AI Forecasting 101: demand, revenue, scenario, and cash

Hey everyone, welcome to Finance AI Academy powered by OneStream. I'm Tiffany Ma, Director of AI Product Marketing, and today we'll be talking to Jack Allen about AI forecasting.

Hello everyone, I'm Jack Allen, Director of AI Demand and Enablement here at OneStream. Really excited for the conversation today. So Jack, what are some of the key use cases for AI and ML in forecasting? Can you give me some examples of what you've seen at customers? SensibleAI Forecast really supports a wide range of forecasting use cases across different industries, what we typically see is the customers start with areas where they've already seen some strong and reliable data. So think of things like revenue forecasting, demand forecasting, or unit sales.

Those use cases are usually the quickest wins because the data is dense, the patterns are relatively straightforward, and the impact on top line planning is immediately visible. Once organizations get those core forecasts in place, they almost always expand into additional areas. We see customers moving into cash forecasting, expense, and labor forecasting. . There are other operational KPIs that are unique to their industries as well that they look to forecast, kind of growing on those core forecasts there.

The real advantage of OneStream is that everything lives within the same platform, so it becomes very natural for customers to build on what they've already created. So Jack, you hit on something interesting, which is that the real advantage of OneStream is that everything lives within the same platform, so customers are able to build on what they've already created.

Can you expand on that? Many customers in OneStream have already built great planning solutions in their platform, so Sensible AI was created to enable the ML modeling to flow seamlessly into those planning solutions. This is really enhancing the ability for us to scale ML for finance. One of the biggest concerns we hear in the market regarding readiness to adopt AI is data quality and governance.

How much data do customers need, and how clean does the data have to be? Does OneStream help with data preparation? Data quality and governance are absolutely top of mind for organizations exploring AI, and the good news is that you don't need to perfectly have a clean data set to get started with Sensible AI Forecast. From a historical standpoint, we typically recommend the following guidelines for optimal performance.

For daily data sets, three or more years of history. For weekly data sets, five or more years of history. And for monthly data sets, eight or more years of history.

Almost no organization has a perfectly clean time series data set, and we don't expect them to. Our team reshape, cleanse, and transform data sets so that they allineati alle esigenze di SensibleAI Forecast. What matters most is forecasting at the right level of dimensionality. Many customers naturally want to forecast at the lowest level of detail, but if the data is too sparse at that level, it can actually hurt prediction accuracy and make trends harder to interpret.

OneStream provides native tools. So think of anomaly detection, data cleansing strategies, and transformation workflows that can help prepare the data efficiently. We help the customers get their data into a state where AI genuinely can drive better and more reliable forecasts.

What types of data are we typically leveraging for AI-powered forecasting? Is it just internal data, or does external data also come into play? For AI-powered forecasting and OneStream, we typically work with a combination of internal, operational, and financial data, along with external signals that can help provide broader context to the models.

along with external signals that can help provide broader context to the models. So most customers start by pulling data directly from their source system. So think of things like Snowflake, Oracle, SAP, or other ERP platforms. And while Sensible AI forecasts can run entirely on a customer's internal data set, many organizations enhance their models with external signals.

This might include macroeconomic indicators, industry trends, weather patterns, and even geographic holiday calendars. Internal data provides the foundation, and external data adds valuable context. Together, delivering a much more accurate and meaningful forecast. Switching topics now, what's the business impact of having faster, more accurate forecasts? What kinds of decisions or planning does this impact for customers?

There are a lot of different business impacts that come from faster and more accurate forecasts, but ultimately it comes down to making better decisions.

When organizations can trust their forecasts and get those insights earlier in the cycle, it changes how they plan, how they allocate resources, and how quickly they can respond to market shifts. For example, manufacturers use more accurate forecasts to optimize raw material purchasing. That directly translates to cost savings and fewer supply chain disruptions. Another big impact is reducing bias in the planning process.

When forecasts are fully data-driven rather than influenced by gut feel or overly optimistic assumptions, organizations get a much clearer picture of what's actually coming. This leads to more realistic plans and ultimately better business outcomes. We also see a huge boost in workforce planning. Companies can make sure that they have the right people in the right place at the right time. The theme is consistent.

Better forecasts leads to better decisions. And that has a measurable ripple effect across the entire business.

All right. Thank you so much, Jack, for taking the time to share your experiences and your insights with all of us today. Yeah. Thanks for the great questions and happy to be here.