AI and ML provide some unique forecasting capabilities that are exciting and have the chance to change your planning process completely. But, we need to update the forecasting and planning process to address their strengths and weaknesses. For these highly capable forecasting tools to be successful, we need to be aware of the distinction between a budget, a rolling forecast, and a continuous planning process.

Budgets Are Not a Forecast

When I start working with a new business on forecasting, I always ask for their existing forecast. Usually, I am handed a budget when I ask for the forecast. Budgets contain forecasts, but they also contain plans, consensus, compromises, and directives. I call this Budget a Forecast or “BaaFling.”

That undefined soup of stuff is causing issues. An AI/ML forecast is not going to take over planning, consensus, and decision-making, despite what the hype and pontification about AI is saying right now. At least, not with the current iteration of LLMs.

Plans Are Not a Forecast

When we have a manual forecasting and planning process that humans run, it is easy to brush over the fact that planning can happen while we are building a forecast. When we upgrade to automated forecasting, the lack of human input is often crippling. So it’s important to recognise these as two distinct things.

At the pinnacle of planning, the continuous planning process, we can have the same issue. A monthly rolling forecast is not a continuous plan. It’s easy to let the rolling forecast get bogged down as we plan. However, a huge key to being agile and allowing forecast automation to shine is to decouple the planning phase from the forecasting phase. But, to do so, we need to update our definitions, expectations, and the process we follow.

What Is a Forecast? – Definitions

For the sake of clarity, I’m going to define several terms and the distinctions of how I’m using those terms. I quite often hear people use the term forecast to encompass the outcome of a continuous planning process. This is the norm for publicly traded companies and their investor calls. However, that lack of distinction will be one of the primary causes of failure for AI/ML forecast automation and falling short of what these technologies are capable of. So, let’s sharpen the pencil on some definitions:

1. Target: The agreed-upon target for the year, often expressed as a revenue or profitability number.

2. Plan: The plan or agreed-upon steps to get to the Target.

3. Plan Expectation: What do we think the outcome of the plan will be? The business change that will overcome the current business trend momentum or market drivers.

4. Forecast: The most likely outcome, given current trends and drivers, EXCLUDING the Plan and Plan Expectation. The exclusions here are the key to ML forecasting success.

5. Driver-Based Forecast: This is a forecast that includes outside market drivers. FP&A Trends does a fantastic job of communicating the importance of a driver-based forecast.

6. Machine Learning Forecast: Automated machine forecast. An AI/ML-based forecast can be considered an advanced driver-based one.

7. Continuous Planning: The combination of a rolling forecast and the development of a new plan in response to that rolling forecast.

8. Budget: The annual consensus on the chosen combination of Targets, Plans, Plan Expectations, and Forecasts. Sometimes, it is a snapshot of a continuous planning process.

9. Annual Operating Plan (AOP): A new way to say “budget” that acknowledges that there is a plan within the Budget. Getting to the plan is the point of the annual budgeting process.

10. Baffling: Budget as a Forecast. A recipe for failure.

Sample Process Walkthrough

To clarify this, let’s walk through a fictional company that has implemented an annual budgeting process and a monthly rolling forecast process with selective planning.

2025 Annual Operating Plan (Budget)

Leading up to the 2025 annual budget, we have 2024 actuals, which we use to establish the most likely outcome using trends analysis. In its simplest form, this would be the expectation that if we change nothing, 2025 will look an awful lot like 2024. But we have decided to set up a target of 10% growth. So, the 2025 “Budget” revenue number would be 110% of 2024. At the time the Budget is completed, the plan expectation and target are the same.

As shown in Figure 1, the 2025 revenue target is established during the annual planning process (aka budgeting). 

Figure 1

Q1 Results Analysis

When we get Q1 2025 results, they show that our plans were only 50% effective compared to our plan expectations. We are only seeing a 5% increase in revenue over 2024 instead of the expected 10%. Now, this is where things get interesting.

The most common thing reported by finance is the 3+9 number, which would be 3 months of actuals plus the remaining 9 months of Budget. This number is easy to calculate, but it is very misleading. We have to recognise that the plans to increase revenue by 10% are not meeting expectations. They’re only 50% successful. So reporting 3 + budget 9 doesn’t tell us anything useful in terms of end-of-year outcomes. A 3 + forecast 9 (rolling forecast) will capture that. It will say that the most likely outcome will be 2024 trends plus a 5% lift due to the plans we implemented during budgeting, meeting 50% of their plan expectation.

If we have a proper rolling forecast process (which has this critical distinction between forecast and plan expectations), then we will be able to report 3 + forecast 9, where the 9 is the forecast, not the Budget. This version of 3+9 is the most useful to management because it says where we’re most likely to end up without any changes to the plan.

Figure 2 illustrates the difference between reporting actuals plus forecast versus actuals plus Budget when compared to 2025 targets. This visual highlights how rolling forecasts provide a more accurate picture of future performance, especially when plans fall short of expectations.

Figure 2

Are We Going to Update the Plan?

If we do nothing, then we are going to miss targets. We can’t change the Target. We now have a decision to make. Do we need to change the plan?

This is a decision we make every month. The continuous planning process says we are always going to update the plan. In this selective planning process, we can choose not to make a change if we’re OK with the current expected outcome, or we can choose to go into the planning phase.

Part of business partnering is to present a clear picture of management effectiveness. The plan isn’t working, but because we highlighted this early, management was able to effectively update their plans and get back on track to meet targets. Bad news early is good news. We have time to fix it.

Tip: Make sure you have a distinct name for the new plan. This creates clear communication that the new plan is distinct from the initial Annual Operating Plan, even though the 2025 Targets have not changed.

Figure 3 illustrates the new Plan to hit targets. This new plan was enabled by FP&A reporting the 3 + forecast 9, which made the gap to target very clear.

Figure 3

Monthly Rolling Forecasts

This is where FP&A can truly shift from reactive to proactive. Rolling forecasts offer the fastest and most precise view of what’s likely to happen next.

Let’s not underestimate the power of this. By decoupling the plan from the forecast, we not only free up time, but also unlock the full potential of AI and Machine Learning. These technologies thrive on fast updates and large volumes of driver-based data, not on layered human assumptions embedded in annual budgets.

Figure 4 presents the ideal situation where the rolling forecast is updated monthly. This view becomes a constant, flexible guidepost for management, not just a projection but a tool for identifying gaps and enabling timely planning discussions.

Figure 4

Closing Summary

In closing, the strengths of AI/ML are best suited to fast, continuous updates of a forecast, but they are not able to do the planning for us. By setting up the process to take advantage of AI/ML’s strengths and decoupling the part that takes time, we can produce high-value, high-capability data, improving overall business planning and performance.

To make this work, we need to rethink our definitions, align expectations across the business, and update how we integrate forecasts and plans into day-to-day decisions.

Start by asking this simple question: Is your forecast truly a forecast, or is it still a budget in disguise? The answer could be the difference between agility and inertia.