How to Make Enterprise Gen AI Work

"If your organization is still relying on ad hoc employee experimentation with gen AI, it's time to shift gears. While experiments like using Claude to draft emails or ChatGPT to brainstorm can yield learning and productivity benefits on an individual or unit level, they are typically unstructured and unmeasured and rarely yield large-scale results. (We believe that this is part of why companies aren't yet seeing bottom-line impact from AI investments.)"

"Daniel J. Politzer is a research associate at Stanford University in the Management Science and Engineering Department, and the founder of Alerce Advisors, a consultancy that offers AI advisory services and fractional management. Thomas H. Davenport is the President's Distinguished Professor of Information Technology and faculty director of the Metropoulos Institute for Technology and Entrepreneurship at Babson College, a visiting scholar at the MIT Initiative on the Digital Economy, and a senior adviser to Deloitte's Chief Data and Analytics Officer Program."

Organizations relying on ad hoc employee experimentation with generative AI often see only localized gains. Experiments such as using Claude to draft emails or ChatGPT to brainstorm can deliver learning and productivity improvements at individual or unit levels. Those experiments are typically unstructured and unmeasured and rarely produce large-scale results across the enterprise. The lack of structured deployment, governance, and measurement contributes to limited bottom-line impact from AI investments. A shift away from informal experimentation toward coordinated, measured, and scalable AI programs is necessary to capture enterprise-level productivity and financial benefits.

Read at Harvard Business Review