teams are drowning in data. Large datasets can overwhelm to the extent that it becomes easy to miss important trends. Metaphorically, the forest has so many trees that require constant pruning that many analysts forget the forest entirely. In turn, this can result in missed opportunities, or worse: misdiagnosed scenarios.
The access to instant analysis of complex information that generative AI offers presents potential for strategic advantages. You can test more scenarios, analyze more outcomes, and ultimately generate more value for the business.
Generative AI can eliminate some of the most tedious parts of financial tasks, creating considerable efficiencies. Insights offered by AI could enable your financial planning and analysis team to reduce manual work and focus more on tasks that enhance the company’s profitability.
Why are FP&A teams apprehensive about using ChatGPT?
Business intelligence (BI) data is the secret sauce that drives business decisions. Financial data is a valuable asset used by businesses to assess and predict performance. Given that this information discloses a lot about a company’s financial health, it’s particularly sensitive in nature.
For AI to provide company-specific insights, it must access relevant financial and business intelligence data. However, this data is among the most valuable resources a business owns and in the wrong hands can have severe consequences. Undesirable outcomes could include loss of strategic advantage, disruption of stock prices and an erosion of stakeholder trust.
Data leaks are already a threat to the security of financial data. Events like the data breaches at Revolut and X (previously Twitter) last year caused considerable damage to customer and investor trust. In light of the security problems that ChatGPT has already experienced, it is understandable that financial teams may be reluctant to voluntarily share such sensitive information with a system that connects millions of users all over the world.
How can finance teams understand and identify key data risks?
ChatGPT is powerful but imperfect. It has the power to analyze and simplify complex data very quickly. But, it still fails at higher level competency tasks such as passing accountancy exams. While it’s comforting that the CFO’s job is safe for now, AI can’t be trusted as a standalone “source of truth.” So, what actually happens when ChatGPT gets it wrong?
When dealing with sensitive data in industry, mistakes have led to extremely damaging unethical events in the past. One of the biggest problems with ChatGPT is that it is a ‘black box AI’, meaning that it does not provide the user with visibility over how it makes its decisions.
In 2019 Apple was accused of discriminatory treatment of customers via its Apple Card because it failed to take information on customers’ gender into account. Apple’s problem with bias in its AI and machine learning (ML) models had a direct backlash on the product. But, it also serves as a warning for when AI makes decisions that impacts day-to-day credit worthiness that impacts its users.
This ties directly to data security. Businesses that share data with ChatGPT need to understand the potential risks involved. Without the ability to fully comprehend the reasoning behind the decisions made by ‘blind’ generative AI, users are left vulnerable to risks concerning how their inputted data is used, who it’s shared with, and why. Without answers to these questions, there can be no absolute assurance of data security.
Can finance teams mitigate the risks of generative AI to access the benefits?
For AI to provide valuable insights, finance teams must input high-quality data in large volumes, followed by training the AI to analyze this data. But first and foremost, users need to ensure they have the right set of data to begin with.
With the right data collection tools implemented across the business, it is possible to own, collect and access valuable data into all aspects of the business. Robust data in terms of project completion, labor costs, and employee expenses relies on flexible and intuitive data. Using interoperable project and employee time tracking tools, like Kissflow for project timelines, Expensify for budgets or Beebole for timesheets, is key. Only proper tools that work seamlessly with existing programs are capable of capturing the granular detail needed for useful outcomes in the finance function.
For maintaining optimal data hygiene, it’s crucial to have comprehensive data. It’s important to use a system in your business that works seamlessly with other internal systems. Once this is achieved, finance teams can ensure they comprehend the data they are using to power generative AI, and that it’s in line with their intended goals.
However, the use of more valuable data increases the risk, necessitating further actions to ensure businesses share only the data they would be comfortable disclosing publicly in case of a data breach. For managing this, having top-notch talent onboard is invaluable.
Even with top-quality data, impactful decision-making using generative AI relies heavily on experts asking the right questions. The most skilled financial professionals will always utilize the tool most effectively for their specific objectives. Businesses must implement processes to ensure the end-to-end results are accurate.
In the case of ChatGPT and other generative AI tools, much of the responsibility falls on providers. It is incumbent upon these services to incorporate transparency into their programming. However, businesses utilizing the service need to be diligent in maintaining data cleanliness and precise phrasing when interacting with generative AI.
An interesting note will be the parallel development of generative AI and the professionals using it. Training programs that address the inherent risks will be vital for success in embracing advancements.
Is generative AI a realistic solution for the finance function?
As technology evolves, ChatGPT will need to prove its commitment to and success in safeguarding the data it handles. Until then, businesses must proceed cautiously to ensure that strategic gains are bolstered by safe and ethical usage. Top-tier finance personnel can command the tool, balancing data security with potential benefits.
The security of ChatGPT is still under scrutiny. Other generative AI systems will undoubtedly emerge as alternatives. Still, the rise of ChatGPT has prompted businesses to delve deeper into their operations to discover potential hidden value through high-quality business intelligence data. Gathering this data is just the initial challenge. The ongoing task for finance teams is to make sure they’re not missing opportunities to leverage existing data to extract powerful insights that can enhance strategies for operations, spending, and profitability, irrespective of whether generative AI provides the solution or not.
