General-Purpose Technologies, such as AI, must cause the enterprise to rethink and redesign its workflows, organizational structures, and business models. Most information technologies targeting the enterprise, e.g., business intelligence, refine and optimize existing processes rather than forcing their complete redesign. Failing to distinguish between a true General-Purpose Technology and a standard information technology leads CEOs to the dangerous trap of considering AI committees and Centers of Excellence as the appropriate bodies to introduce AI to their corporations.
Introduction
A small number of information technologies require the enterprise to transform and redesign its workflows, organizational structures, and business models. We call them General-Purpose Technologies, or GPTs. Today’s AI is an enterprise GPT. Failing to understand this about AI causes CEOs to respond in a limited way to the AI Stakeholder Squeeze. The AI Stakeholder Squeeze results from the CEO Squeeze and the Employee Squeeze. The forces causing them are shown in the associated graphic. By failing to understand AI’s importance as a GPT, CEOs miss the opportunities to radically improve their corporation’s performance and change the trajectory of their enterprise.
Over the years, we created many new information technologies. However, only a very small number are General-Purpose Technologies. For example, consider machine electrification. It resulted in the complete physical and operational redesign of manufacturing. Factories were redesigned around the logical flow of materials. Productivity increased, costs dropped, and work was specialized by task. The Internet, another GPT, led the enterprises that embraced it properly to dismantle geographically constrained supply chains and fundamentally redesign their business models around real-time digital distribution and e-commerce.
Most information technologies that are targeting the enterprise, e.g., Business Intelligence, are not GPTs. As such, they do not require the radical, cross-functional transformations that General Purpose Technologies (GPTs) demand. Instead, they refine and optimize existing processes.
Committees and Centers of Excellence
When innovators and early adopters adopt a new information technology, the reflex of legacy enterprises in the early and late majorities has remained the same: set up a centralized Committee and a Center of Excellence (CoE). The Committee serves as the strategic governance layer, responsible for developing enterprise-wide policies, managing major vendor relationships, and establishing compliance and security guardrails. It operates as a corporate gatekeeper, dictating top-down adoption rules. The CoE serves as the centralized technical execution arm for the specific technology, responsible for evaluating new tools, building proofs of concept, and establishing enterprise-wide best practices. Most often, it operates as an isolated technical lab, lacking domain knowledge and the P&L authority required to fundamentally change the business.
In the mid-to-late 1990s, Business Intelligence (BI) and the commercial internet entered the corporate consciousness. Most CEOs treated them identically. They set up a “Data Warehouse Committee” to manage BI and a “Web Committee” to manage the Internet. In both cases, their goal was to find ways to bolt new capabilities onto existing business processes.
BI enables employees to extract insights from organizational data that inform business strategies and operations. The companies that decided to treat the Internet as a normal technology rather than a GPT approached it simply as an alternative marketing channel. They built websites, but they did not redesign their physical supply chains, their inventory management, or their customer engagement models. Even though they saw incremental improvements, such as lower marketing costs, they completely missed the trajectory-changing reality of e-commerce, global digital distribution, and other transformational opportunities the Internet offers. As a result of this misreading, many of these corporations were subsequently decimated by the digital natives who understood that the Internet was not an IT project, but a technology that enabled a new way of conducting business.
The AI Committee and the Center of Excellence
Generative AI has taken the world by storm. Many corporations are responding by setting up AI Committees and AI Centers of Excellence. However, these efforts will not alleviate either the CEO Squeeze or the Employee Squeeze. At best, they will be able to enhance some business processes with AI, but will not transform the company into an AI-first enterprise. At worst, they will exacerbate the Stakeholder Squeeze because legacy organizations tend to sabotage such transformation. As Tom Davenport notes, while early IT systems like ERPs often calcified workflows into rigid structures, generative AI finally provides the capability to enact the radical, flexible redesign of work that executives have long sought.
Innovation theorist Carlota Perez has shown that every major technological revolution follows a predictable lifecycle. It begins with an Installation Period, an era of hype, venture funding, and isolated experimentation. During this period, corporations force the new technology into old institutional structures. To capture real economic value, the enterprise must enter the Deployment Period, crossing a “Turning Point.” This requires establishing a new institutional framework. In her recent work, Perez notes that the Digital Era’s Turning Point has been the longest and most difficult in history because it is the first revolution that mechanizes mental work.
Today’s legacy enterprise is trapped deep within this exact friction. AI Committees and Centers of Excellence are artifacts of the Installation Period. They are exploratory bodies. They evaluate tools, but they lack the authority to mandate what modern operations truly require: end-to-end workflow redesign.
Three Systemic Failures
The efforts of AI Committees and CoEs result in three systemic failures:
- The Illusion of Action. The AI Committee provides a shield to the CEO. When boards and shareholders exert pressure for immediate use of AI, the executive can point to the committee and the dozens of pilots generated by the CoE as proof of execution. However, this creates merely the illusion of action. Committees may recommend adopting third-party AI solutions, but have no authority to make organizational, business process, or business model changes.
Because CoEs lack the budgetary authority and P&L ownership to force actions, such as converting the pilot into a deployable system and deploying it across one or more organizations, they default to building proofs-of-concept (POCs) relying on small, siloed data. These pilots fail to accurately measure the Productivity J-Curve and commit the enterprise to a Pilot Purgatory. - Institutional Knowledge Gaps. The AI committee members rarely understand the corporation’s operating model. They are often disconnected from the strategic bottlenecks present in that model. Moreover, the technical fluency that their members frequently lack obstructs their ability to determine which bottlenecks to alleviate with AI. AI technology providers are admitting that their products’ capabilities outpace the abilities of corporations to understand and implement them.
CoEs face the opposite deficit. While their staff may understand the intricacies of generative AI, they lack the deep domain expertise and tacit knowledge of the business workflows they are trying to automate and the associated organizational structures that should be modified. As a result, the enterprise defaults to pilots that fail to deliver structural transformation. - The Reaction of the Corporate Immune System. The threats posed by AI, ranging from fear of impending layoffs to the chaotic burden of learning new tools and altering legacy processes, inevitably activate the organization’s “corporate antibodies.” Contrary to the assumption that resistance occurs only on the company’s front lines, the production of these antibodies begins within the AI Committee. We have seen several examples of delays and rejections under the guise of governance and risk management. Employees become the integration layer between broken legacy processes and the CoE’s POCs. They experience cognitive exhaustion, a phenomenon recently termed “AI Brain Fry.” Combined with the anxiety about their employment future, initial skepticism about AI becomes active operational resistance.
The Way Forward
To avoid the AI Stakeholder Squeeze and transition into an AI-first company, legacy enterprises must view AI as a GPT and abandon their committee- and CoE-led approach. Instead, they must set up Ambidextrous Pods, flat organizations that act as mini startups, merge business logic, technology, HR, and legal, and are led by a P&L owner. The pod structure is necessary regardless of whether the corporation has autonomous business units, e.g., Johnson & Johnson, or functional units, e.g., Apple. In parallel, the corporation must also set up its AI Factory to bring together its AI technical resources at the scale necessary to achieve its technology goals.
The Ambidextrous Pod flips the paradigm of the Innovation Labs that corporations used in the past. It rebuilds the company’s existing core workflows and creates new ones as appropriate. The pod is called “ambidextrous” because it creates a bridge between exploration and exploitation. Its deliverable is a fully codified workflow blueprint, complete with updated HR job descriptions and legal guardrails. The rest of the company exploits and manages the proven workflows, using the AI Factory’s power.
Moderna’s CEO mandated the reinvention of the company’s end-to-end mRNA design workflow. He created a team of the company’s data scientists, biologists, HR, and regulatory experts. He merged HR with IT, giving Tracey Franklin the title of chief people and technology officer. Her mandate was to determine which roles would be performed by humans and which by agents. These were the first steps on the company’s path to becoming an AI-first enterprise. Subsequently, the company partnered with OpenAI and developed its own generative AI models. It created an AI Factory to accelerate drug development. It launched a broad employee re-skilling program.
The AI-motivated organizational restructuring undertaken by companies like Microsoft, Meta, and Amazon, Google’s business model evolution, and OpenAI’s business model experiments should also help us understand how adopting AI as a General-Purpose Technology leads to organizational, business process, and business model changes.
Surviving the AI Stakeholder Squeeze requires legacy enterprises to recognize that AI is a General-Purpose Technology. They must then embrace the opportunities it offers. AI Committees offer an illusion of action. Centers of Excellence generate disconnected pilot projects. Together, they lead to sub-optimal enhancements and prolong the AI Stakeholder Squeeze. The Ambidextrous Pod, with the AI Factory, enables the enterprise to shift from experimenting with AI to becoming AI-first.
My podcast with Ben Lorica, where we discuss AI Committees, Centers of Excellence, and Ambidextrous Pods