We have entered a new era where AI is capable of thinking and reasoning like humans, tackling complex problems that have long been the exclusive domain of expert professionals. This significant development emerged just a few months ago with the release of OpenAI’s first “reasoning” models, which can comprehend and solve problems by drawing logical inferences and adapting to new information. More recently, DeepSeek and Anthropic have also made notable contributions to the field with their own reasoning models, which have been developed at an unprecedented pace and cost.

Let’s delve into the process of AI “reasoning” and explore the implications of this groundbreaking capability for your business.

Uncovering the Breakthrough and its Significance
Unlike current AI models that primarily rely on pattern recognition to provide instant responses, reasoning AI employs a more deliberate and multi-step approach. It utilizes chain-of-thought reasoning to break down intricate problems into manageable components, allowing it to explore different paths and adjust its approach when necessary, much like humans do when solving problems.

The traditional method of enhancing AI model performance involved feeding them massive datasets during training. In contrast, reasoning models leverage a strategy called test-time compute, which involves allocating more processing power and time during the actual problem-solving stage. This enables the AI to think more deeply and provide more comprehensive and accurate answers.

While reasoning AI is not perfect and still lags behind humans in terms of common sense and contextual understanding, its capabilities make it an extraordinarily powerful tool. It can solve complex problems that other systems struggle with, and its applications are vast and promising.

A notable example of the power of reasoning AI can be seen in the work of Ethan Mollick, a professor at the Wharton School of the University of Pennsylvania. Mollick used OpenAI’s o1 reasoning model to identify a math error in a research paper that had briefly sparked concerns about the safety of black plastic cooking utensils. The AI model quickly pinpointed the mistake, demonstrating its ability to understand context and apply logical reasoning.

As Mollick noted, “When models are capable enough to not just process an entire academic paper but to understand the context in which ‘checking math’ makes sense, and then actually check the results successfully, that radically changes what AIs can do.” This capability has significant implications for various fields, including research and development, where AI can propose hypotheses and simulate outcomes on its own.

Reasoning models have achieved impressive results on intelligence benchmarks. For instance, OpenAI o3 outperformed human experts on the GPQA Diamond benchmark with a score of 87.7%. On FrontierMath, a set of challenging math problems, o3 scored 25.2%, marking a significant improvement over previous models. Additionally, o3 scored 87.5% on the ARC-AGI test, surpassing both previous AIs and the baseline human level.

While AI is not poised to replace human expertise and judgment, its ability to reason and provide scalable, always-on support represents a powerful new paradigm. This development will require leaders and organizations to adapt and redefine their approach to work and problem-solving.

Decoding the Potential Impact of Reasoning on Business
The emergence of reasoning AI holds tremendous promise for businesses across industries. Its potential applications in research and development are particularly significant, as AI can now propose hypotheses and simulate outcomes independently. This could revolutionize fields such as renewable energy and pharmaceuticals, cutting years off traditional R&D cycles and driving breakthroughs.

More broadly, reasoning AI will disrupt many of our assumptions about work. Leaders should consider two key implications: First, these models can perform cognitive labor equivalent to or better than humans, perceiving, understanding, reasoning, and executing tasks at levels that approach or surpass human abilities. For every task, leaders should ask themselves, “Can AI do this job?” If the task doesn’t require uniquely human skills like judgment, nuance, originality, or emotional intelligence, the answer is now yes. We need to reimagine the division of labor between humans and AI and develop new approaches to managing that labor.

Second, reasoning models will change the economics of work. Historically, acquiring reasoning capabilities meant hiring humans, but that’s no longer the only option. Organizations can now rent or purchase cognitive labor on a consumption basis, similar to acquiring other inputs like electricity or equipment. This shift will have a profound impact on businesses and industries, and I expect the disruption to come from AI-native firms rather than incumbent companies. AI natives will have a competitive edge due to their early adoption and integration of AI into every process.

While we are still in the early days of AI reasoning, I am confident that this technology will unlock new possibilities and opportunities for businesses that I have not yet begun to imagine.

For more insights on AI and the future of work, subscribe to this newsletter.