69 percent of companies use AI. 8 percent use it at scale across multiple areas. These two figures from the EY AI Readiness Check 2026 describe the state of digitalization in controlling more accurately than any trend statement. The tools are available. What’s missing is the process behind them.
This article categorizes digital transformation in controlling into five stages, provides real-world examples, and identifies the bottlenecks where projects typically get stuck.
What does digitalization mean in practice for controlling?
The term is often equated with purchasing software – but it actually refers to something else: the transition from manually maintained individual spreadsheets to a unified model in which planned, actual, and reported data are based on the same structure.
This transition occurs in stages that build upon one another – from centralized data storage through automated reporting and integrated planning to AI-supported management with forecasts and assistance functions. No stage can be skipped – and the actual benefit comes not from the tool itself, but from unambiguous definitions and clear accountability for approval.
Specifically, digitalization changes three things, only one of which is technical:
- Data storage, because numbers are now stored centrally rather than in different versions across multiple drives
- Workflows, because reports follow a fixed logic rather than being recompiled every month
- The role of the controller, whose focus shifts from data preparation to analysis.
The difference becomes apparent when the two are compared side by side.
Which metrics should actually be included in the report is a separate question. The section on KPIs in reporting provides some guidance on this.
What stages does a controlling function go through on the path to AI-driven management?
This tiered model has proven effective as a framework in our projects. It does not describe an ideal path, but rather the reality we encounter in companies.
The most common misjudgment concerns the leap from Level 2 to Level 5. It simply doesn’t work. Without standardized definitions, a model will reliably produce incorrect results. Metadata is key to the successful use of AI in this context.
Integrated corporate planning describes what Level 4 looks like in practice. The question of the planning direction is clarified by top-down and bottom-up planning.
What examples of digitalization in controlling are effective today?
Five use cases that we implement on a regular basis, along with their respective effects and prerequisites.
Example 1: Automated comments in the monthly report
The system detects discrepancies and generates a draft comment. The controller reviews the statement and its cause. This saves time on writing, but not on evaluation.
Example 2: Rolling Forecast with default values
Instead of manually updating the data once a quarter, a monthly forecast is generated for each account. This requires a history of at least 24 months with the same structure. This can also be achieved using rolling forecasts and AI-driven projections.
Example 3: Ad-hoc Analyses without an IT ticket
Departments ask questions directly of the model. This only works if the definitions are unambiguous; otherwise, two truths emerge.
Example 4: Planning in minutes instead of months
Scenarios can be calculated in a single session if the calculation logic is embedded in the model rather than in individual tables. A real-world example can be found in “AI-Driven Business Planning in Minutes Instead of Months.”
Example 5: Agents for recurring processes
An agent performs several steps independently, such as verifying data, generating reports, and reporting anomalies. This requires approval logic.
How do you get started with digital transformation in your own controlling department?
- Classify the current state into one of the five levels. Without this classification, every decision about which tool to use becomes a guessing game.
- Measure the effort required for two to three recurring tasks in hours per month. This figure will later serve as proof of the benefits.
- Refine definitions and master data for the area that will be converted first. This step is most often skipped and is the most expensive to make up for later.
- Implement one use case completely, rather than three halfway. “Completely” means: it runs as part of the monthly workflow, not alongside it.
- Establish a testing routine before the second use case begins.
- Build the team’s expertise based on the current case.
Which platform supports this depends on the planning component. IBM Planning Analytics addresses the distinction between planning and analytics tools in comparison to Power BI.
Define the next step, not the one after that
We’ll assess where you stand and name exactly one next step, with effort and expected impact. If that step isn’t a project, we’ll tell you that too.
What mistakes most often slow down digital transformation?
We see four patterns time and time again.
Tool before process: The software is selected before it’s clear which workflow needs to change. The result is a new system with old workflows.
Too many use cases at once: Three half-finished implementations yield less benefit than one complete one.
No measured baseline: Without a baseline before the start, no benefit can be demonstrated. This explains why 47 percent of companies cannot assess the economic value of their AI applications.
Digitalization without accountability: If no one is responsible for maintaining the definitions, the model falls apart within a year.
This leads to the second most important point of this article: Digital transformation in controlling is rarely a technical project and almost always a structural one. The companies that are achieving measurable results with AI in controlling today have successfully completed steps two and three. Those who skip this work are buying speed based on a data foundation that cannot support it.
Why BI2run takes a model-based approach to digital transformation in Controlling
We build planning and reporting models using IBM Planning Analytics and support finance teams in applying AI to management accounting. Through this work, we understand both sides of the equation: the model, which requires a clean structure, and the monthly process, where everything must come together in the end.
If our assessment shows that you need to address master data and definitions first, we’ll tell you so. This answer is less convenient than a project proposal – but it saves you more money.
Fact Sheet
- 54.4 percent of German companies were actively using AI in May 2026, up from 40.9 percent the year before. Industry leads at 58.7 percent. Source: ifo Institute, business survey, June 2026 survey.
- 41 percent of companies with 20 or more employees actively use AI, up from 20 percent the year before. 48 percent plan to deploy it. Source: Bitkom, April 2026.
- Only 31 percent of AI-using companies apply AI to data analysis. Text work is clearly ahead at 71 percent. Source: Bitkom 2026.
- 69 percent of companies use AI, 8 percent have scaled it across multiple areas. Source: EY AI Readiness Check 2026.
- 47 percent are unable to assess the business value of their AI applications, and 24 percent have initiated governance measures. Source: EY AI Readiness Check 2026.
Glossary
Frequently Asked Questions about digitalization in controlling
What should be digitized first?
The process that involves the most recurring effort and the cleanest data. This combination delivers the fastest demonstrable benefits and builds trust for the next step.
How long does it take to progress from Level 2 to Level 4?
In medium-sized companies, the timeframe is usually between six and eighteen months. The key factor is not so much the technology as the time required to refine definitions and master data.
What role does Excel play in the post-digital era?
Excel remains a useful interface, for example, for data entry and ad hoc reports. What is no longer needed is Excel as a storage location for planning data and as a calculation engine.
Why do so many projects get stuck at Level 3?
Because automated reporting significantly reduces the workload, thereby easing the pressure. The move toward integrated planning requires coordination across departmental boundaries and necessitates a decision by management.
When is it worth adopting AI features?
Starting at Level 3, once definitions are in place and reports are running reliably. Before that, AI automatically incorporates the existing inaccuracies.
How much does it cost to get started?
That depends on the level and scope. The question can only be properly assessed once the current number of hours spent on the process in question has been measured. This measurement costs nothing and changes the discussion.
