McKinsey warns that people, not models, slow enterprise AI

On June 10, 2026, McKinsey published an interview with Sandra Durth and Suman Thareja about the second edition of Rewired. The text is less about models and more about company management. Its central argument is that the success of a technology and AI transformation is not determined only by tool quality, but by the organisation’s ability to change people’s work. In practice, this means who can use AI, who checks outputs, who changes the process and who is accountable for results. Sandra Durth says in the interview: Every employee must become a technology-enabled problem solver. That changes the training agenda. It is not enough to train a small expert team or launch an internal prompt library. If AI changes decisions in sales, operations, HR or finance, ordinary employees must understand it. Otherwise a gap opens between a few enthusiasts and the rest of the company. Durth also points to the difference between short-term productivity and lasting advantage. A company can use AI to write emails and meeting notes faster. That saves time, but competitors can copy it easily. A more durable advantage appears only when the company rebuilds processes, performance measurement and decision-making. In a bank, that may mean a different approach to risk. In manufacturing, faster capacity planning. In retail, more accurate management of prices and stock. Suman Thareja adds that the usual training route is no longer enough. Learning has to happen in the flow of work, he says. This is a practical warning for companies that treat AI training as a one-off workshop. People usually leave with a few tricks, not a new workflow. Learning must happen on a real task, with real data and quick feedback. The business impact is direct. When only individuals learn AI, the value remains in their heads and the company cannot manage it. When entire teams learn, shared rules, measurable procedures and shared accountability emerge. That is the difference between an experiment and operations. For Czech firms, the uncomfortable point is that this cannot be bought ready-made. A supplier can provide the tool. Changes in roles, approval, measurement and responsibility must be solved by leadership. Terms: Scaling AI means moving from isolated experiments to AI use in normal company operations. AI literacy is the practical ability to assign a task, check an output and understand risks. Flow of work means learning directly on a task a person actually performs

McKinsey warns that people, not models, slow enterprise AI

The full article is available to users registered at Hard Skills