This piece first appeared in German. The English version is a rewrite rather than a line-by-line translation, and the German original stays online in the archive: Erfahrung schlägt Effizienz.

Boston Consulting Group sorted companies by what they actually get out of AI and found 5 percent generating measurable value at scale. Another 35 percent are scaling and beginning to see something. The remaining 60 percent, after substantial investment, are reaping hardly any material value at all.

Almost everyone has the technology. Very few have the return. That gap is not a technology problem, and reading it as one is how a company ends up buying more tools for a problem the tools were never going to solve.

Efficiency measures pace, and says nothing about direction

Efficiency has been the organizing goal of business for decades: optimize the process, cut the cost, speed up the cycle. AI looks like the obvious next step in that sequence, because it takes over routine, finds patterns and produces output at a speed no team can match.

The logic has one weakness: it measures tempo and stays silent on direction. Deploy AI without embedding it in a strategy and what you get is efficiency with no gain in understanding: more produced, less understood.

Judgment is what turns information into meaning

Judgment is the work that happens between data and decision. A model can calculate, compare probabilities and derive hypotheses. Producing meaning is a different operation, and it needs experience, context and values in the same place.

Experience supplies the patterns that let you place something new, and context establishes when and why a decision makes sense. Values keep efficiency from becoming an end in itself. In a working world trained on speed, this is the form productivity actually takes, and it looks less like reacting faster and more like understanding better.

Delegating the doing delegates the learning hidden inside it

Most organizations treat AI as relief, which is right and incomplete. Hand over everything that no longer seems necessary and you often hand over the thinking that was buried in it.

A junior who used to work through two hundred pages of due diligence was learning structure, skepticism and pattern recognition while doing it. When a model takes that task, the work disappears and so does a piece of the learning process. The same holds for tax advisers, auditors and bankers. A system can check data without developing a feel for risk, and it will flag a deviation without telling you whether the deviation matters.

Experience is not a nostalgic factor on the balance sheet. It is what translates a technical result into a human decision.

Experience is the filter that decides what is relevant

A model has no intuition, and it treats all information as equally worth having for as long as the information sounds plausible. People work the other way around, sorting what matters from what does not, and that sorting capacity comes from having seen things before.

In practice the division is clear enough. AI produces a hundred possible scenarios and judgment decides which of them is realistic, while experience is what separates the true from the probable from the wishful. The deeper a company integrates AI into its work, the more this filter is worth, because it is the only defense against saturation by information.

Experience turns strategic again

For a long time the ideal was young, fast and digital. That standard is shifting, and experience is becoming strategic again, less as a form of authority and more as an ability to read uncertainty.

Experience in this sense has nothing to do with standing still. It means remaining able to judge, and so able to act, when the situation is ambiguous. Orientation in a world full of data is the new capital.

Leadership moves from control to interpretation

The job of leading changes with it. Distributing tasks is no longer the core of it; conveying meaning is. AI delivers results that leaders have to explain, weight and place, which means turning technical precision into human orientation.

That looks like understanding before deciding, asking before knowing, and reflection ahead of oversight. Curate, don't command. Anyone unable to orchestrate complexity loses authority over it, which makes judgment an organizational survival strategy and not a private virtue.

Three ways to build it

Contextualize knowledge. Data and tools describe the world without explaining it. Companies have to create the rooms where interpretation and argument happen.

Bring experience in systematically. Mentoring, reverse mentoring and case shadowing are strategic investments in collective thinking capacity, whatever the HR budget line calls them.

Keep responsibility visible. AI may support a decision and must never obscure who made it. Responsibility stays human, and that is the right outcome.

Efficiency is reproducible

AI raised the tempo without replacing the thinking. The more the machines take over the doing, the more the capacity to interpret, to doubt and to decide becomes the thing that separates one firm from another.

Experience beats efficiency because it supplies context, and context is what turns data into meaning. Technology can calculate a great deal, and meaning is still made by people, which is why judgment remains the competitive advantage worth having.

Sources

Grouped by the section they support.

Opening

  • Apotheker, Beauchene, de Bellefonds et al. (2025), The Widening AI Value Gap: Build for the Future 2025, Boston Consulting Group, September 2025, bcg.com. Supports the study of more than 1,250 companies: 5 percent achieving AI value at scale, 35 percent scaling and seeing some returns, and 60 percent reporting minimal gains despite substantial investment.

All other sections

  • No external source. The argument about efficiency, judgment, experience and leadership draws on the author's professional experience in restructuring, turnaround work and consulting.
The link has been copied!