Few companies have made as much of an impact on global business over the last few years as OpenAI, the AI research company best known for ChatGPT and DALL-E, among other LLM-based tools.

Understanding and quantifying this impact in real-time is the job of Aaron "Ronnie" Chatterji, Chief Economist of OpenAI and former White House Coordinator for CHIPS and Science Act Implementation during the Biden administration. In an event hosted by Columbia Business School’s Distinguished Speaker Series and AI in Business Initiative, Chatterji drew on his experience across academia, government, and industry to unpack how AI is reshaping business, productivity, and the global economy.

In conversation with CBS Dean Costis Maglaras, the David and Lyn Silfen Professor of Business, Chatterji explained the implications for leadership, the future of work, and decision-making required as AI becomes a core general-purpose technology.

AI as a General-Purpose Technology

Chatterji framed AI as a general-purpose technology with the potential to reshape entire economies, much like electricity, computing, or the internet. As with those earlier technologies, the most profound effects of AI are unlikely to appear all at once or evenly across sectors.

While AI capabilities are advancing rapidly, Chatterji emphasized that real economic impact depends on deployment, diffusion, and organizational change. Benchmarks and demonstrations often highlight what models can do in controlled settings, but translating those capabilities into sustained productivity gains at scale takes time. In economic terms, productivity is a lagging indicator, especially for technologies that require firms to rethink workflows, roles, and incentives.

“AI is helping us do a ton of stuff faster and better, but it isn't necessarily quantified,” Chatterji said. “... The CapEx channel is important, but what people really mean when they talk about AI productivity is, for a given set of inputs, generating more outputs. That is the lagging indicator for AI's impact.”

This gap between technical progress and economic measurement helps explain a growing tension in public discourse: AI already feels transformative in daily life, yet its effects are not fully visible in traditional macroeconomic statistics. That disconnect, Chatterji suggested, is evidence that general-purpose technologies like AI take time to work their way through complex systems.

Where AI Creates Value Today

Much of AI’s current economic value, Chatterji noted, appears outside the places where economists and policymakers typically look. Consumer use of tools like ChatGPT is dominated by practical guidance, writing assistance, information-seeking, and other activities that save time and improve quality but do not show up directly in GDP. Economists refer to this as consumer surplus: real value that accrues to individuals without being captured in formal economic output.

At the same time, enterprise adoption of AI is accelerating. OpenAI’s business users are sending dramatically more messages year over year, reflecting growing experimentation with AI in professional settings. However, Chatterji cautioned against assuming that adoption alone guarantees productivity gains. One of the central challenges he identified is what he calls “capability overhang,” or the gap between what AI systems are already capable of doing and how most people actually use them.

Within organizations, a small group of power users is often extracting far more value from AI than the average employee. Closing that gap is less a technical problem than a leadership one. It requires managers to rethink training, workflows, and expectations, and to help translate individual experimentation into organization-wide change.

Rethink Your Job

While jobs can be understood by some as simply bundles of tasks, Chatterji argued instead that work often evolves as technology changes. When personal computers entered the workplace, for example, they automated many tasks, but they also enabled workers to take on higher-value responsibilities.

“I’m bullish on humans,” Chatterji said.

Recent research reflects a similar pattern with AI. In some settings, lower-performing workers benefit most, using AI tools to catch up to more experienced peers. In others, high performers use AI to extend their advantage. The common thread is that AI tends to complement human labor more often than replace it, reshaping how value is created within roles rather than eliminating those roles outright.

For leaders, this shift raises new expectations. Basic AI fluency is quickly becoming table stakes, but it is not enough. Chatterji emphasized the importance of judgment, intuition, accountability and similar qualities that cannot be delegated to a model. Even as AI handles more technical tasks, humans remain responsible for interpreting results and making decisions in context.

The Global Stakes of AI

The pace and shape of AI adoption vary widely depending on organizational size and institutional context, according to Chatterji. Smaller firms can often move faster, building AI-native workflows without needing to reengineer legacy systems. Larger organizations face greater constraints, including regulation, reputational risk, and internal complexity. 

In those environments, Chatterji argued, change management becomes a critical leadership skill. AI cannot simply be “dropped in” like a traditional software product; employees must be brought along for adoption to succeed.

Globally, attitudes toward AI diverge sharply. In countries like India, AI is widely viewed as an opportunity to leapfrog existing economic hierarchies. In parts of Europe, it raises concerns about competitiveness and regulatory balance. Meanwhile, rapid deployment in China highlights different norms around coordination and scale. For business leaders, Chatterji warned, understanding AI’s impact requires a global perspective.

In the U.S., the CHIPS and Science Act, which Chatterji helped implement, reflected a renewed embrace of industrial policy in the United States—driven by concerns about national security, supply chains, and technological leadership. Similar considerations are now shaping debates around AI infrastructure, from energy and data centers to advanced chips.