The Future of AI in FP&A
One of the mega trends of our time is the advent of Artificial Intelligence as a mainstream tool. Though AI has been around in various forms for decades, it is only in recent years that it has become a topic of conversation as a usable, transformative tool.
I have done some reading on AI (see A Brief History of Artificial Intelligence: What It Is, Where We Are, and Where We Are Going: Wooldridge, Michael: 9781250770745: Amazon.com: Books) and have attended a few conferences sponsored by Finance Alliance | World Finance Forum. The purpose of this post is to summarize where I think we are with AI as it relates to FP&A. The most burning question on peoples’ minds tends to be whether AI will replace FP&A Professionals in the workplace.
In its current state, AI’s ability to run 3 statement models, forecast a business, develop a 5 year plan and prepare a deck for the board and investors is pretty limited. Remember at this stage it is a language model. So, if you try to use AI for these types of analytics you ultimately get a “how to” list for forecasting. Even the ChatGPT add in for excel is pretty limited to functions that produce words, not numbers.
For the future, my initial answer is well, it depends.
Here is a list of 6 practical considerations for assessing the potential impact of AI (using Gemini or ChatGPT for example) in FP&A:
- Company Stage and How You Connect Decision Making with Your Data
The promise of AI is in its ability to one day process large volumes of data in order to draw important conclusions about the dynamics of a business. The challenge is that large volumes of sales data, customer data, product usage etc is typically reserved for later stage organizations. Companies in their infancy will not have developed this volume of information and more likely than not have not yet made the necessary ERP investments to get there (very few start ups have FP&A as native to the organization). Therefore, the opportunity to connect data and decision making is limited to Stage 4/5 organizations.
2. The Type of FP&A Work That You Are Doing
FP&A work tends to span a rather broad spectrum that connects the dots from financial results post financial close to telling the story of those results to communicating that information to a wide variety of customers (boards, investors, bankers) and ultimately making recommendations around long term planning. Those activities go way beyond data science and scenario analysis that could be offered by generative AI. The art of FP&A is in the contextualization of information and more importantly “reading the room”. Even with the ability to program in preferences, AI will not be able to operate in a way that leverages the five human senses. Experienced FP&A professionals and especially those with M&A experience can tell you that there is so much nuance in bringing companies together (or breaking them apart) that more often than not, the decision is made despite what the financials say.
3. The Promise of Automation and Systems
Harnessing data and achieving maximum utility in its use has been the promise of every system tool since the first instances of Essbase and SAP BI. I have discussed this topic previously in my blog. No matter how sophisticated the system, the challenge always lies in how a business leader wants to tell the story of the business. Sometimes it is supported by the data but more often it is not. Telling the story of the business using historical information can sometimes support stories of performance but the information may not be aligned with strategic objectives and investment plans. Unless all preferences and unexpected outcomes are programmed into the learning model you are using, it is a tall order to replace and FP&A person with AI.
4. What Are Your Resources Like?
Companies who are at a stage with a large dataset may very well be among the earliest adopters of AI in FP&A. The challenge is that in the initial adoption it is highly likely that FP&A teams will need data scientists to partner with in the development of the tool. There are several thought leaders who discuss the merger of data science and FP&A. Smaller teams with resource constraints may have difficulty with this. More importantly, a data scientist is not an FP&A person and vice versa. FP&A leaders going down this path will need to be able to synthesize disparate skillsets to bring a full picture with effective business recommendations to the C Suite.
5. Can Your Organization Absorb the Opportunity Cost
There is an opportunity cost associated with database automation – whether it is a data warehouse (like in the old days) or ultimately a sophisticated AI system that tells you everything you need to know about your business. That opportunity cost is institutional knowledge. No data set in business is perfectly clean. Experienced FP&A professionals know that spending time scrolling through tens of thousands of lines of data allows you to see the anomalies. You can think of these as “one-time events” that can dirty your data. For example, in any given quarter you may have a blue bird customer fall from the sky that distorts the trend line in your business. Unless you program all of the potential one-time events into an AI model, you will get distorted view of what is happening.
6. Risk Management and AI
The fact that AI is coming to FP&A in some form and even with limitations is unavoidable. The most important thing that organizations can do is to protect themselves now by instituting AI policies ASAP. Understanding the exposure is a first step and especially so with a remote workforce. AI’s use in creating work product in the absence of company policies for its use can lead to less than desirable outcomes. Companies need to be clear about what employees share with the learning models and more importantly what they should not share. Moreover, AI and any of its use in creating work product has to be treated like an employee. Work still needs to be reviewed and contextualized. Think of it as one of those situations where you literally get what you ask from generative AI but it may not be what you really wanted. That’s where FP&A leadership comes in.