Johnson & Johnson cuts 9,000 AI projects down to its highest-value use cases after a year of experimentation
Apr 18, 2025
Key Points
- Johnson & Johnson cuts its generative AI portfolio from nearly 9,000 use cases to a focused set after finding that 10-15% of projects drive 80% of the value.
- The company prioritizes four high-impact applications: a sales rep co-pilot for healthcare professionals, drug discovery optimization, supply chain risk management, and an internal HR chatbot.
- The shift signals that most enterprise AI spending may not survive scrutiny—what sticks is information retrieval, copywriting assistance, and narrow domain applications rather than broad transformation narratives.
Summary
Johnson & Johnson is cutting its generative AI portfolio sharply after a year of broad experimentation. The company went from pursuing nearly 9,000 individual use cases down to a focused set of highest-value applications. CIO Jim Swanson found that only 10 to 15% of use cases were driving 80% of the value, prompting the shift from what he called a "thousand flowers" approach to disciplined prioritization.
The company started with a centralized governance board that allowed use case ideas to germinate across divisions. That structure generated sprawl. Employees proposed thousands of projects, from transcribing paper receipts to broad AI experimentation across sales, operations, and support functions. Managers pursuing promotion saw AI as a way to signal modernization. Executives needed an AI transformation story for Wall Street.
J&J is now focusing on four categories. A sales rep co-pilot is the closest to production. It coaches representatives on how to engage healthcare professionals about new treatments. The tool started in the innovative medicine business, covering oncology and other therapeutic areas, and is expanding to the medtech segment, which sells robotics and hardware like hip replacements. The mechanism retrieves information from internal knowledge bases and helps with copywriting when a sales rep receives an inbound question from a buyer. Drug discovery is the second high-value use case, where generative AI helps researchers identify the optimal moment to add a solvent to turn a liquid molecule into a solid. Supply chain risk management is the third priority, relevant given current shortages and trade friction. An internal HR chatbot that answers policy and benefits questions rounds out the set, handling an estimated 10 million employee interactions annually with the services team.
The signal matters for vendors and enterprises alike. Consulting firms like McKinsey sold broad transformation narratives into large corporations. The Pareto result suggests most AI spend in large enterprises may not stick. What survives appears to be information retrieval at scale, copywriting assistance, and narrow domain applications like molecular chemistry. Generic internal chatbots and employee-facing tools, while easy to deploy, ranked lower once J&J measured actual impact.