If you find value in content like this, sign up for my Talent Edge Weekly newsletter.
This article presents findings from a field experiment conducted by the BCG Henderson Institute, Boston University, and OpenAI’s Economic Impacts team, exploring how GenAI can assist workers in completing tasks beyond their current capabilities. Over 480 BCG consultants participated, completing three tasks requiring data science skills they typically lacked: writing Python code to merge and clean datasets, building a predictive model for sports investing, and validating and correcting statistical outputs from ChatGPT. While GenAI couldn’t fully automate these tasks, it provided assistance. Only half of the participants had access to GenAI tools, and their performance was compared to a control group of 44 data scientists who completed the same tasks without GenAI. A key finding showed that GenAI significantly expanded participants’ coding and statistical abilities, with those using it achieving 86% of the data scientists’ benchmark, compared to just 29% for those without it. Despite the limitations of the research, such as a small sample size, the study provides insights into the tasks GenAI can assist with. With this as the backdrop, I am resharing two resources: A Deloitte AI Institute report, which includes a framework for determining which work tasks are best performed by a) AI, b) humans with the help of AI, and c) humans by themselves. An Accenture report, which includes a framework for categorizing four types of impacts that AI can have on jobs and work tasks.
