Decision transparency will become a new obligation for organizations.

AI can help or do a great deal of harm. And although this may sound slightly surprising, I believe that in the coming years many companies will discover that the hardest part is not making a decision about a person, but explaining it properly.

For years many decisions in organizations were made in a way that can be summed up in one sentence: “because that is what we decided”. And importantly, this discretion applied to promotions, pay rises, bonuses, recruitment, potential assessments, development, moves between teams and the ending of employment – processes that are hugely important for how a given team or department works and, in consequence, for the whole company.

Sometimes there were very good arguments behind that “we decided”. Sometimes a manager’s experience. Sometimes knowledge of the situation. Sometimes actual results. But sometimes also habit, subjective judgement, the pressure of the moment or inconsistent standards. For a long time organizations could live with that, but should it stay that way?

In my view that time is coming to an end. We are entering a new reality in which what matters increasingly is not only what decision we made, but also: on the basis of what data, according to what rules, by whom, with what justification and whether similar situations were treated similarly. This is especially important in HR, because decisions about people affect careers, pay, development, security and trust in the company.

In my work I have repeatedly seen situations in which the decision itself was rational, but the organization had trouble justifying it well afterwards. Why did that person get a promotion? Why was that pay rise higher? Why do two people in a similar role have different salaries? Why did one manager apply an exception and another did not? Why was a candidate rejected? Why was one person classified as talent and another not?

Often the problem was not that there was no answer. The problem was that it was hard to justify unambiguously to a manager. And here, fortunately, AI can help.

It can gather the data. It can point to the sources. It can compare similar cases. It can show inconsistencies. It can help prepare the justification. It can remind you about a required approval. It can leave a decision trail. Does that sound good? Many managers using askee stress how much easier this makes their everyday work.

I would like to share one more experience with you, though. There is a certain risk here too. AI works like a black box. If it generates recommendations without sources. If it does not distinguish facts from opinions. If it reinforces earlier errors. If it gives a manager a false sense of objectivity. If the organization cannot explain why the system suggested something. These are some of the problems we had to deal with when designing askee

The AI Act introduces particular requirements for high-risk systems. For such systems, human oversight is meant to prevent or minimise risks to health, safety and fundamental rights. This matters especially where AI can influence decisions about people, including in employment and people management.

On top of that comes the pay transparency directive. The European Commission indicates that its aim is to strengthen the principle of equal pay through greater pay transparency, and EU member states are to implement the rules by 7 June 2026.

This means transparency stops being a slogan. It becomes an organizational requirement. Companies will not only have to have policies. They will have to show they can apply them. Not only talk about fairness. They will have to be able to explain differences. Not only make decisions. They will have to document them.

And here a very important question appears: will AI increase the transparency of decisions, or create another layer of obscurity? Because if a manager says “the AI suggested it”, that is not a justification. It is rather a reason to ask further questions: based on what data? According to what criteria? Was the data current? Did the system take company policy into account? Did a human verify the recommendation? Were similar cases treated similarly? Can the employee understand the logic of the decision?

Mature AI in HR should not act like an oracle, but like a system supporting accountability – as Agnieszka Surowiec wrote recently. It should help a person see the grounds for a decision. It should show the sources. It should separate data from interpretation. It should point out risks. It should make auditing possible. It should strengthen human oversight rather than bypass it.

A real-life example: a decision about a pay rise. AI can help check the salary against the band, the history of pay changes, the level of competence, results, scope of accountability and a comparison with similar roles. But the final decision should belong to a human. And that human should be able to explain why the decision is justified.

A second, frequently encountered example: promotion. AI can help check readiness for the role, the competence gap, the requirements of the position and the history of results. But it should not automatically settle the promotion.

Or perhaps another one: identifying talent. AI can point out patterns, but HR has to watch whether it is not simply rewarding the most visible people, those rated best by particular managers, or those matching a historical pattern of success.

Transparency does not mean all decisions have to be identical. It means the differences should be explainable. This matters especially in pay. Two people can earn different amounts, but the company should be able to point to the reasons: experience, scope of accountability, level of the role, competencies, results, location, market, the history of the position. If it cannot, the problem does not start with communication. The problem starts in the decision system.

AI can help put such a system in order. That is exactly what askee does. But only when it is designed with explainability, sources, control and accountability in mind. Otherwise it can create something very dangerous: the appearance of objectivity.

In my view the appearance of objectivity is worse than open subjectivity. Because it is harder to challenge. That is why I believe decision transparency will be one of the most important management topics of the coming years. Not only because of regulation. Also because of employee expectations. People increasingly want to understand why a company makes particular decisions. They want to know whether the criteria are clear, whether the process is fair and whether similar cases are treated similarly.

My main message is this: AI can help with that, but only if it does not hide decisions behind technology. I wonder – could your organization calmly and specifically justify its most important decisions about people today?