Albert Bourla kept turning one question over in his mind: why hadn't the company's use of AI produced more impressive results. "Every night before falling asleep, I would review which decisions were right and which went wrong, and that feeling was uncomfortable." Last month, on the eve of attending the annual Concordia Summit in New York, the Pfizer (PFE) CEO reflected on the two years since the company rolled out generative AI across its entire organization. "I had expected faster, disruptive progress. AI has not been slow, but it is far from the pace I wanted, and that left me deeply frustrated."

Admittedly, AI has already sharply reduced the time required for some business processes, including administrative work such as data reporting and clinical trial planning. He said Pfizer (PFE) will launch 20 Phase 3 clinical trials this year, double last year's scale, covering trials for obesity and oncology treatments. Even so, these late-night reflections ultimately led him to overturn some of his earlier decisions.

He explained that he initially believed AI adoption had to be driven centrally by headquarters and led by technology experts in order to achieve substantive change. But later he realized that AI-related skills are becoming commoditized, and that the more critical step is to hand decision-making power to business-domain experts. Who is better suited than the head of manufacturing to advance the digital transformation of production? The same logic applies to R&D, marketing, and other fields.

Several months ago, Pfizer (PFE) switched to a distributed AI adoption model: leaders of each business domain decide on AI-related projects on their own and are held accountable for the results of transforming workflows and business operations, while a central team remains in place to provide supporting resources. He said that since then, internal resistance to AI rollout has eased somewhat, and the pace of change has clearly accelerated.

Bourla also took steps to prevent innovative ideas from being sealed off in silos. He said that in the AI era, he streamlined the executive team while expanding each executive's scope of responsibility. This more agile senior team now meets far more frequently to assess various real-world situations, and it has also become a communication platform for top leadership to understand how projects are actually running.

He added further: "Looking deeper into the organization, we have established cross-departmental groups we call 'clans.' Take engineers, for example: there are engineers in R&D, engineers in manufacturing, and engineers in digital. They belong to their respective departments, but they share common professional interests, and within these groups there can be full cross-boundary exchange and collision of ideas."

Although individual employees have already felt productivity gains, how to maximize the value of AI still has no standard answer for many companies. Bourla also did not claim to have found a universal solution. "I am not sure what the standard answer is, but I have made my own choices, and I hope these choices are the right ones."