Demystifying AI for business leaders

Is Artificial Intelligence truly intelligent? The instinct to attribute human-like cognitive abilities or even self-awareness to a responsive machine says more about us than about the machine itself. Even the brightest among us can be dazzled by recent developments in AI. Without understanding the technology, we can easily misjudge what we are seeing.

Professor Marc Torrens recently gave a presentation at Esade Business School, designed to demystify AI for business leaders. AI isn’t sorcery or salvation; it is a set of mathematical systems that are powerful in certain domains, limited in others—and frequently misunderstood.

The history of AI

Artificial intelligence actually dates back to 1956, when it emerged as a formal discipline within computer science. For decades, researchers worked with so-called symbolic AI: systems that used the knowledge we input to solve structured problems.

Advancements came in the 1990s. “In the 90s, due to digitalization, we created vast amounts of data,” says Torrens. “Machine learning systems no longer had to rely on handcrafted rules, but instead detected patterns in large amounts of data.”

Over the past four or five years, generative AI has accelerated rapidly. The leap forward is that it can now create new content. Large Language Models (LLMs) are trained on vast amounts of text and learn statistical relationships between words, phrases, and concepts. They do not ‘understand’ in a human sense, but they calculate probabilities and produce outputs.

Is AI intelligent?

The scale is staggering. “We can see large language models as a knowledge repository,” says Torrens. “If a human being worked 12 hours a day, seven days a week, it would take that person 300,000 years to process the volume of documents a large language model ingests during training.”

The comparison is striking, but also misleading. Torrens uses the example of a four-year-old child, who processes a comparable magnitude of information over those four years. The difference is that children not only learn from text, but also from sight, sound, touch, cause and effect. The text that LLMs train on is an inadequate representation of reality.

This distinction is crucial when assessing whether generative AI is truly intelligent. Silicon Valley rhetoric often suggests that Artificial General Intelligence (AGI)—systems with flexible, human-like reasoning—will be the next development. Researchers tend to disagree. Intelligence requires more than pattern recognition. “AI can produce plausible conversations, but it doesn’t understand the concept of a vision,” says Torrens.

The philosopher Daniel Dennett captured the paradox succinctly: “We are creating machines that are very competent without any comprehension.” Competence without comprehension can appear to be intelligence. But it is not the same thing.

Why AI feels intelligent

Our own psychological makeup means we are wired to anthropomorphize. Torrens cites the example of a child who might draw the sun with a smiling face. “We routinely attribute our own capacities and characteristics to the things around us,” says Torrens.

When a chatbot produces fluent and context-aware responses, we instinctively assume there is intelligence behind the words. Yet these systems are stochastic. They are non-deterministic, which means that for the same query, you receive different results. This variation gives the impression that AI is creative, even human. The flip side is that the answers cannot be assumed to be correct.

What AI really means for business leaders

Today’s AI systems are limited. Some algorithms can drive cars. Some win at chess and poker. But just like an algorithm trained to drive a car cannot suddenly play chess, an AI model optimized for language cannot autonomously manage business strategy. It’s important to understand the specific specialty of each AI system, and then it can be used effectively. AI can analyze documents, classify images, detect anomalies, draft reports, and tease out patterns that would be invisible to human analysts.

AI creates real value when tailored to clear goals, guided by people, and built into smart workflows

The opportunity lies in vertical application. “The interesting thing for business is to be able to personalize AI models,” explains Torrens. Companies that define a clear problem, curate relevant data, and maintain human oversight can create tangible value. Used well, AI can enhance existing human outputs, enabling teams to produce more reports, test more scenarios, or respond more quickly to customers.

But there’s a caveat. Increased efficiency in one area of the business could cause a bottleneck elsewhere. If AI allows employees to generate twice as many reports, someone still needs to read them. Therefore, leaders need to plan and redesign workflows to truly benefit from AI. Human judgment is still required.

Europe’s opportunity

While the frontrunners in AI are touting a leap to AGI, we currently don’t know how to reach that goal.

“We don’t have any more data to train the systems, and we are already detecting diminishing returns,” says Torrens. While there will be incremental improvements to AI, researchers believe we may be hitting a wall in terms of enhancing its capabilities.

Beyond AGI lies the speculative notion of superintelligence—systems exceeding human capabilities across domains. Such projections attract extraordinary investment and headlines, but they are a long way off. Discussions of an ‘AI bubble’ have already begun, and recent reports about OpenAI potentially requiring government funding have raised concerns.

For Europe, this could present an opening.

“A lot of people have been saying Europe is getting behind because we haven’t been investing as much as the US or China in AI technology,” says Torrens.

“But at the end of the day, if this technology isn’t bringing the value promised by Silicon Valley, then this is an opportunity for Europe. Europe can make a difference by designing systems and applications that use AI technology for a specific case or in a specific industry,” he adds.

As we increasingly turn to chatbots for guidance and companionship, worries arise about losing human independence

Risks to consider

The concerns of the techno-pessimists that overreliance on AI systems can erode human autonomy gain some traction when you consider that we already outsource navigation to mapping apps and music discovery to recommendation engines. Increasingly, people ask chatbots not just for information, but for advice and companionship.

Is AI sustainable? Behind the polished interfaces are human labor forces, often working under harsh conditions in the Global South, reviewing disturbing content to train and moderate these systems. From a resource perspective, AI data centers require significant amounts of electricity and, in many cases, large quantities of water for cooling.

These concerns, taken together, point to a broader question about what lies beneath the convenience and apparent intelligence of these systems.

Understanding a powerful tool

AI is not magic; it is mathematics at scale—probabilities layered upon probabilities, trained on vast archives of data. Business leaders should neither overestimate it nor fear it reflexively, but instead understand what it is actually doing.

Artificial intelligence is potentially one of the most powerful tools we have ever developed. But it is not, at least not yet, thinking.