AI should not replace managers

AI should not replace managers. It should end the era of decisions made “on a hunch”. That is my view, based on the project I am currently working on – askee – an AI assistant for managers. I would like to share my thoughts with you here:)

In my view the biggest mistake would be to judge AI in management by whether it will replace managers. The real question is: will it help them stop making decisions in a state of permanent under-information? Over many years working as a manager and on rollout projects I have seen one very repeatable situation: managers do not want to make decisions “by gut feel” at all. They very often simply have no other option.

A decision about an employee, a team, a budget, priorities or an organizational change is rarely made in ideal conditions. Usually there is time pressure, incomplete data, several conflicting opinions, expectations from above, budget constraints and the awareness that every decision will have consequences. And then the manager does what they do best (this was also my case): they combine experience, intuition, memory, conversations with people, the latest reports and their own read of the situation.

And to be clear – there is nothing wrong with that. The problem only begins when intuition becomes the main management system. Because intuition is valuable, but it can be selective. Memory is useful, but it can be unreliable. Experience matters, but it can entrench patterns. And time pressure means a manager often chooses not the best decision, but the decision that can be made here and now.

That is why I believe AI should not be presented as the “manager’s successor”. That is a very shallow and, fundamentally, harmful simplification. A manager is not merely a decision-making machine. They are a person accountable for context, people, relationships, tensions, consequences, communication and responsibility. AI should not and cannot take that over. What AI, as a genuinely innovative technology, should do is end the era of decisions made without sufficient context.

A good management decision very rarely follows from a single indicator. It emerges from a combination of several things: goals, results, the history of actions, competencies, risks, constraints, company rules, earlier cases and possible consequences. The problem is that it is very hard for a person to gather all of that quickly, reliably and without omissions.

And this is exactly where I personally see the role of AI. Not as an autopilot. Rather as an intelligent “second look” at the situation that helps with many things. AI can help a manager see what they have not taken into account. That was also a conclusion from the conversation I had the chance to have with Łukasz Jarota. It can point out an inconsistency in the reasoning. It can recall earlier decisions in similar situations. It can show risks that usually only surface later. It can help prepare several scenarios and the consequences of each. In short, this is a completely different conversation from: “will AI replace the manager?”.

A more appropriate question is: will a manager supported by AI make better decisions than a manager who has to reconstruct everything from memory? It is worth looking at the current reports too – the Microsoft Work Trend Index 2025 shows that companies are starting to move from simply using AI on individual tasks towards the concept of the “Frontier Firm”, meaning organizations in which people and AI agents work together. According to the report, 24% of leaders said their companies had rolled out AI across the whole organization, while only 12% were still at the pilot stage. This shows the question is no longer “will AI appear at work”, but how deeply it will change the way we manage and work together. I am convinced this trend has only deepened in 2026.

At the same time I see a significant risk. Many companies may treat AI as a way to increase the pace of work rather than the quality of decisions. And those are two different things. A manager may get a summary of the situation faster and still make a bad decision. They may prepare an e-mail faster without solving the problem. They may generate an action plan faster without noticing that the plan does not match the team’s reality.  That is why the real value of AI in management does not lie in acceleration alone. It lies in improving the quality of judgement.

In practice I see several areas where AI can be particularly helpful to managers. First: preparing for a decision. Instead of starting from a blank page, a manager can see an organised picture of the situation, the possible causes of the problem, data confirming hypotheses and the areas of uncertainty.

Second: testing assumptions. AI can ask a manager a question they did not ask themselves: “are you taking into account that the drop in results may come from overload rather than a lack of motivation?”, “have you compared this person with similar roles?”, “does this decision not set a precedent?”.

Third: simulating consequences. A good decision is not only a choice of action. It is also an awareness of what may happen once it is implemented. AI can help analyse the options: quick intervention, a development conversation, changing the goals, additional support, shifting tasks, a personnel decision.

Fourth: documenting the justification. This is especially important in HR, pay, promotions and corrective action. A manager should be able to explain why they made a decision, not merely state that “that is what they decided”.

The IBM Institute for Business Value study of CEOs in 2026 carries an important signal: 64% of chief executives said they were comfortable making important strategic decisions based on inputs generated by AI, and 83% agreed that AI is changing roles on boards and in top management. This shows that AI is moving ever more firmly into the area of decisions, not just task automation.

For me this is both an opportunity and a warning.

An opportunity, because managers can get a tool that helps them act more consciously. A warning, because it is very easy to cross the line between supporting a decision and shifting accountability onto the system. And that is a very real danger – some managers may uncritically accept every AI recommendation, convinced that it is the simplest and at the same time the best way to make a personnel decision. That is exactly what I discussed recently with Łukasz Sowiński – AI, which will certainly produce a good qualitative analysis, cannot be the sole source of a management decision.

Why? Because accountability should stay with the human. Not because the technology is weak. Because management decisions concern people, priorities, conflicts of values and consequences that cannot be reduced to calculation alone. AI can help a manager see more. But it is the manager who has to decide what to do with it. That is why, for me, mature use of AI in management is not a manager asking: “what should I do?”.

It is asking: “what are the possible interpretations of this situation?” “what else do I not know?” “what are the risks of my decision?” “are my assumptions consistent with the data?” “how can I justify this decision better?” “what should I monitor after implementing it?” – these are some of the questions that, in my view, should be asked.

That is a different level of work. It is not about replacing the manager. It is about raising the quality of managerial thinking. This is the key aspect worth paying attention to. And that is precisely why I believe the biggest change is not that AI will make decisions for managers. The biggest change is that managers will no longer be able to justify decisions by a lack of access to data, when that data can be analysed from the perspective of the full situational context.

Because if technology can help gather data, show the options and point out the risks, then a decision made “on a hunch” stops being a necessity. That gives us entirely different possibilities and defines an entirely different responsibility.

A question I think is worth putting up for discussion: do managers in your organization really make decisions based on the full picture of the situation — or have they simply learned to act despite not having it? Give it some thought, or better still – talk about it with your managers, and if you are a manager, do some self-reflection in this area:)

And here a very important role for HR appears too – not as the department responsible for HR processes, but as a source of support for managers. What I am proposing, though, is not the traditional operational support that HRBPs often provide to their business units today, but giving managers a tool for working independently that lets them make the highest-quality personnel decisions in their own teams. In my view HR is increasingly accountable for the quality of decisions about people across the whole organization. And it should be, although I sometimes get the impression this is forgotten.

Questions such as: Who should we develop? Who should get a promotion? How do we justify a pay rise? How do we spot the risk of someone leaving? Is the problem the employee, the manager or the way work is structured? Are the pay differences justified? Are managers’ decisions consistent? Is company policy applied in the same way in similar cases? These are the questions that really determine the quality of HR. Questions, or perhaps more the answers…

Over years of working with clients I saw very similar problems, or perhaps I should write challenges. Companies had HR systems, data, reports, procedures, development paths and policies. But when a specific decision about a person came up, suddenly the picture of the situation had to be assembled by hand. The HRBP asked the manager for details. The manager looked for the history. Someone exported data. Someone checked the rules. Someone recalled an earlier case. And the decision was still often made under time pressure.

And this is not about HR not doing a good job. It is that many organizations do not have a good system for supporting decisions about people. They have very extensive resources in the form of processes, procedures, reports and policies, but no easy way to combine them into one practical recommendation. This is exactly where I see the real role of AI in HR.

Not as a “magical recruiter”, nor as a tool that assesses people on a human’s behalf. Not as a system that independently decides on a promotion, a pay rise or a dismissal, but as a layer that helps HR and managers better understand situations involving people.

In its Global Human Capital Trends 2025 report, Deloitte writes about the tensions organizations have to resolve in the areas of work, workforce, organization and culture. The report stresses the need to build “human performance”, meaning an organization’s ability to connect business results with people’s potential. This matters a great deal, because it shows HR can no longer function only as an administrator of processes. It has to help the organization make decisions that affect both the result and the people. For me that is the heart of it.

Let us imagine a few situations:

A manager says: “I want to give this person a promotion”. Well-designed AI should not answer “yes” or “no”. It should help check the criteria for the role, the results, the competencies, readiness for the next level, the development history, the risks, comparable cases and the promotion rules.

Another situation: HR analyses the risk of people leaving a team. AI should not produce a list of “people to retain”, as if people were records in a database. It should help see the signals: a lack of development, overload, pay inequalities, no conversations with the manager, falling engagement, changes in the structure, a history of unaddressed problems.

Another situation still: a company is preparing for greater pay transparency. AI can help HR organise the criteria behind pay differences, check the consistency of decisions, gather data for justifications and point out the places where the organization does not have sufficient documentation.

Would you agree with me that these are the real applications of AI in HR? Far more important than generating a job description.

But of course there is another side to this. AI in HR has to be designed with particular responsibility, because it concerns people. The EU AI Act classifies many AI applications in employment and workforce management as high-risk. This covers, among others, systems used for recruitment, selection, assessment, decisions affecting working conditions, promotion, ending employment or allocating tasks.

That is why a conversation about AI in HR cannot be a conversation about productivity alone – it always has to be a conversation about accountability too. AI in HR does not release HR from thinking. On the contrary — it requires even greater maturity and proactive involvement in the process of bringing solutions based on this technology into the organization.

That is precisely why I believe the future of HR is not that AI will take over HR processes. The future of HR is that, thanks to AI, HR can finally stop being merely a supplier of reports and procedures and become a partner in making decisions about people.

I invite everyone to the discussion, and especially to assess my way of thinking. This article is exactly the basis on which askee was created as a solution addressing challenges of this kind while using the latest technology based on AI.