And starts with the board. How can companies implement artificial intelligence before everyone does it? [WYWIAD]

Jan Michalski, partner, Gena leader in Central Europe in Deloitte: Based on our experience of working with clients, it is worth starting with two elements. The first is to describe the internal policy of artificial intelligence (Ai Governance) – so as to ensure compliance with regulations, transparency, but also to ensure the credibility of the models used.

The second step should be the analysis of the value chain and key processes. It is necessary to identify places where artificial intelligence and automation can bring the greatest benefits.

On this solid basis, it is worth developing a coherent strategy for the implementation and functioning of AI in an organization – balanced between the pursuit of efficiency and the ambition of innovation, strongly rooted in specific business needs, and not based solely on technology itself.

It is also necessary to early inclusion in these works of various functions in the company: from HR and operations, through risk management departments, to IT – so that the transformation is comprehensive, real and systemic. With this approach, AI will not be a one -time project, but a stimulus for a permanent change in the operating model.

On a global scale, more and more boards are taking the topic of AI. At the same time, from this year's Deloitte report “Governance of Ai: A Critical Imperty's Word” it follows that 66 percent companies still declare limited knowledge in this area, and 31 percent He does not treat AI as one of the key topics under strategic supervision. This shows that leaders who are already investing in building a solid frame and strategies can gain a clear competitive advantage before the topic becomes a market standard.

There is an insufficient AI implementation rate in Europe. What can you do to make specific decisions made at the management level are made faster?

There can be a involvement on the side of the highest management. As in the case of business transformation, initiatives related to artificial intelligence will be implemented faster if the presidents of companies and members of the board are actively involved. It is also important to equip leaders at all levels of organization with appropriate knowledge about AI and give them a clear mandate to implement these solutions. Without competent and motivated leaders, it is difficult to break the inertia and go from pilot to scalable implementation.

We also see that the best results are focused on specific use of artificial intelligence, those with a clear return on investment. It is worth building the first successes on them, which will convince the organization to the value of AI scaling.

How to implement ai at home

How can the management board determine AI's risk appetite and balance it with pressure on innovation and time-to-market?

It is necessary to consciously manage the portfolio of AI initiatives and determine clear criteria for acceptable risk. It is worth determining what level of risk we are as a company ready to accept in projects with high innovation potential, and where more conservative actions are needed.

The key is balance. In addition to initiatives carrying low risk and quick return, and which at the same time build trust in AI, there should also be more innovative projects, with a greater risk, but giving a chance for a significant added value. Let's not forget about the already mentioned solid framework of AI Governance and transparent risk escalation procedures. This is particularly important, because, according to our report, as much as 70 percent. The organization provides that building such mechanisms and dealing with challenges regarding management of AI will take them at least 12 months.

Thanks to this approach, the Management Board can assess the impact of AI on the organization on an ongoing basis, react to emerging threats and at the same time not inhibit the innovation where the risk falls within the adopted limits.

What practical steps help to implement and monitor from the level of the Framework Council of Ethics, Safety and Compliance in the context of AI?

I think that the most important is the establishment of the AI Governance structure, the implementation of politicians regarding data, compliance and responsibility of models. It is a good idea to create the so -called AI Council, i.e. a team with the participation of business representatives, HR department, lawyers and technology experts, so as to monitor the topic in a comprehensive way.

Directors for artificial intelligence are welcome? Not necessarily

Is the appointment of a single “AI-Director” is enough or do you need to increase the technological competence of the entire board and directors?

AI is not a single function, it is the basis and direction of broader transformation. It is necessary to raise the competences of the entire management board to understand the potential of AI in all business areas such as operations, customer service, risk and strategy. Otherwise, artificial intelligence can become the domain of a narrow group of experts, not an element of the DNA of the organization.

The role of “AI-Director” should be to accelerate the organization's transformation by supporting and educating the entire management and joining individual teams, not lonely artificial intelligence management in isolation from the rest of the business.

What forms of management board education develop the ability to think about AI best?

Strategic workshops with identifying the impact of artificial intelligence on the value chain and the operational model are definitely working, but practical sessions, for example, building AI agents and simulations with their use will also be irreplaceable.

Board members should also get involved in design work to understand well what can be achieved using this technology in the organization now.

Remember, however, that AI is not only the domain of the board. In the most advanced organizations in terms of implementation, leaders consciously equip employees with AI tools. Almost half provide them with at least 40 percent. your staff. This approach allows not only to develop competences faster, but also to build an innovation culture, in which employees discover new technology applications and become its ambassadors.

Which organizational factors most often block the transition from piloting AI to full scale and how can the management remove them?

The most common barrier is a low level of trust or a lack of competence to transform within the organization, and what is associated with it, resistance to this technology at medium management levels. The transition to the full use of AI can also block the lack of clear scaling criteria in the organization. If we do not have a specific purpose, it is difficult to go in one direction.

Therefore, the management should set a implementation path, support testing on real data, simplify decision -making processes and reward initiatives with the greatest scalability potential.

What KPI and early warning indicators do you recommend to measure the influence of AI on value for business?

The method of measuring success is largely dependent on the purposes of implementing AI. We often show two solutions in conversations with clients. The first is horizontal mocking, focused on employee education and personal productivity. In this case, we measure success by adopting technology in the organization – i.e. the number of people who use AI tools every day or the number of generated prompts per day.

The second solution is vertical. We measure the use of AI in business processes. They often relate to the effectiveness of the process, the effort of involved people, but more and more often also creating a new business value that would not be possible without AI.

How to build a culture combining “data hygiene” with openness to experiments and what does the management have to do to keep it?

This is not a new challenge for management because we see here analogies to invest in other innovations. To maintain openness, you need initial success, an investment in AI, which will create a significant business value.

Many companies are already at this stage when traveling with AI – I like to call it scaling – that is, when they expect from AI to translate into a business result, not just effective experiments. Such the first success is a very strong base for building the culture of openness to further experiments with this technology.

What should you remember

Which regulations (e.g. EU AI ACT) are the most important for management board today and how to prepare an organization in advance for subsequent requirements?

There are no more important and less important regulations – only those that are already in force and those for which you need to be prepared, because they will apply to us in a moment. Of course, EU AI ACT is a key document here, which, among others He imposes on companies that are AI suppliers, the obligation to educate employees (so -called AI writings) or prohibits the use of artificial intelligence systems with an unacceptable risk.

In turn, from August 2, 2025, providers of general purpose AI models (GPAI) will come into force on the provisions on management principles and obligations regarding these models. To prepare well for them, you need to understand for what purposes we use AI in the company, what solutions are created and manage the AI GoverNance process in such a way as to meet all EU AI ACT guidelines.

Although Ai Governance can be associated with a slowdown in innovation, we observe that particularly large companies emphasize that clarity and transparency makes it much easier to create and implement solutions. This is because we clearly see the whole process, the roles of people who participate in it and the framework in which we want to use AI in our organization.

What you advise management in the context of Agentic AI and the growing autonomy of systems. How to include these topics in the agenda early?

It may seem that Agentic AI is still a distant future. However, I think that it is worth including this topic in the agenda today. First of all, you should identify business processes in which such a solution can be used and build a business case on it. At the same time, from the beginning of experiments with autonomous systems, the principle of human participation at critical decision -making points should be implemented. Even if the AI agent works alone, man should be able to supervise, verify and possible correction of decisions at important stages. The autonomy of systems does not necessarily mean that they work without control or transparency – that's why we build emergency scenarios and carry out immune tests.

Are there good exit strategy practices – the procedures for fast disabling or replacing AI models that turn out to be unbelievable or unethical?

For each critical model, the company should create a plan to withdraw or replace the model in the case of degradation of its quality, clear “cut-off” criteria based on trust, compliance and accuracy indicators, ready alternatives (e.g. switching to a manual process or simpler decision-making rules) and mechanisms of testing and audit of models in the life cycle.

The role of the board is to enforce the existence of such procedures and make sure that they do not work only on paper, but were realistically tested in the organization. Exit Strategy is not only a technical issue today, but a condition of trust, internal and market, to the entire AI technology.