Proof of Life: Keeping Your Family in Your Pocket with Everyday Connection, No Matter the Distance;Losing What Cannot Be Written Down: Mandatory AI Adoption, Tacit Knowledge Loss, and Organizational Vulnerability 10 views

During my summer 2025 internship at Becton Dickinson (BD), I built an news automation system using Power Automate and AI Builder. That system replaced a $17,000 annual subscription and changed how analysts did their jobs. Building it raised a question: What happens to workers when AI takes over the judgment they used to build? My technical project and STS research are the same case, examined from two different directions. The technical portion of my thesis produced a fully functional automation workflow. The system pulled healthcare news from RSS feeds, extracted relevant content using AI, and delivered daily summaries to BD's commercial and marketing teams. Before the system existed, analysts read hundreds of articles per month. They applied their own knowledge to decide what was relevant and what each team needed to know. After the system launched, they reviewed AI selections instead. The system ran on a set schedule, every weekday at 8am, without analyst input. BD saved $17,000. By every standard engineering measure, the project was a success. In my STS research, I argue that mandatory AI adoption erodes the foundational skills and tacit knowledge that make employees valuable, ultimately making them more replaceable and their organizations more vulnerable when systems fail. Tacit knowledge is the judgment workers build through repeated practice. It lives in the doing, not in a document. Polanyi (1966) described it as knowing more than we can tell. When AI takes over a task, the worker stops practicing it. The skill fades without anyone noticing. I applied the Butollo et al. (2025) framework to the BD case to trace what the automation removed across three categories: skills, creativity, and authenticity. Analysts lost the ability to filter without AI, the cross-market judgment that filtering built, and the ability to explain why a specific article mattered to a specific team. These losses compound. A reviewer of AI output is not the same as a producer of original analysis. The $17,000 saved appeared on a balance sheet. The tacit knowledge lost did not appear anywhere. Considering the technical and organizational dimensions together changed what I saw. Measured by engineering standards, the BD automation was a success. Measured by STS standards, it created a where environment where the conditions for maintaining expertise were removed by design. STS gave me a way to ask not just whether the system worked, but what it did to the people who used to do that work. Engineering decisions are not made in isolation. They happen inside organizations, they change what workers practice, and they determine who retains the judgment to step in when a system fails. That is what STS made visible to me.

BS (Bachelor of Science)

AI; automation; tech

All rights reserved by the author (no additional license for public reuse)

2026-05-08