AI is the Best Thing for Controlling Since Excel, but There is a Trap

written by: Jasmina Očko, consultant for controlling – Controlling Kognosko

To write this article, I went back exactly one year and read my article ‘Smarter Together – How People and AI are Now Building the Future Together‘, published in Lider in October 2025, to check myself – was my view on the topic of artificial intelligence in controlling realistic at that time and, above all, did it predict anything. And I can happily say – it certainly did. I endorse everything said then even today, only with additional, significantly greater experience in implementation. Now my main question is: what have we learned so far, do we know what we need to do, are we smarter today?

There is probably no one left who is not aware that technology is advancing faster than humans can keep up. In fact, it has never been about technology – it has always been about people. Artificial intelligence has entered controlling, that is, management tools, through the front door, without knocking and without asking if we are ready for it. It is normal to expect that controlling embraces new opportunities and utilizes them as best as it can.

What AI Can Really Do

It has long been struggling with an ever-increasing influx of data and continues to relatively slowly and laboriously convert it into wisdom. There are many reasons: there are still many different data sources, along with manual processing, poorly implemented ERPs and BI systems. Employee turnover in controlling, rapid company growth that is not accompanied by organized processes, etc., do not help either. The list of reasons is long.

The emergence of artificial intelligence in the domain of controlling now opens up the possibility of eliminating some of the aforementioned barriers. AI can much more easily merge data, model it, and automate processing. It can also automate data entry with much stricter garbage control, analyze them much more easily and quickly, and produce comments. It would be a great help if we could use artificial intelligence to identify deviations earlier, explain causes, assess risks, and offer scenarios to management.

This is a great opportunity for controlling to truly revolutionize. However, it is also an opportunity for us to mess things up once again. The question is not whether artificial intelligence can speed up and automate part of the controller’s work. The question is whether controlling, thanks to artificial intelligence, can be a better advisor to management. The answer is: it cannot if we do the same things we did before artificial intelligence.

What Needs to Change First?

The mental framework. Controlling needs/can/must be the ideal partner for management that will prepare the foundation for artificial intelligence, because no one, absolutely no one in the company knows so well and precisely where the garbage is, where processes break down, where there are no rules and procedures, where water flows as it finds its way. Controlling is the ideal partner for changes towards AI, but only if those who make decisions understand that.

What are the Correct Steps

We can consider the moment when management really starts to seriously engage in defining the business development strategy and business model as a prerequisite for the strategy and a well-developed and prepared plan for digital transformation as the beginning. AI is just the icing on the cake afterward, the grand finale as a self-evidence, not a starting point, which often happens in practice.

Here is an example that illustrates this: not long ago, a heated discussion arose on LK with many different comments on what company owners with excess profits will do now that AI tools automate everything and when productivity rises to a significantly higher level, i.e., when significantly fewer employees will be needed for the same job. It is hard for me to say where to start analyzing what is wrong with that sentence. So let’s go in order.

1. Step: Purpose Analysis. To create a positive effect, we must first automate segments of the business and processes that bring newly created value. If we are improving something that is already burning money without effect, automating such a process will not change much. For example, a company introduces AI in procurement to automate the ordering process without analyzing the procurement process itself and asking why they have four hundred suppliers with whom they maintain contacts, why for half of the procurements they do not have contracted suppliers, and why they do not return excess purchases to the supplier even though it is agreed. Or another example: in controlling, reports are created that management does not read and does not act upon. Speeding up reporting will not solve the problem; it will only bring reports faster that no one will read.

2. Step: Investment Analysis. If processes that truly create new value are automated and we genuinely benefit from it, we must first compare the effects with the total investments in the AI investment. And no, we are not talking here about the cost of a monthly license. More serious AI projects require knowledge to be developed to measure: they need a significant tidying up of business processes, gathering and consolidating data from different systems, connecting with ERP, CRM, DMS, HR systems, and other applications; AI regulatory and information security consultants are needed, development and testing of user interfaces and automated workflows. Additionally, there are fees based on the number of calls, tokens, or the amount of processed data, training, and general raising of the competency level of a broader circle of employees, and much, much more. We can also mention that during this implementation period, productivity will fall, that time is needed for learning and adaptation, that increased challenges in communication and interpersonal relationships will arise (because not everyone accepts changes equally well). It is also possible that some job titles, scopes of work, and responsibilities will change.

3. Step: Competency Analysis. To reach the holy grail of automation, we cannot (actually we can, but it makes no sense) automate only individual, randomly selected fragments; it must be a rounded story. And for that, strong competencies are needed, primarily from the leading people or, more precisely – the first person in the company. If they do not have those competencies, there will be no meaningful development of artificial intelligence.

These are just some of the challenges we face in the upcoming periods of business automation on the wings of artificial intelligence. The fact is that AI in the field of business management and controlling is the best thing that has happened to us since the advent of the first spreadsheet calculator, Excel, but we really need to be fully aware of the prerequisites for all of this to truly come to life.

Demystifying Myths

Not long ago, a heated discussion developed on LinkedIn about another topic: now that AI is here, we can start thinking about a four-day work week! The only answer to that question would be the following: if your company can realize newly created value in four days of work that can finance all stakeholders for the whole week (with a special emphasis on employees and their decent standard and on the profit needed by the owner for further growth and development) – you are a candidate.

If you cannot do that, you must work five or six or even seven days a week (this is best seen in small and micro-entrepreneurs, who have not had days off, weekends, and holidays for years). And this again has little to do with artificial intelligence. Exceptions aside – if you are an architectural firm that has an AI tool that draws a complete house in half an hour and sells that project for a fantastically high price, then you might start resting on Thursday. But realistically, many domestic businesses achieve (by European standards) minimal profit with full-time work. Theoretically, automating a large number of processes could help finally raise average profitability to the usual average of developed economies in a five-day work week. If we achieve that, it will be fantastic.

AI Only Estimates Outcomes

Artificial intelligence operates on the principle of probability, estimating the most likely outcome based on a sea of data. This is not mathematics with a calculator, and therefore the results can be inconsistent, imprecise, or even incorrect. Therefore, artificial intelligence should not be treated as a calculator or a final decision-maker. Together with a group of colleagues within the ICV International Controller Vereina (International Association of Controllers), I worked on a project related to AI and we conducted an experiment: the four of us from four different countries used the same data and the same instructions for AI on four different AI platforms. We received quite different responses. This is an extremely useful exercise that I recommend to everyone – it will make it much clearer how it works.

However, instead of being disappointed, we looked at it pragmatically: if you have four controllers, even in the same company, and you give them the same data and the same task, a similar thing will happen. The interpretation of business results is absolutely never just mathematics, just data. That is the problem; if it were not so, it would be easy. Each controller draws their conclusions based on their experience, mental framework, knowledge, and beliefs. We should expect the same from artificial intelligence.

Beware of ‘AI Expertise’

One should beware of the false sense of expertise that now widely popular tools like ChatGPT, Claude, or Copilot truly provide. By using such powerful tools, a person suddenly feels that everything is simple and logical, just much faster than they would know themselves.

It is not easy to lose sight of reality, especially when we talk about young controllers who, objectively, have not yet acquired real knowledge. An unskilled person can easily gain a sense of expertise. We must always be aware that conversations with a machine are based on limited segments of business; the data is fragmented, most often without context and the big picture. The controller, on the other hand, must understand the company’s history, the dynamics between departments, unspoken rules, and tacit agreements that are not in the system.

All in all, we can look forward to the future in terms of technology that launches controlling to the sky, but now all the strength must be directed towards developing the awareness of the entire company, especially those who decide – that through organizational development, culture, learning and developing new competencies, data management, awareness of processes, and continuous improvement of all this, we must systematically engage every day. This cannot be skipped – artificial intelligence will not solve this, it will not disappear. And controlling is the best business partner that can help a lot in this.