I came to AI from management, not technology. Before focusing on AI-enabled operating models, I spent 7.5 years leading a major university hospital with approximately 3,000 employees, after earlier work in strategy, consulting and data analytics. That experience shaped how I look at AI today: not primarily as another software category, but as something that can fundamentally change how an organization is designed and managed.
Most companies still approach AI as another technology rollout: buy tools, give employees access, train them to write better prompts and hope productivity improves. That is a reasonable starting point, but it is not transformation. The more important question is much more fundamental: if we designed the company today, knowing what AI can already do, would we build its processes, roles and management structure in the same way? In many organizations, the answer is clearly no.
Most business processes were designed for a world in which cognitive work had to be performed almost entirely by people. People researched, summarized, analyzed, prepared reports, transferred information between departments and turned raw inputs into decisions. AI changes that basic constraint. Simply inserting an AI assistant into the existing workflow may make an inefficient process faster, but the real opportunity is to redesign the workflow itself: decide what should be produced by AI, what requires human judgment, where responsibility must remain human and where a person should intervene only when something exceptional happens.
This also changes the role of management. Many highly qualified professionals still spend a surprisingly large part of their time producing work rather than managing it: preparing analyses, consolidating information, drafting presentations or manually checking outputs. In an AI-native organization, managers increasingly define objectives, allocate work, set standards, evaluate results and resolve exceptions, while production capacity comes from a combination of people, software and AI agents. AI therefore does not make management irrelevant. It makes good management more valuable, because delegating work to intelligent systems requires much clearer objectives, accountability and quality control than simply assigning tasks to another employee.
The same principle applies to autonomy. The goal should not be to automate everything as quickly as possible. Autonomy should be earned. A sensible organization first lets AI assist, then execute bounded tasks, validates the quality of those outputs and only afterwards increases autonomy where the evidence supports it. The right architecture is not “human or AI”, but a deliberately designed combination of AI execution, independent validation and human decision gates.
The economics matter more than the technology. Counting licenses, prompts or employees using AI tells management almost nothing. What matters is whether a redesigned process reduces total cost, shortens decision time, improves quality, lowers rework and releases scarce senior attention for work that actually requires it. An AI system that generates an answer in ten seconds but creates an hour of human correction is not productive. The relevant measure is the net economic value of the complete workflow.
One lesson from managing complex organizations is that technology by itself rarely transforms anything. Processes, responsibilities, information flows, incentives and decision rights have to change with it. I see AI in exactly the same way. The companies that gain the greatest advantage will not be those that simply buy the best models. They will be those that redesign how the company works around a new reality: intelligence is no longer supplied only by people.
