How Finance Can Beat the Bots With Prescriptive Analysis

Matthew Dziak

Be a Prescriber, Not Just a Predictor of Outcomes

By now you are either one of the 100 million users of chat bots like ChatGPT, or you have heard the stories of how LinkedIn co-founder Reid Hoffman wrote the book Impromptu, in a few months using ChatGPT.

There is no question a portion of roles will be either greatly reduced or limited due to the adoption of artificial intelligence (AI) and machine learning technology. According to a recent report from Goldman Sachs, 300 million jobs, including  35% of business and financial operations could be automated, making it the fifth most industry to be impacted by AI. 

It may be human nature for us to fear the unknown or workforce displacement, however, it is more likely that AI will weave its way into many aspects of business, and our lives, and the ones who harness its capabilities will be the ones who thrive.

For finance, it’s one thing to leverage AI and machine learning to predict a likely outcome with relative certainty, but that is just the beginning. The goal is to use that awareness to analyze and model scenarios and implement proactive measures to combat the risks and opportunity loss of not being agile in your plans. 

However, predictability is only part of the equation, and to get to the solution, where you are making the most impact, you need to consider all factors. The devil is in the details and the details are what matter most. Predicting the future has more applications than determining the winning lotto numbers or revealing strangers fortunes. Surely, if you had that superpower, you’d be reading this from your own private island (if at all for that matter).

The reality is we can not predict the actual outcomes, but we can analyze trends, determine probable outcomes and adjust strategies as needed - and this is where finance can be the champion of strategic insights.

Having a glimpse into the likely outcomes is essential for any business, especially in the face of uncertainty. With the bots handling the tedious operating tasks, finance must focus on analysis and develop insightful game plans for potential scenarios facilitated through prescriptive analysis.

What is Predictive and Prescriptive Analytics? 

Predictive analysis reveals what might occur and factors in historical trends and metrics. This is essentially how models work and is a bedrock use case of machine learning tools like ChatGPT, which gathers billions of data points and insights from the internet and its developer’s (OpenAI) database to answer questions posed by users.

Prescriptive analysis suggests ways to optimize future outcomes and improve or maintain trends identified in predictive analysis. Although both are necessary, in the hierarchy of analysis, prescriptive analysis comes out on top and is what high-impact FP&A teams provide. 

For a full breakdown and comparison of descriptive, predictive and prescriptive analytics, read through our previous blog post

Prescriptive Analysis Examples

Scenario planning is a core function of FP&A and one that gained more traction during the pandemic, which revealed which companies had the fortitude and established processes to remain agile in their approach to growth and potentially pivot or double-down on initiatives. The goal is to inform business unit leaders of their progress toward a target and educate them to take a proactive approach and prepare game plans for what might happen - instead of reacting to new trends and extraneous factors. 

If you (or the bots) predict input costs will rise 5 to 7% in the first half of next year, that’s great insight, but now what? To be a best-in-class finance function, ask yourself the following questions:

  • What can be done with that information that is helpful to executives and budget owners?
  • How will that impact our COGs compared to our plan? 
  • How will that impact our current and projected bottom line? 
  • What are our options to offset these cost increases?
  • Are there other vendor options or can we explore new payable terms? 

These are the prescriptive measures that finance must identify and provide answers to be effective business partners. Finance holds the keys to the flow of operational information and through critical analysis, you can unlock the necessary drivers to fuel optimal decisions. 

For example, a software or IT systems vendor might identify users of a given industry have a 15% higher rate of professional service revenue on average compared to others. This insight could lead you to recommend to your head of sales and customer success that they should consider offering the service as a package during the negotiation period, citing the need for customers with similar use cases.

To drive home the point, consider a real-life analogy beyond the scope of business. Say you had the misfortune of being diagnosed with type-2 diabetes. After a few months pass and you return for a check up, you don’t want your doctor to simply tell you they think your diabetes diagnosis will lead to significantly higher blood pressure and send you on your way. 

You want your doctor to either prescribe medication to remedy the symptoms, suggest you implement a medium-intensity cardio routine a few times per week or recommend a nutritionist to dial in your diet to reduce your consumption of refined sugars and carbohydrates.

The same goes for your organization and, more specifically, your finance function. As a trusted advisor, you are like a scientist testing and confirming a hypothesis. You must gather and structure your financial and operational data (something our FP&A Platform can do for you), synthesize the results, test assumptions and the possible outcomes, and prescribe measures to produce the desired results. 

Embrace The Bots, Don’t Fear Them

Artificial intelligence (AI) and the rise of popular machine learning chatbot applications like ChatGPT has the potential to revolutionize business operations in a similar sense that Google Search brought endless web information to our fingertips. In its first three months post-launch, ChatGPT amassed 100 million users, a record surpassing all other tools and applications.

The boom should come as no surprise as efficiency is the primary concern for many businesses and a focal point when expenses and COGS are scrutinized. For Finance, ChatGPT can be useful to: 

  1. Determine which Excel formula to use to extract a segment of a given data set
  2. Use historical data to forecast potential future outcomes
  3. Produce a range of possibilities based on adjustments to trends
  4. Provide descriptive analysis highlighting variances in budget to actual performance

Naturally with any new application, there is an innate disposition to resist, especially when the potential to be replaced becomes a fear for end users. The reality is, embracing technology and AI allows you to do more in a fraction of the time, giving you superhuman-like power to make an impact on your finance function in a fraction of the time.  

Like Batman conquering his fear, you must face your fears head on, and learn to use the tools surrounding you to your advantage. Dare we say, become one with the bots. However, keep in mind that machine learning and language tools like ChatGPT are only as accurate and informative as the data it collects. 

Chat With Caution

Let’s take a step back and ask, how did we get to this point? 

For years, professionals have tuned to Google Search, YouTube University, and professional forums to get quick answers to their most pressing matters. 

For example, computer programmers and developers rely on the popular forum Stack Overflow to figure out the syntax of a code or reasons for bugs and errors, among many other topics. Recently, Stack Overflow decided to ban the use of ChatGPT generated questions and answers on its forums. 

“Overall, because the average rate of getting correct answers from ChatGPT is too low, the posting of answers created by ChatGPT is substantially harmful to the site and to users who are asking and looking for correct answers.” - Stack Overflow

Chatbots make it incredibly simple to get an answer to a specific question in a simple interface with no distractions from pop-up ads or live chats.These tools aggregate a range of data points, identify patterns and translate that information into a relevant answer. 

The ability to deliver immediate answers in an instance makes it no surprise that ChatGPT has amassed over 100 million users in its first three months since launching in November 2022. Despite the popularity, there are many concerns related to privacy, data security, and the validity of information that remain. 

Samsung’s Snafu

In April Samsung employees revealed internal source code secrets while working with ChatGPT, which now also belongs to and is stored on ChatGPT’s database. Samsung, a conglomerate that did $245 billion in revenue in 2022, certainly has billions of reasons to be concerned about its future product roadmap and how much intellectual property is now available to OpenAI and potentially hackers and competitors. 

The leak lead to a company-wide generative AI tool ban at Samsung in May.

There is no question there are dozens of other examples like Samsung’s situation, and there is no telling just how many trade secrets and new product release details the chatbot database now holds. 

Italy’s Data Concerns

Italy became one of the first countries to ban use of ChatGPT after data privacy concerns surfaced along with the improper collection and storage of information. The temporary ban issued by Italian regulators required OpenAI to meet specific criteria to ensure its compliance in its management and adherence to privacy and data collection. On August 28, the Company announced it had satisfied the proposed issues and ChatGPT access was restored.

Although this matter was resolved within a month, ongoing political debates and use cases of the generative AI technology is yet to be established and any matters are subject to change among regulators.

Three Key Takeaways

  1. To beat the bots, and be a high-impact finance professional, focus on being a prescriber of solutions and strategies, answering the why and how when analyzing metrics and variances. 
  2. Leverage machine learning and automation tools to work more efficiently and deliver results faster.
  3. Although you should use technology to your advantage, be cautious of a chatbots' results and recommendations, as they are only as accurate as the data they can collect.