During this year's study week at Harvard Business School, we explored the current technological leap and its impact on future business models.
We are living in a transitional period. We already have 5G telecommunications infrastructure, which enables us to transmit data at a much higher volume, faster, and more efficiently. While there are initiatives to take advantage of these opportunities, business models are still lacking. The metaverse is taking a break. The intelligent networking of machines, robots, and sensors in factories—for more precise planning and control of production processes and real-time responses to disruptions—is not yet a given. The same applies to digital twins, virtual representations of systems that could be used to run simulations aimed at improving efficiency. It is a transitional period similar to the six years between 2002—when the 3G telecommunications infrastructure was introduced—and 2008, when the 3G iPhone established a groundbreaking business model based on this technology in the market.
We are living in a transitional period because the potential of artificial intelligence (AI) is only just beginning to unfold. As a result, many existing business models are still based on the fact that we cannot predict the future. We build logistics centers because we don’t know which items consumers will buy; so we have to keep as many in stock as possible to be able to deliver quickly. Airports have become shopping districts to make unexpected or security-related wait times more bearable for passengers. Prada and Marco Polo are banking on this. AI, on the other hand, enables predictions. This is already the case with the next word that Copilot suggests by default when I’m writing an email. Or with a query to ChatGPT, which uses existing information to generate information that isn’t yet available. AI is now becoming increasingly intelligent at an accelerating pace because it has begun to learn on its own (“deep learning”).
It’s becoming clear to me that any human prediction about whether and when machines will surpass humans is built on shaky ground. This is true even for experts like economist and AI specialist Bryan Caplan, who in January 2023 ruled out the possibility that any version of GPT would be capable of achieving a perfect score on the standard economics exam used by American universities before 2029—only to be proven wrong two months later with the release of ChatGPT 4.0. The last ten years of AI development have taught us that no one should bet against the possibilities of deep learning, even if computing capacity, energy consumption and costs, and data quality still pose limitations at present. The significantly more efficient large language model from the Chinese company DeepSeek, which was unveiled to the public in January 2025, and the fact that Google’s Gemini 1.5 significantly outperforms ChatGPT while being ten times cheaper, suggest that these bottlenecks could soon be a thing of the past. In his highly recommended essay “Situational Awareness,” published in June 2024, AI architect Leopold Aschenbrenner demonstrates that a superintelligence emerging from deep learning will be a reality by 2028.
With 5G, the infrastructure for the next major technological leap is already in place, and artificial intelligence is evolving autonomously at an exponential rate: What does this mean for our companies, for their future business models, and for the development of strategies? It’s high time to think about the interim period from the end forward and ask “what if” questions.
What if we knew for certain today that,
- that starting in 2028, we will no longer use a chatbot like ChatGPT in a sequential question-and-answerand answer format using only the internet, but that AI will be able to think longer, independently engage in dialogue with human and artificial experts both inside and outside your company before responding, and ultimately provide a well-developed, thoughtful analysis of a problem—including proposed solutions, an evaluation of those solutions, and a recommendation? In other words, will we have artificial employees on board?
- that the virtual employee uses a computer just as people do today—that is, independently sets up Zoom meetings, participates in them, takes minutes, conducts online research, sends emails to people, and reads and analyzes documents?
- that AI is capable of independently and proactively performing calculations, tests, and simulations—taking into account all conceivable and inconceivable external factors—more effectively and comprehensively than any human could with the aid of a computer?
- that predictions will be so precise and machine-to-machine interactions so seamless and perfect that all business models based on a lack of predictive capability will become obsolete?
What impact will these and other conceivable developments have on our business model? Will it have become obsolete by 2028? If not, why not?
How are you dealing with the avalanche of artificial intelligence that is now bearing down on us? What conclusions are you drawing from this for your companies?
