When Ford Motor Co. recently announced that it was bringing back hundreds of experienced engineers to strengthen product quality, it delivered one of the most important lessons of the artificial intelligence (AI) era.
After years of expanding the use of AI and automated engineering systems, Ford acknowledged that inspection systems failed to deliver the results.
Kumar Galhotra, its chief operating officer, admitted: “We had been relying more and more on automated quality systems and not getting the desired results. We brought back technical specialists, and they hunt for failure points before a part ever reaches the plant floor.”
Ford’s experience should not be dismissed as an isolated engineering setback. It is an example of the illusion of wisdom that AI creates. A process can be automated, but not wisdom. Interestingly, history offers a useful parallel.
During the 1990s, organizations rushed to embrace enterprise software and workflow automation, believing that digitizing existing processes would automatically improve ‘productivity.’ Ironically, many simply automate inefficiency.
Management thinker Michael Hammer captured this lesson in his famous dictum: “Don’t automate, obliterate.” Technology should redesign work rather than merely accelerate flawed processes. AI risks repeating that mistake on a far greater scale.
Much of the public debate has focused on whether it will eliminate jobs. However, the deeper and less appreciated risk is epistemic. AI is changing not only how we work, but how we decide what to believe. Two pieces of work are relevant in the context of Ford. Herbert Simon observed over half a century ago that a wealth of information creates a poverty of attention. As information becomes abundant and cheap, human attention and judgement become scarce.
Three decades later, James Surowiecki (2004) argued, albeit in a different context, that the “wisdom of crowds” emerges only under specific conditions: individuals must think independently, possess diverse information and perspectives, and combine their judgements through effective mechanisms.
Collective wisdom, therefore, depends not merely on having many opinions but on those opinions remaining genuinely independent and diverse. AI threatens at least two of these conditions: As millions of people rely on the same AI model, their judgements get correlated and lose independence.
Also, large language models often produce similar consensus-like answers, reducing the diversity of viewpoints. Therefore, instead of amplifying the ‘wisdom of crowds,’ AI risks creating what one can call the ‘illusion of wisdom,’ as many people arrive at the same answer not because it is correct, but because it was generated by similar underlying models.
There are other information-based theories that can help us understand the dynamics of such suboptimal choices in the AI era. For example, economists Sushil Bikhchandani, David Hirshleifer and Ivo Welch have shown how informational cascades arise when individuals rationally ignore their private information after observing others’ actions.
Similarly, Stephen Morris and Hyun Song Shin identified a different but equally powerful mechanism. When everyone observes the same public signal, rational decision-makers place excessive weight on it because they know everyone else is observing it too.
The common signal becomes a focal point for coordination. If it is wrong, mistakes become widespread even though people may be responding rationally to the same source of information.
Generative AI combines elements of both mechanisms while creating a new challenge.
First, AI eliminates information scarcity while making independent human judgement relatively scarce. As machine-generated answers become instantaneous, comprehensive and inexpensive, fewer people invest the effort required to verify them independently.
Second, repeated exposure to accurate and confident AI-generated responses encourages users to outsource not merely information retrieval but judgement itself. Confidence could gradually become a substitute for competence.
Third, AI acts as a common informational anchor. Decisions grow increasing correlated because they originate from the same statistical engine. In classical herding, people imitate one another. This differs. Nor is this simply an informational cascade.
AI is yielding a world in which independent reasoning is crowded out by the same machine-generated starting point.
The consequences would be profound. The greatest danger is the gradual erosion of society’s capacity for independent judgement, which may be worse than an overload of misinformation.
Ironically, genuine expertise will be more valuable in a world where the rarest skill will be the ability to recognize when information is incomplete, misleading or simply wrong. Only genuine expertise will be able to break an ugly loop of artificial wisdom.
Three decades ago, Michael Hammer warned managers not to automate bad processes. The AI era demands an even more fundamental warning: do not automate wisdom. Machines can process information at superhuman speed, but wisdom is forged through experience, judgement, accountability and ethical responsibility.
If society mistakes statistical predictions for wisdom, the greatest casualty of this technology’s rise will not be employment but humanity’s capacity to think independently.
The author is professor, Madras School of Economics.