Intelligent Forecasting, Return on Data, and KPIs in the Age of AI

Digital transformation and data go hand in hand. Most executives get this: today, the vast majority view data as a critical asset to be tapped, and new AI capabilities have made this data more beneficial than ever, says Michael Schrage, a research fellow at the MIT Initiative on the Digital Economy.

But here’s the catch, Schrage asserts: Most executives can’t name the five most valuable data assets within their organization, or their firm’s return on data (ROD). That’s a problem, he says, because if a business isn’t paying close attention to the strategic value of its data, making the right AI investments will be much harder. And as the age of generative AI and machine learning (ML) takes off, the potential ROI on those investments for CFOs and financial planning and analysis (FP&A) teams is massive. Achieving those returns requires merging and marrying your most valuable data with AI.

The ways AI can power a paradigm shift in forecasting and metrics are coming into focus, Schrage says. Trained on customized data sets, large language models (LLMs) can turn KPIs into intelligent mechanisms that help create new value, instead of just tracking and protecting value. And with ML supporting FP&A activities, forecasting can become a source of dynamic insights to support operations and workforce planning—far more than just an accurate financial picture of the future. KPIs become tools for better key performance insights and key performance investments.

In other words, “The generative revolution changes everything,” Schrage says.

Future-Ready Metrics

With data now a lifeblood connecting and defining the digitally transformed organization, metrics have become increasingly important for value protection and value creation. Schrage’s most recent research focuses on combining generative AI, ML, and KPIs, asking whether KPIs can evolve beyond being simple measures and instead become software agents capable of learning.

This idea is no longer a hypothetical, given dramatic developments in generative AI over the last few years. Aided by AI, metrics can become forward-looking tools to support forecasting and scenario planning activities, helping finance leaders be more proactive and strategic. 

Using ChatGPT, you can actually ask the KPI questions that matter most, Schrage says. “What could we do to improve you? Is the data helpful to you? What new data would improve your forecast? What would you say under this circumstance? Do you think this scenario would be good for you or bad for you?” 

The bottom line: AI can help the finance function drive value with more predictive and future-oriented KPIs, whether it’s classic metrics such as revenue, profit, and sales, or increasingly important hybrid financial custom-facing metrics such as customer lifetime value or customer churn rate.

“What if your customer lifetime value KPI could talk to your customer churn KPI? The math of that kind of thing is interesting,” Schrage says.

Data + AI = Intelligent Forecasting

The potential value of predictive insights unlocked by AI is huge. So long as LLMs are trained on quality data—bad data is a major stumbling block for organizations in adopting AI, a global Workday survey found—finance leaders have the opportunity to rethink what forecasting can and should be. 

“Is a forecast about being correct and precise, or is it about being a source of insight and conversation for how the business should prepare and prioritize, go to market, or respond to customer needs?” Schrage asks. With custom LLMs or ML in the mix, forecasting becomes a richer, more cost-effective way of gaining actionable insights for the organization.

Take Team Car Care, for instance. As the largest franchisee of Jiffy Lube, the company operates more than 500 outlets across 26 sites. A key business question it faces is: How many people are going to show up at its Jiffy Lube locations today? To find answers that inform sales and workforce planning decisions, Team Car Care’s planning team created a custom forecasting model within Workday Adaptive Planning, leaning on its ML capabilities.

As a first step, the team developed an invoice data feed into Workday Adaptive Planning detailing the date, duration, and location of each vehicle serviced, as well as what customers purchased. It also brings in historical location-specific weather data. 

“We know that rain or snow changes customer patterns for oil changes,” says Matt Castonguay, senior vice president of finance, analytics, and supply chain at Team Car Care. “We designed the model to normalize data to make forecasts more meaningful.”

Today, as this forecasting model keeps “learning” from the latest data, the company can confidently predict how many vehicle bays it should have ready for customers at each location, given the day of the week, time of day, and weather conditions.

When it comes to scenario planning, Schrage sees multiple potential forecasting use cases emerging via generative AI apps built atop customized LLMs. Expect to see emerging AI/LLM ecosystems (such as Hugging Face, LangChain, and OpenAI) allow businesses to connect data streams from platforms such as Workday with generative AI systems to populate scenarios with real data and defined parameters, he says. 

“The ability to interconnect, to create interoperability between Workday and specific scenarios—it could become really meaningful, something that could be part of a compliance stress test for one’s FP&A initiatives,” Schrage adds, observing that the costs for these kinds of analytics are coming down.

Starting Points

The scope and purpose of a custom forecasting model or scenario will vary business to business, of course. But all organizations should start their AI and ML planning journey from the same place, Castonguay says. 

“You have to start with your business needs,” he adds. “What problem are you solving for? Figure that out and go from there.”

Data can be another entry point, Schrage adds. “If data is an asset, then what are your most valuable assets—and how can you get greater value from them?”

Answering that question is made harder by data silos, which remain common. Nearly two-thirds (59%) of organizations report their data is somewhat or completely siloed—and just 4% say their data is fully accessible, according to the latest Workday global survey. Those numbers need to change.

To boost accessibility and value extraction in the age of AI, Schrage suggests the finance function needs to step up and take charge of data.

“For the future of capital allocation, CFOs should be first among equals,” Schrage says. “They should be the drivers of change.”

That makes sense considering that finance—once slower to adopt AI than other functions—is gearing up its AI activity. Gartner® predicts that by 2026, 80% of large finance teams will rely on internally managed and owned generative AI platforms that have been trained on their own proprietary business data.

Watch an on-demand video of the entire conversation that took place at Workday Rising.