The pace of technological development, regulatory amendments and changes in the market environment means that knowledge, skills and established ways of working can become outdated remarkably quickly. This does not mean that expertise has lost its value. Quite the opposite.
In the context of Artificial Intelligence (AI) Deep expertise is needed, among other things, to assess what AI produces, recognize its limitations and understand the context in which it is being used. But expertise can no longer be understood simply as accumulated knowledge. Increasingly, it also means the capacity to learn, question, and adapt, and many times also unlearn.
Learning is becoming part of the work itself.
Organizations often respond to technological change by providing training. AI training is certainly necessary. Article 4 of the EU AI Act even requires providers and deployers of AI systems to ensure a sufficient level of AI literacy of their staff and other persons dealing with AI systems on their behalf.
People need to understand what AI is and how it works. They need to know what AI is used in their organization and what its opportunities and risks are. They need to have a sufficient level of AI literacy to be able to use it responsibly, and what legal, ethical and operational risks it may create. But a training course is not the same thing as organizational capacity.
Is there a culture of continous learning?
No organization can train its people today for every AI development they will encounter tomorrow. The more important question is therefore whether the organization has created the structures, culture and incentives that allow people to keep learning as circumstances change. Is there a culture of continous learning?
This requires time and resources for learning, but also something less tangible: permission not to know. AI develops so quickly that even experienced professionals will regularly encounter technologies, terminology and applications they do not fully understand. Leadership cannot be based on pretending otherwise. An organization with a healthy learning culture allows people – including senior leaders and board members – to ask basic questions, challenge assumptions and acknowledge gaps in their understanding.
Learning also means unlearning.
The ability to say “I don’t know enough about this yet” is not the opposite of expertise. It can be an important expression of it.
Sometimes the more difficult challenge is not learning new things but unlearning old ones. AI is already changing how information is gathered, documents are produced, analysis is conducted, customers are served and decisions are prepared. Some established processes will remain valuable. Others may no longer make sense.
Organizations therefore need to distinguish between practices that exist for good reasons and practices that exist simply because this is how we have always done it.
This is how we have always done it is not a valid reason to keep doing so.
This is particularly challenging for experienced professionals. The more expertise we accumulate, the more deeply certain assumptions and ways of working can become embedded in our professional identity. Yet experience should not become an obstacle to curiosity. In a rapidly changing environment, expertise needs to combine accumulated knowledge with the willingness to reconsider it.
Unlearning does not mean abandoning professional judgment. It means being prepared to examine whether the assumptions underlying that judgment still hold.
This is where capacity-building becomes a governance issue. Good governance is fundamentally about an organization’s ability to exercise power, make decisions and pursue its objectives responsibly. That requires appropriate structures, clear responsibilities, accountability and access to relevant information and expertise.
Continuous learning is part of good governance.
If AI increasingly influences how an organization operates and makes decisions, those responsible for governing and leading the organization need sufficient understanding to exercise meaningful oversight. A board does not need to consist of AI specialists. Senior management does not need to understand every technical detail of a large language model. But they need enough understanding to ask relevant questions, assess the implications of proposed uses of AI and recognize when additional expertise is required.
The same applies throughout the organization. If employees are expected to use AI responsibly, they need more than access to tools and a policy telling them what not to do. They need the competence to exercise judgment: when AI can help, when its output needs to be challenged, what information should not be shared with it, when human intervention is necessary and when a seemingly efficient solution may create risks elsewhere.
Governance therefore cannot be separated from competence. Responsibility without sufficient capability is weak governance.
The objective is to build organizational learning capacity.
The objective should therefore not be to make everyone an AI expert. Nor should organizations chase every technological development simply because it is new. The objective is to build organizational learning capacity: the ability to identify what needs to be understood, acquire the necessary competence, share knowledge across functions, challenge outdated assumptions and translate learning into better decisions and practices.
This also means treating learning as part of strategy rather than as a separate HR initiative. Decisions about AI, competence, governance, investment and organizational development are increasingly interconnected. Capacity-building should therefore be considered alongside strategic planning, budgeting, risk management and governance rather than after those decisions have already been made.
There will never be a point at which an organization can declare itself fully trained for the AI era. Perhaps that is precisely the mindset we need to leave behind. In a world of continuous change, the most valuable capability is not knowing everything. It is the ability to keep learning and to build organizations that can do the same.