AI’s second act: Why intelligent agents will redefine business competition

The next AI race will reshape how businesses compete

For much of the past two years, the artificial intelligence conversation has revolved around increasingly powerful large language models, ever-larger computing clusters and the race toward artificial general intelligence (AGI). Yet focusing solely on the technology risks overlooking what may prove to be AI’s most significant commercial transformation.

The next phase of AI will be defined not simply by better models, but by intelligent agents capable of making decisions, using tools, interacting with digital environments and continuously improving through feedback. As these systems mature, they are expected to move beyond assisting workers to carrying out increasingly complex business functions on their own.

This transition marks what can be described as AI’s “second act.” If the first phase of generative AI focused on building increasingly capable foundation models, the next will center on how businesses reorganize themselves around those capabilities. Success will depend not only on developing AI technologies but also on redesigning industries, workflows and business models to take advantage of them.

For companies, this creates two distinct but interconnected arenas of competition. The first is the business of AI—building the infrastructure that powers artificial intelligence. The second is AI for business—using AI to reshape industries, create new sources of value and redefine competitive advantage.

Understanding the difference between these two battles may become one of the defining strategic questions of the coming decade.

The next global infrastructure

Every major technological revolution has been built upon a new layer of infrastructure. Electricity transformed industry by providing universal access to energy. The internet created a global platform for information exchange, fundamentally changing communication, commerce and entertainment.

The next infrastructure will not distribute energy or information but intelligence.

Rather than just a collection of standalone applications, AI may become a global network of intelligent agents operating beneath virtually every industry. Just as businesses today depend on electrical grids and internet connectivity, tomorrow’s organizations may rely on networks of autonomous AI systems capable of performing cognitive work, coordinating operations and making routine decisions.

Under this framework, AI becomes less a software product than a foundational utility supporting economic activity across sectors.

If such a network emerges, its impact could extend far beyond technology companies. Every industry built on top of that infrastructure—from healthcare and finance to manufacturing and logistics—would likely be reshaped as intelligence becomes embedded directly into business operations.

Two battles, one value chain

This transformation creates a six-layer AI value chain.

At the bottom sit energy systems and physical infrastructure, followed by semiconductor ecosystems, next-generation cloud computing platforms and foundation models. Together, these layers form the technological infrastructure required to build increasingly capable AI.
Above them are agent-based systems and industry transformation—the layers where AI is applied to solve business problems and redesign commercial activity.

This distinction matters because companies do not all need to compete in the same way.

Technology companies may focus on building the underlying infrastructure, from chips to models and cloud services. Most other organizations, however, are unlikely to develop frontier AI themselves. Instead, their competitive advantage will come from applying those technologies to reinvent how they operate.

Competition within each layer is also becoming increasingly interconnected.

Success in semiconductors, for example, depends not only on producing powerful chips but also on software compatibility, memory capacity and manufacturing ecosystems. Likewise, data centers require coordination across electricity generation, cooling systems, networking equipment and supply chains.

The same logic extends throughout the value chain. Competitive advantage increasingly comes not from excelling at a single technology but from integrating multiple capabilities into a coherent system.

This partly explains why many of the world’s leading AI companies are pursuing greater vertical integration, bringing together infrastructure, models and applications within a single ecosystem.

Understanding where a business fits within this broader value chain—and how it connects with adjacent layers—may ultimately prove more valuable than optimizing any individual technology.

From copilots to autonomous agents

The emergence of agentic AI represents the most important shift in the industry over the past several months.

The first generation of enterprise AI largely revolved around copilots—systems designed to assist humans with writing, coding, analysis or customer service. These tools remained fundamentally human-centered. People asked questions, evaluated answers and made final decisions.

Agents operate differently. Instead of waiting for detailed instructions, they can receive objectives, access software tools, interpret changing environments, make decisions and improve their performance through continuous feedback.

The distinction may appear subtle, but its implications are profound.

Rather than simply increasing individual productivity, agents have the potential to become active participants in business operations. They can coordinate workflows, monitor inventories, analyze markets and complete multi-step tasks with limited human supervision.
As these capabilities mature, AI shifts from being primarily a knowledge tool to becoming an operational system. This also changes how businesses should think about value creation.

During AI’s first phase, competitive advantage largely depended on model performance, computing power and token generation.

The second phase revolves around something different: business scenarios, proprietary data and intelligent agents capable of transforming model outputs into commercial outcomes.
The key question is no longer whether a model can generate an answer. It is whether organizations can convert those answers into measurable business value.

Where AI creates value

Many discussions about AI continue to focus on foundation models, GPUs and computing infrastructure. While these remain essential, the greatest commercial value will increasingly be created elsewhere.

AI’s economic structure can be thought of as an inverted pyramid.

Infrastructure forms the base. Foundation models produce tokens and outputs. At the top sit business systems, operational processes and commercial outcomes. It is at this upper layer where AI generates the greatest economic impact.

Access to powerful models alone will not create lasting competitive advantage because those models are becoming increasingly accessible across industries. Instead, differentiation will depend on combining three elements.

The first is deep understanding of specific business scenarios. The second is proprietary operational data accumulated through years of experience. The third is closed-loop agent systems capable of making decisions, learning from outcomes and continuously improving.

Companies that successfully combine these three components are far more likely to translate AI capability into sustainable competitive advantage than those that simply adopt the latest models. In this sense, value is migrating upward through the AI stack.

The winners will not necessarily be those possessing the greatest computing resources or generating the most tokens, but those most effectively transforming AI into new organizational capabilities, business processes and industry structures.

From adding AI to rebuilding the enterprise

Many organizations today approach AI as an incremental upgrade. Existing workflows remain largely unchanged, while AI is introduced to automate selected tasks or improve efficiency at the margins.

This approach can be thought of as “+AI”—adding AI to an existing business.

It is a natural starting point. Organizations can deploy AI writing assistants, customer service chatbots or coding tools without fundamentally altering how they operate. Because existing structures remain intact, these projects often face relatively little internal resistance. Yet this approach will capture only a fraction of AI’s potential.

The alternative is “AI+”: redesigning businesses around AI rather than simply inserting AI into existing processes.

Instead of asking how AI can support current workflows, companies should ask how workflows themselves would look if intelligent agents were treated as integral participants rather than software tools.

Business processes would be redesigned around what agents can do. Decision-making systems would be built to allow agents to access knowledge, retrieve data, use software tools and complete tasks autonomously. Organizational structures and commercial models would also evolve to accommodate this new way of operating.

The result is not simply greater efficiency but a fundamentally different type of enterprise.

Why leadership matters

Such a transformation cannot emerge organically from isolated experiments within individual departments. Redesigning an organization around AI is ultimately a leadership challenge rather than a technology project.

Questions such as how work should be organized, where decision-making authority should sit, how humans and AI agents should collaborate and what new business models become possible are strategic issues that extend well beyond the remit of an IT department.

As a result, senior executives—and particularly CEOs—must take ownership of AI transformation.

Their role is no longer limited to approving technology investments. Instead, they must become architects of AI-native organizations, capable of redesigning workflows, organizational structures and business models from the top down.

The companies that move fastest may therefore be those whose leaders view AI not as another digital initiative but as the foundation for rebuilding the business itself.

Lessons from previous technological revolutions

History suggests that transformative technologies rarely achieve their full impact through incremental adoption alone. The spread of electricity offers a useful comparison.

Although electric power became commercially available in the late nineteenth century, factories initially continued to organize production around layouts developed for steam-powered machinery. Simply replacing steam engines with electric motors delivered only modest improvements.

It was only when manufacturers fundamentally redesigned factories around electricity—most famously through Henry Ford’s moving assembly line—that the technology’s full economic potential became apparent.

The same pattern can be seen more recently in the mobile internet revolution.

Companies such as Uber and Didi did not simply adapt existing taxi businesses to smartphones. Their business models depended on capabilities that only smartphones could provide, including continuous location tracking and real-time coordination between drivers and passengers.

Likewise, TikTok and Douyin were built around the smartphone’s camera, connectivity and always-on mobile usage rather than treating mobile devices as another channel for distributing traditional media.

In each case, competitive advantage came from designing entirely new operating models around a new technological foundation. AI represents a similar inflection point.

Organizations that merely incorporate AI into existing systems may realize incremental productivity gains. Those that redesign their businesses around intelligent agents could reshape entire industries.

Building organizations around intelligent agents

Thinking about agents as software tools may also underestimate their long-term role within organizations. Instead, they should be viewed as a new category of employee.

Like any employee, agents require knowledge, access to relevant information, appropriate permissions, clearly defined objectives and continuous performance feedback. Without these elements, they cannot improve or take on greater responsibility. This highlights an important shift in mindset.

Organizations should not expect agents to generate value immediately after deployment. Instead, they must invest in developing the knowledge, data and operational context that allow agents to perform effectively.

First develop the agent, then let the agent develop the business. Equally important is ensuring that agents operate within measurable feedback loops.

An inventory management agent, for example, should not simply recommend stock levels. It must also receive clear feedback on whether its decisions reduced shortages, lowered storage costs or improved resource allocation. Without measurable outcomes, agents cannot continuously refine their decision-making.

This closed-loop learning process distinguishes agentic AI from earlier generations of business software.

From individual agents to intelligent organizations

The implications become even greater when multiple agents begin working together. Rather than deploying isolated AI assistants for individual tasks, companies may eventually coordinate networks of specialized agents responsible for different aspects of the business—from procurement and logistics to finance, compliance and customer service.

Over time, these interconnected systems could evolve into an organization’s primary decision-making layer, continuously learning from operational data while coordinating increasingly complex activities across departments.

Humans would remain essential, but their roles would change.

Instead of overseeing every routine process, employees would increasingly focus on judgment, creativity, exception handling and strategic decision-making, while agents assume responsibility for repetitive analytical and operational tasks.

This does not imply the disappearance of human work. Instead, it suggests a new form of collaboration in which people and AI systems contribute different strengths within the same organization.

A new source of competitive advantage

Ultimately, AI’s second phase is not fundamentally about models, computing power or even agents themselves. Those technologies will become increasingly accessible across industries. Nor will competitive advantage come simply from possessing more data.

Instead, lasting advantage will depend on how effectively companies integrate intelligent agents with their own operational knowledge, proprietary data and industry expertise.

Organizations that successfully combine these assets will be able to build AI-native business models that competitors cannot easily replicate.

The challenge therefore extends well beyond technology adoption. It requires companies to rethink how work is organized, how decisions are made and how value is created.

Just as electricity transformed manufacturing and the internet reshaped information, intelligent agents have the potential to become a new layer of business infrastructure underpinning entire industries.

For business leaders, the strategic question is no longer whether AI will change their organizations. It is whether they will continue adapting existing systems—or use AI as the foundation for building entirely new ones.