Blog Author
Heinz is an experienced management consultant for growth and transformation strategies. His expertise comes from many years of global leadership responsibility as well as from his entrepreneurial successes as a sparring partner and consultant for CEOs, Management Boards and HR leaders.
Anita Imbach is an experienced Executive Sparring Partner for critical AI, digital and business transformations. Drawing on extensive management, leadership and transformation experience across startups, Swiss SMEs and large corporations, she works at the intersection of AI governance, executive decision making and transformation in regulated environments.
LinkedIn: linkedin.com/in/anitaimbach
At a steering meeting, a team reported that it was already using Claude for programming and UX design. While AI is still being discussed at the executive level, the team has already moved ahead.
At first glance, the issue seems clear: unauthorized AI tools. But that’s not the whole story. The employees didn’t simply choose a new tool. They decided what AI would be used for and which tasks it would take on.
The reality of decision-making has changed before management has altered decision-making authority.
This raises a new management question: Which decisions remain with humans, where does AI contribute to decisions, and where is it allowed to act on its own?
1. AI doesn’t just change work. AI changes leadership.
Until now, management has primarily determined which employees are authorized to make which decisions. With AI, this logic is changing.
Today, AI provides analyses, options, and recommendations that form the basis for business decisions. As a result, influence over decisions is already shifting before management has consciously determined how far that influence should extend.
The OTTO case study examined by Sturm, Krause, and van Giffen (2026) shows that rules alone are not enough. How AI is used is also shaped in day-to-day work.
AI changes not only how work is done, but also how influence over decisions is distributed within the company.
2. Gaining speed without losing control.
The obvious answer is more control. Yet additional approval steps can limit the benefits that AI is intended to provide.
If AI generates an analysis in minutes while the subsequent decision still goes through the same approval loops as before, one step in the process becomes faster. The company does not necessarily become faster.
This is precisely where the paradox lies for boards. In a Gartner survey of 328 non-executive directors, 91% viewed AI as an opportunity for shareholder value. At the same time, 80% considered their existing board structures and practices insufficient for effective AI oversight.
Executive management and the board must assess whether their existing governance structures still align with the new reality of decision-making.
3. Reality is moving faster than governance.
At another company, certain AI tools were blocked. As a result, employees used personal devices and their own AI tools and explored potential uses for AI agents in customer care.
This could be interpreted as a “shadow AI” problem. To me, however, this interpretation falls short.
The employees have already explored which tasks AI can take on and where AI could eventually act on its own.
Management never consciously made that decision. Yet it happened anyway.
For senior leadership, therefore, it’s not just a matter of which AI tools are approved. They must recognize where employees are already using AI in ways management has not yet consciously authorized.
4. Governance Is the New Management
AI is transforming a central aspect of management. It’s no longer just about what decisions are made, but also about who is authorized to make them.
AI can prepare decisions, influence them, or, for clearly defined tasks, decide and act on its own. Management must determine how far that role should extend.
For me, the line is clear when it comes to fundamental issues: strategy, business model, and corporate direction fall under management’s responsibility. AI can analyze, develop scenarios, challenge assumptions, and prepare options. Where decisions are based on clear rules and reliable data, AI can take on more.
AI’s technical capabilities should not determine what it is allowed to decide. Management should.
5. How much oversight is really needed?
Not every decision requires the same level of control.
The greater the business impact and risk, and the harder a wrong decision is to correct, the tighter the boundaries must be.
More human oversight does not automatically mean better oversight. Approval is of little use if the person responsible cannot evaluate the AI’s output. At the same time, AI will not scale if every AI decision requires approval at a higher level.
How much oversight is necessary depends not on the technology, but on the potential consequences of a decision.
6. Decision rights can shift. Accountability cannot.
AI can take on decision-making authority. Accountability does not disappear.
For important decisions, it must be clear who intervenes if something goes wrong and who is accountable for the outcome. If an AI-assisted or autonomous decision causes significant harm, accountability cannot be clarified only after the fact.
Whoever grants AI decision-making authority must also clarify who is accountable for the consequences.
7. Do you know where AI is already involved in decision-making in your company?
In one of my projects, a manager compiled company data and used AI to create a cost model for a new product.
AI did not decide on the product. It provided the numbers that informed the decision.
AI already influences decisions about business, customers, costs, or risks without making the final decision itself.
AI’s influence doesn’t start when AI makes decisions on its own.
Senior management must know where AI is already shaping the basis for key business decisions.
The European Commission’s current draft guidelines on high-risk classification also show that what matters is not only who makes the final decision, but also the influence AI has on that decision. Human involvement alone does not automatically change this.
8. Management designs the system within which decisions are made.
Back to the steering meeting. Management is still discussing the use of AI. The team is already working with Claude.
What is visible today in individual use cases will become increasingly important as AI becomes more capable of making decisions and acting on its own.
This does not reduce management’s responsibilities. It changes it.
Management will increasingly make fewer individual decisions. Its role is to shape the system in which decisions are made.
That is precisely why: Governance Is the New Management.
In which areas of your company is AI already involved in decision-making without management having consciously defined the limits of its authority?
Sources
European Commission (2026). Draft Commission Guidelines on the Classification of High-Risk AI Systems under Article 6 of Regulation (EU) 2024/1689 (AI Act) for Stakeholder Consultation – General Principles.
European Commission (2026). Draft Guidelines on the Classification of High-Risk AI Systems – Annex III. Article 6(2) and Article 6(3) of Regulation (EU) 2024/1689 (AI Act).
Gartner (2024). Gartner Says 80% of Non-Executive Directors Believe Current Board Practices and Structures Are Inadequate to Oversee AI. Gartner Board of Directors Survey.
Sturm, S., Krause, F., & van Giffen, B. (2026). Beyond control? Governing AI in organizations for responsible use. European Management Journal. https://doi.org/10.1016/j.emj.2026.06.003
