I’ve spent a lot of time recently talking with people about the impact of AI. In most cases, it doesn’t take too long before these conversations take on an air on unreality as people describe more and more fanciful examples of how “AI is going to change everything”. With a mixture of fear and awe they seem entranced by the endless possibilities of where AI can be applied. While some of this excitement is clearly warranted, I can’t help but think that that too many of these discussions do not seem grounded in reality, missing a more fundamental understanding of why AI-driven digital transformation is revolutionizing businesses and society.

In the rush to adopt AI, many appear to have fallen prey to the illusion that AI is magical thinking powered by the alchemy that converts data into insight. With headlines proclaiming AI as revolutionary and transformative, it’s easy to lose sight of a basic truth: AI is fundamentally an economic phenomenon, not a magical one.

Ajay Agrawal, Joshua Gans, and Avi Goldfarb, in their seminal work “Prediction Machines,” provide a simple framework for understanding this economic perspective. They argue that the true impact of AI isn’t about creating sentient machines or digital consciousness, but something much more practical – making prediction cheap.

Prediction, in economic terms, is the process of using information you have to generate information you don’t have. When viewed through this lens, AI’s purpose becomes clearer: it reduces the cost of prediction. This economic framing helps demystify AI and allows leaders to make more rational decisions about implementation and investment.

Consider how we’ve historically approached prediction problems. Weather forecasting once relied heavily on human expertise and limited data. Today, we use sophisticated models that process vast amounts of information to deliver increasingly accurate forecasts. This isn’t magic – it’s the economics of prediction at work.

The same principle applies across industries. In healthcare, AI systems predict patient outcomes and disease progression. In finance, they predict market movements and credit risks. In manufacturing, they predict maintenance needs and supply chain disruptions. Each application represents prediction becoming cheaper, faster, and more accessible.

However, one economic factor that has limited AI adoption is the extraordinarily high cost of training large models. A single training run for today’s largest models can cost millions of dollars, requiring specialized hardware, massive datasets, and significant energy resources. This has restricted model development to well-funded tech giants and specialized AI labs.

Yet, this economic constraint is rapidly changing. As DeepSeek’s release reviewed, training costs are expected to decrease dramatically as computing efficiency improves, specialized AI hardware becomes more available, and techniques like transfer learning and few-shot learning reduce data requirements. Just as previous technological advances followed predictable cost curves, AI training costs will likely follow a similar trajectory – dropping by orders of magnitude in the coming years.

For digital leaders, this economic perspective offers actionable insights. First, focus on identifying prediction problems within your organization where even modest improvements would deliver substantial value. Second, understand that AI’s ROI must be measured against the economic value of improved predictions – not against hyped expectations of magical transformation.

The economics-first approach also highlights the complementary human skills that will become more valuable as prediction becomes cheaper. As Agrawal and colleagues note, when prediction costs fall, the value of judgment, creativity, and interpersonal skills increases. Leaders should invest in developing these complementary capabilities alongside their AI implementations.

By viewing AI through an economic rather than magical lens, leaders can make more disciplined decisions about where and how to deploy these technologies. They can assess AI investments based on concrete economic returns rather than fear of missing out on a revolution. They can prepare their organizations for an era where prediction is abundant while judgment remains scarce.

The future of AI isn’t a magical transformation but an economic evolution – where prediction becomes a commodity and human judgment becomes the differentiator. Understanding this distinction is the first step toward realizing AI’s true potential in your organization.

Originally posted here

Read More AI for good